Description
Ecological studies are defined by three primary categories of activity: the collection of field data, statistical analysis, and modeling. Of the several meanings of the term "model," it is used here in the sense of simulation modeling, when the purpose is to extend the existing state of knowledge to potential scenarios not yet encountered or impossible to evaluate with field methods. This dissertation encompasses all three types of studies. In Chapter 1, we manipulated vegetation and squirrels in a replicated, large-scale field experiment. The intended objectives of the field experiment had a strong applied purpose, to provide managers with a cost-effective tool for restoring hybrid ecosystems with improved suitability for the recovery of Western Burrowing Owl (Athene cunicularia hypugaea) on protected reserve lands. We implemented short-term restoration treatments to re-establish key ecological processes provided by an ecosystem engineer, the California Ground Squirrel. The experiment produced new information about burrow availability for owls, squirrel habitat selection, and the positive feedback of squirrel activity on vegetation structure in grasslands. This chapter represents the type of focused field experiments needed to reduce uncertainty around parameters in the study system. The results can be used to inform management directly or to parameterize simulation models. Chapter 2 presents a statistical analysis that characterizes the variability in available field data sources in order to produce precise and robust estimates of the mean and variance of demographic parameters. We employed Bayesian Markov Chain Monte Carlo estimation to estimate four key life history parameters of Western Burrowing Owl: adult and juvenile-mortality, nest success, and nest productivity. In this suite of vital rates the estimate of juvenile mortality of Burrowing Owls was least well-resolved. While our specific goal was to provide improved estimates of burrowing owl vital rates, this approach is broadly applicable to synthesizing the findings of multiple field studies, linking demographic estimates to simulation models, and informing management decisions. In Chapter 3 we developed, parameterized, and evaluated the stability of an agent based model of burrowing owls that uses explicitly defined individual behavior to examine the relationship between owl settlement decisions and the habitat quality consequences for reproduction and population growth. In an analysis of model sensitivity, juvenile mortality consistently was the most influential input parameter, followed by adult owl mortality and squirrel mortality. We conducted an uncertainty analysis to verify the reliability of the model. Observations of densitydependent owl population growth provided evidence that the model produces biologically realistic population-level emergent behavior. We utilized the model to conduct a factorial experiment comparing the relative influences of owl density and a gradient of landscape proportions of low vs. high habitat qualities, which showed that initial owl density is also an influential driver of population persistence. The model currently enables qualitative consideration of the proximate mechanisms behind population decline, including the factors that attract owls to disturbed edge habitats. The progression of the three types of studies is generally expected to begin with data collection in the field. The data is then evaluated with statistical analyses, which indicates whether the amount and quality of data is sufficient for inference. Simulation modeling is conducted as the final step. Frequently, simulations are delayed until researchers believe they have a thorough understanding of the system. However, this dissertation compressed all three steps into a relatively short period of time. Proceeding directly to simulation created a feedback of valuable information to both our field efforts and statistical analyses, both by identifying the most influential parameters in the system, and by showing the potential range of model outcomes from current levels of parameter uncertainty. This identified needs for additional field data and updated analyses. This dissertation illustrates the value of planning for all three steps at the outset of research projects, to enable iterative improvement in data collection and parameter estimates, in order to provide managers with the best possible set of predictions and recommendations from modeling.
