Monday, August 4, 2008 - 2:10 PM

COS 6-3: Predicting interaction strengths in complex food webs - CANCELLED

Ulrich Brose, Darmstadt University of Technology, Eric L. Berlow, University of California at Merced, Richard J. Williams, Microsoft Research Ltd., Jennifer A. Dunne, Santa Fe Institute, and Neo D. Martinez, Pacific Ecoinformatics and Computational Ecology Lab.


Predicting how the loss of one species alters the abundance of other species in complex network of interactions has been considered almost intractable. While metabolic theory suggests that metabolism, consumption and energy fluxes of direct predator-prey interactions may be universally constrained by simple allometric scaling rules; it is unclear if these simple rules are maintained when species interact in a realistically complex network. We addressed this question by simulating species population dynamics in ecological networks where each species’ metabolic and maximum consumption rates scale as a ¾ power law with its body mass. We carried out simulated removal experiments in 600 ecological networks with over 80 stochastically varied species, link, and network attributes, and quantified the strengths of all possible interactions. The attributes of network structure and species’ traits were then used to predict the sign and magnitude of species interaction strengths, measured as both the per capita and population-level effects of removing one species (R) on the biomass change of another target species (T).


The simple allometric rules that drove pair-wise feeding interactions disappeared in complex networks. However, new surprisingly simple patterns emerged: the absolute magnitude of per capita and population interaction strengths was consistently well predicted by the biomasses of R and T and R’s body mass; and the sign of the interactions was well predicted by the net sign of the shortest and second shortest paths between R and T. This allometrically-based model helps establish a metabolic baseline for understanding other sources of variation in species interactions in natural systems. Perhaps most surprisingly, the ease of predicting interaction strength in our simulations increased with web size suggesting that diversity and complexity may simplify rather than complicate our understanding of natural communities.