Application of multivariate morphometrics to delimit three Japanese species of Arisaema sect. Pistillata (Araceae)

The Arisaema serratum group sensu stricto (Araceae) is a Japanese species assemblage characterized by a well-developed leaf rachis and a long pseudostem. It has been regarded as a complex including many species with few decisive traits. They were previously considered as a single species due to their wide intraspecific variation, but recent monographs have recognized about 13 species. In this group, A. maekawae and A. pseudoangustatum var. pseudoangustatum were recently divided from A. angustatum from the type localities. An earlier study found sympatric populations of the three species in the Chugoku region, far from the type localities, and distinguished them by qualitative observations, but there are no quantitative data to delimit the species boundaries objectively. Since high phenotypic plasticity hinders species delimitation in the A. serratum group s.s. using just a few quantitative traits, I investigated the morphological discontinuity among them using principal component analysis with eleven quantitative characters. In an ordination space including the three species, the morphological variation was roughly divided into A. maekawae and the others, and largely overlapped between A. angustatum and A. pseudoangustatum var. pseudoangustatum. However, the morphological variation clearly distinguished the two latter species when A. maekawae was excluded from the PCA. To conclude, the three Arisaema species appeared to be morphologically distinct. The strong reproductive isolation reported previously also ensures the species distinction according to the biological species concept. The discriminating traits between A. angustatum and the other two species partly differed from those in the type localities, suggesting that examination of intraspecific variation is needed between geographically isolated populations. This study showed that a combination of quantitative traits effectively supports tentative species identifications using qualitative traits and reinforces the usefulness of multivariate morphometrics for future systematic studies.