“Pollination syndrome” is a classical concept depicting suites of floral traits as outcomes of specialized co-evolutionary process between plants and their pollinators, but has been discredited its conceptual utility as a tool to explain whole array of flower-pollinator associations actually observed in the field. Because there are many exceptions or generalist flowers visited by various generalized and specialized pollinators, we developed a hierarchical Bayes model, by taking flower traits (petal color and flower morphology), pollinator's color vision, and environmental conditions (sunshine, cover degree and index of number of pollinator in survey area) into account, in order to quantitatively predict pollinator assemblages building up on particular flowers or floral assemblages.

First, we predicted the floral traits selected by 5 types of pollinator functional groups (bee, wasp, fly, hoverfly and butterfly) with pollinator visitation frequency to 50 plant species. Resultant 5 models with appropriate parameters were combined to create a meta-model to predict proportions of each pollinator to all visitors. We are currently seeking procedures testing how precisely our model predict the observed data, in hope of obtaining mechanistic model to predict relative contribution of floral traits to flower visitation rates by respective pollinators under various environmental conditions.