The Appearance of Stochastic Control in Fish Communities on Coral Reefs: a Hierarchical Approach to System Organization

Abstract

<p>Species membership and population densities of fish communities on coral reefs are highly variable both spatially and temporally across many scales. Several models of stochastic control (i.e., organizing processes that produce these highly variable patterns) have been proposed. These models of stochastic control include: a lottery for available habitat sites, limited recruitment, and opportunistic predation. Although these models of stochastic control differ from each other in their organizing processes, they are similar in that they share two implicit assumptions. First, there is only one level or scale of system organization; second, there is only one significant organizing process. Recent developments in community ecology, however, suggest that the organization of fish communities on coral reefs is complex (i.e., involves multiple levels of system organization). The heuristic value of a multi-level system of organization is that patterns that appear under stochastic control can, at least theoretically, be broken down or decomposed into a series of simpler, non-stochastic patterns. If fish communities on coral reefs are indeed complex as in the above sense, then we need to re-evaluate the supposed single-level stochastic control in these systems. The purpose of this thesis, therefore, is to test the null hypothesis that fish communities on coral reefs are dependent on single-level stochastic processes as has been suggested. The alternative hypothesis is the appearance of stochastic control is a blend of patterns due to multiple levels of system organization. The method I use to test this is conceptually straightforward. The first step is to divide the species membership into several groups. In the second step, for each species group I determine the correlation between species distributions and, for example, some attribute of the habitat template. If the organization of fish communities on coral reefs is dependent on stochastic processes acting within a single level, then independent of how the species are divided into groups, the correlation for all species groups should be similar. I found that: - when the species are divided into groups based on similarities in the number of patch reefs a species is found, then significant differences occur among groups in correlation between species distributions and habitat attributes, - specifically, for each habitat attribute tested there exists a trend of increasing correlation (or decreasing correlation -- depending on the habitat attribute) from species found on most patches to species found on few patches. This suggests that the previous models of single-level stochastic control (e.g., habitat lottery, limited recruitment, and predation) are incomplete because they do not consider consequences due to differences in species' distributions and densities. Instead these results suggest that the structure and organization of fish on coral reefs incorporate a mixture of stochastic and non-stochastic processes that occur (a) simultaneously within the community, and (b) differentially between the species. This implies that fish communities on coral reefs are complex concerning their basic structure, and that unless a community level pattern is decomposed into its component elements, the conclusions made about the nature of these communities can be misleading.</p>