Effects of topology on robustness in ecological bipartite networks

Abstract

High robustness of complex ecological systems in the face of species extinction has been hypothesized based on the redundancy in species. We explored how differences in network topology may affect robustness. Ecological bipartite networks used to be small, asymmetric and sparse matrices. We created synthetic networks to study the influence of the properties of network dimensions asymmetry, connectance and type of degree distribution on network robustness. We used two extinction strategies: node extinction and link extinction, and three extinction sequences differing in the order of species removal (least-to-most connected, random, most-to-least connected). We assessed robustness to extinction of simulated networks, which differed in one of the three topological features. Simulated networks indicated that robustness decreases when (a) extinction involved those nodes belonging to the most species-rich guild and (b) networks had lower connectance. We also compared simulated networks with different degree-distribution networks, and they showed important differences in robustness depending on the extinction scenario. In the link extinction strategy, the robustness of synthetic networks was clearly determined by the asymmetry in the network dimensions, while the variation in connectance produced negligible differences.

Removing interactions, rather than species, casts doubt on the high robustness of pollination networks

“…We applied the interaction extinction approach to robustness assessment recently proposed by Pastor et al () to 12 empirical binary ASP networks previously analyzed in Santamaría et al () using the traditional node (species) extinction approach (Memmott et al ). Although weighted networks would be preferable, only three weighted ASP networks were available.…”

“…Thirdly, based on the extinction plots in NBRA, an index of robustness, R 50 , has been proposed (Dunne et al ). In a similar way, we used the bipartite R 50 index of robustness introduced by Pastor et al () to facilitate the comparison of robustness between scenarios and networks. We refer to this index as R 50‐N‐N to indicate that it is based on the node–node extinction plot of IBRA.…”

“…Based on the previous available results obtained for empirical food webs (Gao et al , Allesina et al ) and simulated pollination networks (Pastor et al ), we expected the simulation of interaction extinctions to produce qualitatively different extinction patterns from those obtained using the node extinction approach. This would imply that the traditional approach to robustness assessment might lead to overly‐optimistic conclusions about the robustness of pollination networks (Gao et al , Allesina et al , Pastor et al ). Alternatively, if these two approaches produced similar results for empirical networks, assessing robustness by removing nodes would be appropriate and simpler.…”

“…In ecology, simulation of interaction removal has been performed for empirical food webs (Gao et al ) and the distinction between redundant and functional interactions is relevant for the robustness of food webs (Allesina et al ). Interaction extinction has only recently been expanded to simulated mutualistic networks by Pastor et al ().…”

Extinction models of robustness for weighted ecological networks

“…1b). Robustness is therefore a product both of structural heterogeneity of the network (eg Pastor et al, 2012) and of the method of producing extinction sequences; we explore each of these contributions to robustness in this paper.…”