UC Berkeley

The built environment, especially roads, urban and suburban development, can reduce the ability for wildlife to move across landscapes. Maintaining landscape connectivity has become a central theme in ecology and conservation, as corridors of intact habitat help maintain ecosystem functionality and, in the face of climate change, may provide migration paths for species. However, the influence of the built environment on connectivity is rarely quantified using empirical data informed by species detection, movement, or genetic structure. Rather, structural connectivity, as opposed to functional connectivity, is estimated using land cover alone. Structural connectivity estimates offer a simple and potentially powerful approach with fewer data requirements for wildlife corridor planning; however, models of structural connectivity are rarely if ever evaluated with empirical species data, limiting our understanding of their reliability and utility. This dissertation fills this gap by investigating the influence of human land use and habitat fragmentation on landscape connectivity using a suite of quantitative modeling approaches and mammals as the focal species, including cross comparisons among these approaches. Specifically, three methods that vary in levels of biological information are used to evaluate how well structural connectivity models perform for individual species, as well as their relationship to functional connectivity.

To begin with, the utility of a structural connectivity model based on the distribution and intensity of land use is evaluated by comparing model predictions to observed land use by a generalist carnivore, the puma (Puma concolor). Findings from this study indicate that generic landscape permeability models can be used with confidence as a guide when prioritizing habitat corridors for biodiversity conservation across fragmented landscapes. Next, the utility of structural connectivity models is further evaluated by examining how the inclusion of specific human land use variables affects model accuracy in a species distribution model for gray fox (Urocyon cinereoargenteus). Findings from this study indicate that species distribution models generated in human-dominated landscapes have higher accuracy when informed by indices of land use. Finally, a combination of spatial and genetic methods is used to evaluate the influence of roads on the functional connectivity for a small mammal, California ground squirrels (Otospermophilus beecheyi). Findings from this study indicate that a combined spatial and genetic approach can be used to identify locations where roads act as barriers.

Given the importance of habitat fragmentation, there is a pressing need to rapidly develop and utilize connectivity assessment methods in conservation planning. Research findings presented here have already impacted mammal conservation planning and management in California in the following specific ways. The structural connectivity model was used to identify priority habitat linkages for inclusion in the Conservation Blueprint for Santa Cruz County, and the combined genetic and spatial approach was used by the Santa Clara County Open Space Authority, CA to identify corridors and restore connectivity in an area with an existing road network. Beyond these regional impacts, future conservation planning worldwide can benefit from using readily available data collected by citizen scientists as input in predictive mapping to increase the sample size and spatial coverage for species distribution modeling.