PS 65-82
Network complexity improves the predictive certainty of community responses to perturbations in allometrically-constrained networks
The complexity of ecological networks has long plagued attempts to predict species responses to perturbations. The more speciose and connected the network, the lower the confidence in model predictions, unless accurate estimates of species interaction strengths are known. Low predictive certainty results from the many direct and indirect pathways that contribute to each species overall response to a perturbation. Consequently, most models of community dynamics rely on simplifying assumptions to reduce network complexity and much effort has focused on quantifying interaction strengths. We investigated how the allometric relationships between body size, interaction strength, and network topology may constrain sensitivity and improve predictive certainty of complex networks.
We simulated allometric trophic networks with topologies varying in species richness and connectance using the niche model. Parameters contributing to interaction strengths (metabolic, ingestion, and production rates) were dependent on species body masses, drawn from empirical distributions of predator/prey body mass ratios. Predictive sensitivity was assessed by systematically varying the uncertainty of interaction strength estimates with respect to the “true” equilibrial responses of each simulated community.
Results/Conclusions
Allometrically-constrained networks had considerably improved average predictability over non-constrained networks. Further, and counter-intuitively, complexity (connectance) increased predictability when networks were allometrically constrained. This occurred because the distribution of interaction strengths is key to predictive certainty and is directly influenced by allometry, species richness, and connectivity. As a consequence, having accurate estimates of only the strongest 5% of species interactions dramatically improves predictability, especially for highly speciose and connected networks.
We conclude that efforts to predict the response of multispecies communities to perturbations will profit by incorporating allometric scaling, embracing network complexity (i.e. avoiding network simplification), and accurately identifying and estimating the strength of the strongest species interactions in the community.