Using occupancy modeling to estimate American marten distribution in Michigan’s Lower Peninsula

Link To Share This Poster: https://cdmcd.co/5mqG3P
Live Discussion Link: https://cdmcd.co/v9DZw5

Maria N. Weston and Paul Keenlance, Biology, Grand Valley State University, Allendale, MI, Angela Kujawa and Robert Sanders, Natural Resources, Little River Band of Ottawa Indians, Manistee, MI

Background/Question/Methods
American marten (Martes americana) were extirpated from Michigan’s Lower Peninsula in the early-20th century due to habitat loss, habitat degradation, and overharvesting. In the 1980s, eighty-five marten were reintroduced in total between two areas of their historical range in Michigan: the Manistee National Forest and the Pigeon River County State Forest. However, a lack of monitoring post-release has led to considerable uncertainty about their current distribution. Determining marten occurrence is a crucial first step when attempting to implement active management techniques for this imperiled species that has both ecological and cultural significance within the state of Michigan. Only after a marten is detected can we begin to understand what requirements are necessary for the survival of this species. To assess current marten distribution throughout the northern Lower Peninsula (NLP), the main objectives for this project were to (1) Use occupancy modeling to estimate American marten occurrence in the NLP using species detections from a large-scale camera trap study, (2) Determine the detection probability of American marten in the NLP, and (3) Determine if habitat covariates or the presence/absence of predators and conspecifics can be used to estimate marten occurrence. We used detection-nondetection data from a large-scale camera trap study encompassing 27,371km2 of state and federal lands to estimate marten occurrence across Michigan’s northern Lower Peninsula.
Results/Conclusions
We modeled marten occupancy as a function of our series of covariates which indicated that marten occupy 19% of sites surveyed from 2019-2020 with a detection probability of 22%. Model predictions indicate high occupancy probabilities (>0.90) when canopy cover is high (>0.84), and when predators were not detected at the site. In this study, we demonstrate the use of occupancy modeling as an aid to determining species distribution and improving management decisions when abundance data is otherwise unavailable.