Author ORCID Identifier

Year of Publication


Degree Name

Master of Science in Forest and Natural Resource Sciences (MSFNRS)

Document Type

Master's Thesis


Agriculture, Food and Environment


Forestry and Natural Resources

First Advisor

Dr. Matthew T. Springer

Second Advisor

Dr. John J. Cox


The elk (Cervus canadensis) was extirpated from its range in eastern North America by the end of the 1800s, prompting several U.S. states and Canadian provinces to begin translocation programs with the goal of reestablishing elk populations. While eastern elk managers have relied on information from western herds to guide population and habitat management, there is a need for region-specific research on the spatial ecology and habitat associations of translocated elk given the stark differences in landscape, climate, predator communities, and harvest regimes across the continent. While the Kentucky elk reintroduction is one of the best documented programs to date, including a wealth of scientific study, a deeper understanding of how elk relate to their environment will be required for the future considering changes in landscape and population dynamics.

We studied the spatial ecology of female elk across the Kentucky Elk Restoration Zone (KERZ), focusing our inquiry on several aspects of reproduction and behavior via GPS telemetry data. First, we analyzed movement, space, and habitat use patterns during the calving season to predict parturition and early neonatal survival. We trained a random forest model to classify focal time windows as parturition or non-parturition, using birth events confirmed by expulsion of vaginal implant transmitters (VITs). We then applied the model to pregnant females for which parturition and one week calf survival status was unknown. Using a probability-based decision rule, our model was highly accurate (89.5%) at correctly classifying elk with known status. When applied to unknown individuals, we found that less than half (48.6%) were predicted to give birth and rear their calf to a week old, compared to 85.1% of known-status females, suggesting that fetal mortality and early calf loss may contribute more strongly to reproductive success in this population than previously thought.

Second, we investigated calving habitat selection by comparing landscape characteristics from confirmed parturition sites to random, “available” locations within elk home ranges. Specifically, we sampled landscape covariates (reflecting vegetation, topography, and human footprint) and both used and available locations and modeled the relative intensity of use with resource selection functions (RSFs) implemented with generalized linear mixed-effects models (GLMMs). We sampled covariates at several spatial grains to capture scale-dependent selection patterns. We found that females selected birth sites with intermediate levels of canopy cover on gentle topography at fine grains, with lower vegetation greenness, higher topographic positions, higher edge densities, and intermediate levels of patch interspersion at coarser grains. This suggests that female elk make multi-scale decisions during parturition to minimize predation risk to their neonate near the birth site while maximizing forage availability in the surrounding area.

Finally, we characterized the behavioral variation present in Kentucky elk by fitting GLMM RSFs including random slopes to female elk data collected during four biological seasons. We used the random slopes, representing group-level selection coefficients, to assess the effects of landscape composition and configuration on variation in habitat selection (i.e., testing for functional responses) with regression splines. We also used k-means clustering to identify general behavioral tactics within seasons, accounting for multidimensional correlation between behaviors (i.e., behavioral syndromes). We found the highest variability in elk responses to open areas, primary/secondary roads, and successional forests, despite strong population-level selection/avoidance of these covariates. While elk exhibited clear functional responses to availability and configuration of these and other covariates, showing the partial context dependence of habitat selection in this population, clustering analysis assigned groups to major behavioral tactics which better predicted intensity of use over global and functional models.

Our overall results demonstrate both population- and individual-level patterns of space and landscape use in Kentucky elk. We provide a framework for identifying successful reproduction remotely, without the use of VITs, which can assist in informing population models that may be limited by sample size and/or ignoring the potential influence of fetal mortality on calf recruitment. We also illustrate population wide patterns of parturition site selection that can be useful in delineating suitable calving habitat and targeting management actions such as the introduction of disturbance. Lastly, the extensive variability in female habitat selection across the annual cycle can help explain shifts in elk behavior and landscapes in the KERZ change, highlighting how flexible individuals in the population are to changing and novel landscapes.

Digital Object Identifier (DOI)

Funding Information

Kentucky Department of Fish and Wildlife Resources

U.S. Department of Agriculture McIntire-Stennis program (project #1021936)

Rocky Mountain Elk Foundation

Available for download on Wednesday, July 12, 2023