A simple stochastic model for complex coextinctions in mutualistic networks: robustness decreases with connectance

Abstract

Understanding and predicting species extinctions and coextinctions is a major goal of ecological research in the face of a biodiversity crisis. Typically, models based on network topology are used to simulate coextinctions in mutualistic networks. However, such topological models neglect two key biological features of species interactions: variation in the intrinsic dependence of species on the mutualism, and variation in the relative importance of each interacting partner. By incorporating both types of variation, we developed a stochastic coextinction model capable of simulating extinction cascades far more complex than those observed in previous topological models. Using a set of empirical mutualistic networks, we show that the traditional topological model may either underestimate or overestimate the number and likelihood of coextinctions, depending on the intrinsic dependence of species on the mutualism. More importantly, contrary to topological models, our stochastic model predicts extinction cascades to be more likely in highly connected mutualistic communities.

Adaptive rewiring aggravates the effects of species loss in ecosystems

“…For instance, theoretical work suggests that extinction-induced rewiring in mutualistic networks like plant-pollinator systems is positive both at the species and the network level 19 (see also ref. 43). In a purely mutualistic system overexploitation is not an issue.…”

Protecting an Ecosystem Service

“…While plant-pollinator interactions are dynamic, the topology or structure of the network is 573 thought to confer a degree of stability or robustness upon the community (Bascompte et al, 574 2003;Kaiser-Bunbury et al, 2010;Memmott et al, 2004;Olesen et al, 2007;Ramos-575 Jiliberto et al, 2012;Thebault and Fontaine, 2010;Tylianakis et al, 2010;Valdovinos et al, 576 2013;Vieira and Almeida-Neto, 2015). There is, however, much debate over the extent that 577 different properties of network architecture confer stability or robustness to species loss 578 (Rohr et al, 2014).…”

“…For example, species that are highly abundant and well connected to 579 many other species in the network, typically generalists, may increase overall network 580 robustness . However, other evidence suggests that increased network 581 connectance (the proportion of possible links between species) may lead to a greater chance 582 of extinction cascades following species loss (Vieira et al, 2015), and the structuring of these 583 connections appears important with predictions that increased network nestedness imparts 584 stability (Allesina & Tang 2012;James et al, 2012). Moreover, behavioural plasticity means 585 that a pollinator species can potentially 'rewire' the network by switching to alternate plant 586 species following extirpation of a partner pollinator species, thereby maintaining the overall 587 community cohesion (Figure 2; Kaiser-Bunbury et al, 2010;Ramos-Jiliberto et al, 2012;588 Valdovinos et al, 2013).…”

“…For example, in areas experiencing significant 601 transitions to more homogenous landscapes such as intensive arable monocultures, short 602 tongued generalist species of insect pollinators may be less affected than long tongued 603 specialists due to the concomitant loss of habitat(s) that maintain specific plants (Figure 2; 604 Goulson et al, 2005). Simulation modelling of empirical data shows that if the most 605 connected species in plant-pollinator networks are lost, then this may trigger a cascade of 606 secondary extinctions of plants or pollinators (Kaiser-Bunbury et al, 2010;Memmott et al, 607 2004;Vieira and Almeida-Neto, 2015). The implication is that if environmental change 608 reaches a level sufficient to extirpate these highly linked species then there is a risk that the 609 whole plant-pollinator network could disassemble and jeopardise the continued delivery of 610 insect pollination in that ecosystem (Lever et al, 2014).…”

Extinction models of robustness for weighted ecological networks

“…One approach has been to use specific sequences of primary extinctions, such as ordered by traits of the nodes (Dunne, Williams & Martinez, 2002;Memmott, Waser and Price, 2004;Pocock, Evans & Memmott, 2012;and Santamaria et al, 2016). A different adaption, developed by Vieira and Almeida-Neto (2015), allowed co-extinction due to feedback between guilds (implemented stochastically, based upon interaction frequencies), so permitting the possibility of cascades of extinctions.…”