COS 113-6
Temporal variability predicts species importance in an ecological interaction network
How can we identify the importance of species in the complex network of interactions in natural ecosystems? Qualitative topological food webs and quantitative modeling of interaction webs have so far been of limited use. We propose that simple metrics arising from integrating temporal community dynamics with interaction web structure might provide a roadmap for identifying important species. Specifically, we hypothesize that the relative variability of a species is inversely related to its impact on the community, because highly variable species with many strong impacts would likely cause the network to collapse. We tested this hypothesis by using experiments and observations from the Tatoosh Island, WA, rocky intertidal zone to generate a species importance index and food web for 102 species, and by using a 20-year data series of community dynamics to document variability (CV) for each species. We also generated multiple feasible parameterizations of a 50-species food web model, characterized the variability of species in stochastic simulations, and assessed the impacts of removing species from the model on community dissimilarity (Jaccard distance) to the equilibrium value. We then explored whether relationships existed between species importance, temporal variability, and network metrics.
Results/Conclusions
The temporal CV was strongly and negatively associated with empirically estimated species importance, generating 60% correct assignments among the four importance classes (25% random expectation; p<0.001). Predictive ability was significantly improved (to 67%) by including a positive relationship with the number of species to which a focal species was linked (degree; p=0.001). The theoretical models showed the same pattern: species importance was consistently negatively associated with temporal CV (p < 0.001 across all models), and positively associated with number of links to other species (p < 0.001 across all models). Our results support the hypothesis that the most important species in persistent networks will be more buffered from environmental variation, and will be more highly connected with other species. Furthermore, the theoretical results suggest that this pattern may be broadly applicable to a wide range of ecological communities, and perhaps many types of networks in general. Hence merging ecological dynamics with network approaches may provide efficient insight into species on which to focus study and conservation efforts to understand and protect complex ecosystem structure and function.