COS 49-6: Evidence for niches in food web data: A likelihood-based approach
David Alonso, Mercedes Pascual, and Stefano Allesina. Univeristy of Michigan
Food webs, networks describing feeding relations in ecosystems, are paradigmatic examples of complex systems in nature. Despite the challenge posed by the intricacy of these networks, simple models have been proposed for their topology that capture successfully particular structural properties. Two of the most prominent, the niche and cascade models, postulate the existence of a single dimension along which species can be ordered, with the resulting hierarchy constraining partially or completely the connections among species. The niche model further generates interval networks, in which all prey of a predator are adjacent on the niche axis, a property clearly related to the concept of ecological niches. Recently, it has been shown that the empirical food webs are close to being interval, a result interpreted as supportive of the niche model. The degree of compatibility with data is not, however, a good measure of the performance of a model, even if multiple characteristics are taken into account. What is missing is an evaluation based on the full topology of the networks. In order to achieve this goal, we have developed a likelihood-based approach. Since none of the two models is completely compatible with data, we first extended both models, providing alternative formulations capable of generating all empirical food webs. Given data from the best studied food webs available, we find that our extension, the minimal potential niche model, always yields better likelihoods. The central assumption of the niche model, the fact that predators consume prey that share similar characteristics, is well founded, even if it does not explain all consumer-resource links in empirical webs. These results open the possibility for a quantitative characterization of the nature of the niche axis and provide better null models for the analysis of ecological networks.