Download
Filename Size Access Description License

Abstract

Ecology has always been a study of interconnection, making the growth of network,analysis something of a necessity as the field has continued to mature. Too,often, however, network analyses are included without full appreciation for,their appropriateness or requisite assumptions. This has led to a criticism of,these methods and a call to better articulate their relationship to and,implications for the underlying ecological system. In this dissertation, I,describe three projects that seek to address this divide. In the first chapter,,I look at how the structure of a network of interactions might be influenced by,the type of interaction being described (e.g. parasitism or pollination). This,is done through a combination of measuring a variety of network-structural,properties and comparing distributions of these values between different types,of networks using principal component analysis. We find that ecological,interaction networks are less different than has been previously proposed. In,the second chapter, we move this argument to inside a single network, in this,case, a food-web. We ask if species pursuing different trophic strategies (e.g.,parasitism or herbivory), have different structural roles within the network. We,attempted detection of these differences by use of a network structure model,known in Ecology as the group model. Though no single (or even combination of,several) simple structural properties is sufficient to distinguish trophic,strategies from one another, the group model consistently and significantly,creates groups of species of predominantly one trophic strategy. In the final,core chapter, we turn to networks of symmetric competition as might be found is,a system approximated by Lotka-Volterra competitive dynamics. We ask what,overarching network structure would yield the most or least stabilizing,community, and then ask how inter- and intraspecific components of that,structure might differentially influence that (de)stabilizing effect when varied,independently of one another. Surprisingly both the most and least stabilizing,configurations have clear, low-rank signatures in the network structure, with a,nested configuration proving most stabilizing and an anti-modular configuration,least stabilizing. In each of these three cases, we utilize a fairly abstract,method, but always with the goal of tying results to empirically observed,network structures. As with all science, only time will reveal success in this,endeavor.

Details

Additional Details

Actions

from
to
Download Full History