Groups and Niches in Food-Web Structure: Bayesian Methods and Consequences for Robustness.

Abstract

Food webs, networks of feeding links, provide a valuable abstraction for studying the structure and dynamics of ecosystems. This dissertation examines the importance of groups and niches to food-web structure, develops methodology for identifying these patterns, links these patterns to ecological and evolutionary relationships among species, and finally explores how these patterns relate to network robustness.

One pattern thought to be important for food-web structure is the presence of compartments, sub-networks of highly connected species. In one study, we explore the large-scale network architecture of a newly compiled plant-mammal food web from the Serengeti by developing a Bayesian method that identifies groups of species and patterns of linkages within and between groups. At the plant level, groups reflect habitat structure, an observation made possible by unusually high taxonomic resolution. Plant groups are coupled by groups of herbivores, which are in turn coupled by groups of carnivores, forming a pyramid structure that differs from the standard notion of compartmentalization.

Food webs have also been observed to exhibit low dimensionality in feeding niche space: species can be ordered so that diets consist of largely contiguous intervals. In a second study, we extend the Serengeti analysis using a hybrid model that simultaneously infers groups alongside niche structure. We find that both elements are important: groups reflect the overall trophic architecture of the web, while niche space constrains the fine detail of feeding interactions. We identify statistical relationships between model parameters and evolutionary relationships, thus demonstrating a relationship between biological traits and network topology.

Network structure determines, in part, how robust a food web is to extinctions. In a third study, we ask how different group structures vary in their robustness to bottom-up extinctions. We find that compartmentalization and interval niche structure reduce robustness. Notably, a coupled-pyramid structure like the one observed in the Serengeti is more robust than a compartmentalized structure, mirroring observations that spatial coupling can stabilize food webs. These results contrast with the prevailing wisdom that compartments are stabilizing, a difference that may result from opposing top-down and bottom-up effects.