In most ecological networks investigated to date, a few species tend to have disproportionately more interactions than the rest of the species. Some have argued that we should consider these generalists as keystone species, because static or dynamic simulation models suggested that their removal would have catastrophic consequences on the whole community. This could have implications for the way we harvest trees, for example, as “keystone trees” in ectomycorrhizal networks have been found to be the largest and oldest ones, which are preferentially targeted at harvest. However, what is unclear is whether these apparently keystone taxa can easily be replaced after a disturbance: this might be the case if we found that keystone taxa fluctuate spatially in mycorrhizal networks. Here, we tested this hypothesis using 18 arbuscular mycorrhizal networks in 3 different habitat types (grassland, early regenerating forest and late-successional forest).
Results/Conclusions
We show that although these networks were structured around highly generalist taxa, the level of generalism of a given species was highly fluctuating across space, and so were preferential interactions. Those results thus suggest that (1) being a generalist locally is an insufficient criterion to be considered a keystone species in an ecological network, and (2) preferential interactions can arise from local and contemporary partner selection without involving coevolutionary events. We argue that such flexibility in interaction patterns (either the number of interactions or the identity of the partners) urgently needs to be characterized in other systems to guide modeling studies on ecological networks. We also discuss the relevance of seeing the assembly of ecological networks as an optimization process.