Nonlinear Modeling of Muskrat Use of Habitat

Where muskrats (Ondatra zibethicus) are burrow dwellers, traditional aerial surveys of lodges to characterize the potential of different areas to sustain the species are not adequate. We present an alternative to identifying critical habitat variables on the basis of muskrat presence indices. We used stepwise logistic regression to create a habitat model based on muskrat presence in early fall along various wetlands of James Bay. We built 2 models, 1 based on presence of burrows only and 1 using the presence of muskrat feeding signs and droppings. Collecting the latter data required less field time than looking for burrows. The burrow model had a classification rate of 88 and 92% for 60-m shore sections of wetland used to build the model and other sections of streams and rivers, respectively. Classification rates for the feeding signs and droppings model were lower (79 and 71%, respectively). Logistic regression analysis on presence/absence of burrow (P < 0.001) selected bank slope, percentages of floating and submerged plant cover, presence of clay-loam soil, and width of shore herbaceous belt as important habitat variables. The number of muskrat burrows in the study area averaged 2.1 ± 5.5/km, and lodges were not found. Slow-flowing rivers represented the best habitats. Methodology presented here may be used to determine variables related to muskrat presence in riverine wetlands. These habitat variables may be used to assign values reflecting the potential of each riparian section of a given wetland to harbor burrow-dwelling muskrats