Date Approved


Graduate Degree Type


Degree Name

Biology (M.S.)

Degree Program


First Advisor

Paul Keenlance

Second Advisor

Joseph Jacquot

Third Advisor

Alexandra Locher

Fourth Advisor

Robert Sanders

Academic Year



American marten (Martes americana) are typically associated with mature coniferous forests. Marten were extirpated from Michigan’s Lower Peninsula due to human impacts, such as fire, logging, and over-harvest. Little is known about the resource selection and distribution of marten in Michigan’s northern Lower Peninsula since their reintroduction in 1985-86. Resource selection functions are valuable tools to estimate the relative probability an animal will utilize an area and predict where they may occur. When creating a resource selection function, potential sources of variation in data collection methods and wildlife populations should be considered to ensure accurate results. We sought to create a resource selection function for marten across Michigan’s northern Lower Peninsula to estimate their occurrence and identify habitat with higher probability of use by marten to maintain or with lower probability of use to improve. We also sought to determine whether home-range estimates derived across different seasons or collar types would impact marten home-range size and habitat selection. Marten were live-trapped, fitted with VHF or GPS collars, and locations were obtained via radio-tracking VHF collars or downloading data stored within GPS collars. We estimated 95% fixed kernel home-ranges for all marten with ≥ 30 locations from one collar type. Characteristics potentially indicative of marten habitat selection determined a priori were measured within each used home-range and surrounding available habitat. Kruskal-Wallis and ANOVA tests were used to compare habitat characteristics, size, and overlap among individuals with home-ranges from both VHF and GPS collars. Forward and backward selection were used to establish the best-fit logistic regression model explaining marten resource selection. Comparisons between home-ranges estimated for 5 marten based on collar type and season revealed no significant differences. Therefore, we combined home-range data across collar types and seasons. We used home-range data from 18 marten to generate the resource selection function, which predicted percent of canopy cover, coniferous forest, and mixed forest were the best indicators of marten habitat selection. We extrapolated our model to Michigan’s northern Lower Peninsula and found ~38% was estimated to have a relatively high probability of being used by marten.

Included in

Biology Commons