How do species abundance distributions influence plant -- pollinator networks?.

摘要

Plant -- pollinator networks provide a tool for understanding patterns in ecological communities. However, apparent network patterns observed in real-world systems may be influenced by a skewed species abundance distribution (SAD), which obscures the differences between biology- and chance-based drivers of network patterns. This dissertation addresses that issue using a four-year experimental network in which the relative abundance of 17 plant species is even, and from which 8,054 specimens of 104 bee species were collected. Patterns observed in that even-SAD field experiment were compared with null models, to answer the following questions: (1) Can the apparent nested pattern of plant -- pollinator networks inform pollinator conservation? (2) Does the number of plant species partners and the number of plant individual partners equally influence the effect of losing plant species from networks, and does this depend on the plant SAD? (3) Is apparent year-to-year variation in interactions driven by biology or by chance, and does this depend on the bee species abundance? This research shows that (1) the abundance of rare and common bee species were uncorrelated across plant species, suggesting that their divergent preferences would require a targeted, not a general, choice of plant species for conservation. However, their preferences were positively correlated when the even plant SAD data were subsampled to simulate a skewed plant SAD network, but only when the most attractive plant species was simulated to be the most abundant. (2) Plant species' differential importance in the process of network disassembly is driven more by number of partner species than partner individuals, but only when the plant SAD is even. Almost no difference between species- and individual- based plant species loss was observed in the skewed network, because subsampling to skew the SAD removed many rare species and caused partner species and individuals to become more tightly correlated. (3) Expected annual dissimilarity in bee species' preferences decreases with bee abundance, suggesting an increased ability to detect non- random changes for common bees. The common bees differed significantly from that null expectation, suggesting that variation driven by biology, not by chance, is only detectable for common species.