"Environmental degradation has become a ubiquitous feature in the modern world. This degradation is resulting in widespread loss and fragmentation of wildlife habitat, leading to increased extinction risks and population declines. In order to stem these threats, it is imperative to accurately predict species' habitats, how to optimize the restoration and protection of these habitats, and better understand how their ecology interacts with habitat requirements. I used both simulated and empirical study systems to investigate these topics and focused heavily on giant panda (Ailuropoda melanoleuca) populations across Wolong Nature Reserve and Southwest China.Given the uncertainty and debate surrounding the relative effects of habitat amount and habitat fragmentation on ecological responses (Chapter 1), I set out to accurately define habitat before further investigating these effects. I found that the grain size of environmental predictor variables had important effects on modeling the distributions of virtual species simulated on real landscapes, and that modeling with grain sizes farther away from the "true" grain size used to simulate the species resulted in lower predictive accuracy and incorrect ecological inferences about the importance of environmental variables to habitat (Chapter 2). I then went a step further and investigated interactive spatial scale effects on species distribution modeling by varying the total extent of the study areas and grain size of the environmental variables used to predict panda habitat and distributions across Southwest China (Chapter 3). I found that increasing total extent offset the negative effects of increasing grain size on model accuracy and that total extent can be optimized as both larger (at smaller spatial scales) and smaller (at the geographic range scale) than the study area of interest. I then further improved the accuracy of our species habitat and distribution modeling by leveraging empirical movement distributions derived from GPS-collar data to transform the environmental predictor variables, and used the resulting habitat map to investigate the effects of habitat amount and fragmentation on functional connectivity in the panda population in Wolong (Chapter 4). I found that the standard deviation of the core area index, a measure of habitat configuration, was the best predictor of functional connectivity. Habitat amount was the second-best predictor and we found that in our study system it could optimized to cover about 80% of a local landscape to maximize functional connectivity. Habitat fragmentation also showed a nonlinear and threshold-dependent relationship with functional connectivity-important findings to consider in the spatial planning of protected areas. Finally, I used the noninvasive genetics data to "capture" and "recapture" unique individuals across a core habitat area in Wolong and conduct the first social network analysis of pandas (Chapter 5). I found strong evidence of two to three social clusters in the population, defined as groups of pandas that associated with each other at a significantly higher rate than individuals outside the cluster. These clusters may represent cryptic family structuring, as genetic relatedness was a significant positive predictor of associations between individuals. My detailed approaches to investigating the habitat and ecology of giant pandas used throughout this manuscript resulted in unique insights into this threatened habitat specialist species, and we recommend they be applied widely to other species. Optimizing the way in which we predict and conserve habitat in each landscape and system, as opposed to relying on expert opinion or competing theories, will be increasingly important as environmental degradation continues in the Anthropocene."--Pages ii-iii.
Read
