Identifying a Ranking of Plant Preferences for a Pollinator

Pollinators are an integral part of agriculture and the ecosystem. However, due to changing land use, populations of wild pollinators are decreasing and plant distributions are changing all around the world. To understand how plant-pollinator networks will adapt over time, we would like to understand how pollinators choose flowers to visit. We will model a pollinator’s interaction with plant species in two ways: first using a probabilistic multinomial approach to fit a preference score to each plant and second to explain our findings from the multinomial model using the traits of the flowers themselves. Our findings show that a model with preferences performs better than a model which does not have preferences. While this model shows potential in finding plant preferences, it does not fully explain the distribution of plant-pollinator interactions. To try to explain the interactions more fully, we incorporated the traits of the plants into the score of the plant. We found that the traits do have some effect on the score of the plant, but again do not fully explain the interactions in this particular model.
Key Words: Plant-pollinator networks, machine learning, computational ecology, pollinator behavior