Species abundance distributions and differences in plant attractiveness influence plant-pollinator network disassembly
Many studies show that ecological networks are robust to the removal of specialists, but the loss of generalists causes secondary extinctions of generalist-dependent species. However, the skewed species abundance distribution (SAD) of ecological communities means that apparent differences in species’ community disassembly roles may be driven by differential abundances and attractiveness, rather than by generalization per se. To explore this issue, we established a field experiment consisting of 102 monospecific plots of 17 plant species, with each plot standardized for relative abundance of plants. We collected bees visiting flowers to create an empirical network based on an even SAD for the plants. We then re-sampled from this even community to simulate a skewed SAD similar to natural communities. We compared secondary extinction rate from even or skewed-SAD plant communities by simulating plant extinctions either randomly, or in ascending or descending order of plants’ partner richness (network ‘degree’) or partner abundance (network ‘strength’). Because of our experimental design, disassembly patterns from the even-SAD network reflect differences in plant attractiveness to pollinators, but not differences in plant abundance that drive either actual interaction probability or sampling probability. Disassembly patterns from the skewed-SAD network thus reflect differences in plant abundance and attractiveness combined.
In our four-year field experiment, we captured 8,054 specimens of 104 bee species. We found that in both even and skewed networks, secondary extinction patterns deviated from random, but the extent of deviation and the relative effects of losing species according to degree versus strength depended on the plant SAD. Systematic plant loss deviated more from random in the skewed than in the even-SAD network, suggesting that both plant relative abundance and attractiveness influence disassembly as it is measured from real-world, skewed-SAD networks. However, in our skewed-SAD network, the effects of degree and strength on disassembly were similar, but when the SAD was even the effects of degree-based loss were greater (deviated more from random) than the effects of strength-based loss. This is likely due to the higher richness of rare pollinators in the even SAD network that were not detected when the SAD was skewed, and the tighter correlation between plant degree and strength that resulted from subsampling to create a skewed SAD. Thus in the skewed network the plant species with the most interactions also had the most partner species, and rare species that contribute to the robustness of pollinator richness are not detected.