Using wildlife occurrence data to test permeability estimates and ecological indices used in urban planning

Increasing urban development degrades ecosystems partly by diminishing natural area connectivity and quality, ultimately reducing and homogenizing urban biodiversity. To support biodiversity, ecological planners in Edmonton, Alberta (hereafter the City) have implemented tools to incorporate wildlife habitat into land use planning. These tools include circuit-based simulation models that used permeability estimates based on coyote (Canis latrans) movement to approximate connectivity for urban mammals. Two other indices, biodiversity potential and ecological connectivity, estimate the ecological value of natural areas based on patch characteristics. To evaluate the predictive capacity of these tools, this thesis compared the predicted permeability and habitat values with animal occurrence data from GPS collars, camera traps, and small mammal track tube arrays. In Chapter 2, I explored how habitat selection by 19 urban coyotes fitted with GPS collars was affected by health status (via infection with sarcoptic mange; Sarcoptes scabiei) and season (summer vs. winter) using two modelling approaches. I used two seasonal compositional analyses to explore the selection of broad categories to obtain selection estimates as log-ratios of proportionate use to compare with feature-specific permeability ratings. I then built a RSF model to assess fine-scale habitat selection and derived a habitat suitability index (HSI) to compare with cumulative landscape permeability values used in circuit-based models in linear regressions. From compositional analyses, whether coyotes used or avoided habitat was consistent between seasons, but used natural forests, natural shrubland, modified grass/shrubland and residential areas more in winter. The RSF model showed that coyotes largely avoided developed areas, but selected steeper slopes and areas closer to natural areas, modified forests and grass/shrubland, and residential areas. Coyotes with mange were more likely to use human-dominated areas, especially in winter. The feature-specific permeability ratings used in circuit-based models undervalued residential and developed areas in both seasons and maintained grass in summer, while overvaluing most natural vegetation types and the North Saskatchewan River in winter. The landscape permeability estimates were predictive of the RSF-derived HSI, but less so in winter and when coyotes had mange, and the model fit was poor. In Chapter 3, I used data from 89 camera traps and 47 track tube arrays placed throughout Edmonton, Alberta to measure the occurrence and relative abundance of three groups of terrestrial mammals (small, medium, and large) and 13 species. I used these as response variables to evaluate the predictive capacity of two ecological indices used by the City in zero-inflated Poisson and linear mixed models. I also modelled detections of the three groups and five species using various remotely-sensed and site-based variables. The indices of biodiversity potential and ecological connectivity used by the City correlated variably and generally with the occurrence or relative abundance of groups and species. As biodiversity potential increased, large mammals occurred more often, and white-tailed deer (Odocoileus virginianus) were more abundant, but the abundance of small mammals declined. By contrast, higher ecological connectivity predicted more abundant small mammals, but less abundant snowshoe hare (Lepus americanus) and white-tailed deer. I found high variability in the predictiveness of remotely-sensed and field-measured variables among and within species groups, with patch-level covariates predicting only small mammal abundance from track tubes. I found generally adverse effects of human disturbances, such as urban density, human activity, and off-leash areas, but with some positive associations with detections of domestic dogs. In combination, my results suggest that expert-derived estimates of landscape permeability used to model connectivity by the City reflect habitat selection by urban coyotes. However, the accuracy of such models could be improved by using empirical data, such as those provided by GPS collars. Doing so could identify the effects of individual variation and season, and the high capability of using developed areas by urban-adapted species like coyotes. My results from the tests of ecological indices suggest that a larger buffer width may be necessary to represent connective habitat for larger species, and a lower weight for wetland habitat may better reflect habitat quality for mammal species in Edmonton. The accuracy of ecological indices used by urban planners could be increased by considering surrounding vegetation density and type, and human infrastructure and activity. Further development of such indices will assist Edmonton and other cities retain biodiversity and the many ecological benefits provided by wildlife.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.