Plant-hummingbird pollination networks exhibit limited rewiring after experimental removal of a locally abundant plant species.

Mutualistic relationships, such as those between plants and pollinators, may be vulnerable to the local extinctions predicted under global environmental change. However, network theory predicts that plant-pollinator networks can withstand species loss if pollinators switch to alternative floral resources (rewiring). Whether rewiring occurs following species loss in natural communities is poorly known because replicated species exclusions are difficult to implement at appropriate spatial scales. We experimentally removed a hummingbird-pollinated plant, Heliconia tortuosa, from within tropical forest fragments to investigate how hummingbirds respond to temporary loss of an abundant resource. Under the rewiring hypothesis, we expected that behavioural flexibility would allow hummingbirds to use alternative resources, leading to decreased ecological specialization and reorganization of the network structure (i.e. pairwise interactions). Alternatively, morphological or behavioural constraints-such as trait-matching or interspecific competition-might limit the extent to which hummingbirds alter their foraging behaviour. We employed a replicated Before-After-Control-Impact experimental design and quantified plant-hummingbird interactions using two parallel sampling methods: pollen collected from individual hummingbirds ('pollen networks', created from >300 pollen samples) and observations of hummingbirds visiting focal plants ('camera networks', created from >19,000 observation hours). To assess the extent of rewiring, we quantified ecological specialization at the individual, species and network levels and examined interaction turnover (i.e. gain/loss of pairwise interactions). H. tortuosa removal caused some reorganization of pairwise interactions but did not prompt large changes in specialization, despite the large magnitude of our manipulation (on average, >100 inflorescences removed in exclusion areas of >1 ha). Although some individual hummingbirds sampled through time showed modest increases in niche breadth following Heliconia removal (relative to birds that did not experience resource loss), these changes were not reflected in species- and network-level specialization metrics. Our results suggest that, at least over short time-scales, animals may not necessarily shift to alternative resources after losing an abundant food resource-even in species thought to be highly opportunistic foragers, such as hummingbirds. Given that rewiring contributes to theoretical predictions of network stability, future studies should investigate why pollinators might not expand their diets after a local resource extinction.