Figure 1. Importance of different pollination processes for plant pollination success as a function of floral abundance, as determined by the mathematical model. Comparison of pollination success expected for low, medium and high values of A) pollinator abundance, B) pollinator specialization, C) pollen removal and D) carryover at variable floral abundance (see Table 1 for parameter values). E) Importance of the quantity component of pollination, represented by pollinator abundance, and the quality component, determined by pollinator specialization and carryover as a function of floral abundance (measured as the average between the importance of pollinator specialization and carryover). F) Hypothetical representation of the pollination systems offering the highest pollination success at different floral abundance. Specialization on pollinators with high pollen carryover and level of specialization is favoured at low floral abundance (hypothetical examples; hawkmoths and hummingbirds). At intermediate abundance, specialized pollination by more abundant pollinators (hypothetical example; bees) is favoured. At high floral abundance, most pollinators are not sufficiently abundant to remove most pollen grains and generalization becomes more advantageous. When floral abundance is too high for the pollinator community to remove most pollen grains, reliance on abiotic pollen vectors is expected.
Figure 2. Plant-pollinator networks resulting from simulated communities of different plant species and pollinator attributes. A, B) At intermediate average flower abundance (500 flowers), plant-pollinator networks formed with plant species of variable floral abundance result in variable levels of generalization among plant species while C, D) networks formed with plant species of the same floral abundance result in similar levels of generalization. E, F) Networks composed entirely of low-abundance plant species (100 flowers) result in high level of specialization. G, H) Networks composed entirely of high-abundance plant species (1000 flowers) result in widespread generalization. In A, C, E and G, the thickness of the links represents the number of visits of a pollinator to a plant species. In B, D, F and H, grey squares denote interaction between a plant and a pollinator and darker shades represent higher number of interactions.
Figure 3 . Effect of interspecific variation in abundance in simulated plant communities on A) variation in degree of generalization, B) number of shared partners, C) plant-pollinator network nestedness and D) connectance at different average flower abundances. High average flower abundance corresponds to 1000, intermediate high corresponds to 500, intermediate low corresponds to 250 and low corresponds to 100. Symbols and error bars represent mean and standard error.
Figure 4. Degree of generalization and attributes of the pollinators on which a plant species colonizing simulated plant communities evolved as a function of its floral abundance. The parameter values represent the average values of pollen carryover capacity, specialization and abundance of the pollinators on which the new colonist evolved and degree of floral generalization of the new colonist. The parameter values on the y-axis were normalized so that the minimum value corresponds to 0 and the maximum value corresponding to 1. The standard error of the mean parameter values among the 100 simulations is presented as the shaded area around the mean values.
Figure 5 . The three processes generating floral diversification according to the model. Following dispersal by a plant colonist (the red tubular flower species) from a community (represented in A) to new communities, shifts in pollination system can occur as a result of either B) change in the abundance of the new colonist, C) change in plant community composition (change in abundance of the other community members), or D) change in pollinator assemblage. C, E, G) Effect of different amounts of variation in C) abundance of the new colonist, E) community composition and G) pollinator assemblage between communities on variation in evolved pollination systems between species for simulated plant clades colonizing 20 new communities. Variation in evolved pollination systems was measured as the standard deviation of the average attributes values and degree of generalization between species of the plant clade. The parameter values on the y-axis were normalized so that the minimum value corresponds to 0 and the maximum value corresponding to 1. Panel E) do not show values of pollinator specialization (measured as the total flower abundance of all the species it pollinates) and panel G) do not show values of pollinator abundance because those parameters were purposely varied between communities and hence variability was expected for those parameters even in the absence of shift in pollination system. Illustration by Florence Jean and Sébastien Rivest.