Network size, structure and mutualism dependence affect the propensity for plant–pollinator extinction cascades

Citation

Vanbergen AJ, Woodcock BA, Heard MS & Chapman DS (2017) Network size, structure and mutualism dependence affect the propensity for plant–pollinator extinction cascades. Functional Ecology, 31 (6), pp. 1285-1293. https://doi.org/10.1111/1365-2435.12823

Abstract
1. Pollinator network structure arising from the extent and strength of interspecific mutualistic interactions can promote species persistence and community robustness. However, environmental change may re‐organise network structure limiting capacity to absorb or resist shocks and increasing species extinctions. 2. We investigated if habitat disturbance and the level of mutualism dependence between species affected the robustness of insect–flower visitation networks Following a recently developed Stochastic Co‐extinction Model (SCM), we ran simulations to produce the number of extinction episodes (cascade degree), which we correlated with network structure in undisturbed and disturbed habitat. We also explicitly modelled whether a species’ intrinsic dependence on mutualism affected the propensity for extinction cascades in the network. 3. Habitat disturbance generated a gradient in network structure with those from disturbed sites being less connected, but more speciose and so larger. Controlling for network size (z‐score standardisation against the null model) revealed that disturbed networks had disproportionately low linkage density, high specialisation, fewer insect visitors per plant species (vulnerability) and lower nestedness (NODF). 4. This network structure gradient driven by disturbance increased and decreased different aspects of robustness to simulated plant extinction. Disturbance decreased the risk that an initial insect extinction would follow a plant species loss. Although, this effect disappeared when network size and connectance were standardised, suggesting the lower connectance of disturbed networks increased robustness to an initial secondary extinction. 5. However, if a secondary extinction occurred then networks from disturbed habitat were more prone to large co‐extinction cascades, likely resulting from a greater chance of extinction in these larger, speciose networks. Conversely, when species mutualism dependency was explicit in the SCM simulations the disturbed networks were disproportionately more robust to very large co‐extinction cascades, potentially caused by non‐random patterns of interaction between species differing in dependence on mutualism. 6. Our results showed disturbance altered the size and the distribution of interspecific interactions in the networks to affect their robustness to co‐extinction cascades. Controlling for effects due to network size and the interspecific variation in demographic dependence on mutualism can improve insight into properties conferring the structural robustness of networks to environmental changes.

Keywords
environmental disturbance; extinction cascade; mutualism; network structure; pollinator; stability; Stochastic Co-extinction Model; z-score

Journal
Functional Ecology: Volume 31, Issue 6