CONSTRUCTION AND TURNOVER OF MULTISPECIES COMMUNITIES: A CRITIQUE OF APPROACHES TO ECOLOGICAL COMPLEXITY (TIME DEVELOPMENT, STABILITY, MODELS, ECOSYSTEM, APPARENT INTERACTIONS)

This thesis explores a "developmental" approach to a long standing issue for theoretical ecology: the persistence of ecological complexity. Although, when we consider model systems of increasing complexity, stable cases generally become rarer, I ask if we can construct them over time by "trial-and-elimination (T&E)" of potential component populations. To place this exploration in context I examine critically choice of parameter values, interpretations of models, and modeling strategies.; When Generalized Lotka-Volterra (GLV) models govern the dynamics, T&E constructs stable systems, mostly by eliminating unfeasible populations (equilibria negative), thus favoring positive intrinsic increase and interaction terms. This bias increases the frequency of pathological systems, in which some populations grow without limit. Besides this bias T&E does not select populations with atypical parameter values, and the internal arrangement of stable systems cannot be characterized as cyclical or blocked.; When ecosystem-like models, having population interactions and nutrient and energy utilization, govern the dynamics, complex systems can be constructed but their persistence is limited; maintaining complexity requires continued colonization.; Concerning parameter choice: by scaling GLV models to standard representations I refute previous biological justifications for chosen parameter values. Through Monte-Carlo sampling I revise the "semi-circle" law for large random matrices and demonstrate how a measure of complexity, interactance, simplifies parameter choice.; Concerning interpretations and strategy: models can represent relations generating some observations, redescribe such observations, be tools for exploration, or be schemata. Good fit and accurate predictions are insufficient to confirm models as generative representations. We need also to establish accessory conditions, often overlooked and difficult to establish, especially for simple models, e.g., the effective independence of modeled variables from omitted variables. Simple models are suitable for exploration, e.g., the T&E studies, but do not ensure ecological generality.; I illustrate these interpretations for the logistic equation and in analyzing "apparent" interactions, e.g., "predator-prey" interactions between competing protozoan populations. My derivation of apparent interactions incorporates both direct effects between modeled variables and indirect effects of omitted variables to mimic closely the observed population trajectories. Apparent interactions are often counterintuitive and context-sensitive; therefore they are redescriptions, not generative representations.