Can floral preference of insect visitors be applied to species identification of the Japanese Arisaema (Araceae)?

Species specificity in plant–insect interactions can be new diagnostic traits as an objective criterion for species delimitation and identification. To evaluate the taxonomic reliability of floral preference by specialist pollinator species, we used Japanese Arisaema species, which selectively attract different fungus gnat pollinators, as a model system. We comprehensively investigated (1) external morphology, (2) floral visitor assemblage, and (3) population genetic structure of two unidentified Arisaema populations and three candidate Arisaema species (A. aequinoctiale, A. minus, and A. nambae), and compared the performance of species identification between external morphology (conventional method) and floral visitor assemblages (alternative method). First, we selected “correct answer” for species identification of the unidentified populations from the three candidates based on 14 morphological traits, flower visitation frequency of ten major dipteran visitors, and population genetic structure. Then, we compared the identification accuracy among canonical discriminant functions consisting of morphological traits (model M), insect visitation frequency (model I), and both (model M + I). The morphological, floral visitor, and genetic similarities between populations consistently suggested that the unidentified populations belonged to A. minus. The identification accuracy of model I was comparable to that of model M, and the combination of morphological and visitor data (i.e. model M + I) maximized model performance. Although the collection of analyzable datasets of fungus gnat assemblages may impose considerable labor on plant taxonomists, this new taxonomic trait can be an alternative tool for recently diversified species complexes of Japanese Arisaema.