9th International Deer Biology Congress

AUTHORS: Meghan J. Camp, Lisa A. Shipley, and Daniel H. Thornton – School of the Environment, Washington State University

ABSTRACT: Changes in forest management over the last century, such as fuels reductions through thinning and prescribed burning, has the potential to influence populations and distributions of both mule deer (Odocoileus hemionus) and white-tailed deer (Odocoileus virginianus) in forested landscapes. However, because deer are difficult to survey in forested landscapes using traditional methods (e.g., capture and telemetry, transect surveys), biologists currently lack a method to reliably estimate densities within these landscapes. Furthermore, we have a poor understanding of the spatial and temporal segregation of mule deer and white-tailed, or of differences between the species in how they select habitat. Our project aims to address these problems by testing and validating a novel method that uses camera traps to estimate species-specific densities within the Colville National Forest in northeastern Washington, especially in relation to forest management initiatives aimed at reducing the risk of forest fires [1].
During October 2017, we deployed cameras for 26 days in the Colville National Forest to test the feasibility of a larger study to be carried out during the summer of 2018. The camera grid had an array of 30 cameras spaced 500 m apart (Fig. 1). At each camera location, we measured visual obstruction using a cover pole at 7 m and 15 m from the camera to provide an estimate of detectability by the camera in different habitats. During installation of the cameras, we measure distanced from the camera and recorded videos of researchers holding distance markers at 1- m intervals out to 20 m to provide a reference for estimating distances to filmed deer. To separately estimate densities of mule deer, white-tailed deer, and moose we estimated distances to recorded animals by comparing their distances to those of researchers in the reference videos. We fit point transect models, adapted to camera trap data, in program Distance [1]. We included visual obstruction at 7 and 15 m as covariates in the detection function to account for differences in detectability of animals because of vegetation. We sampled a total area of 12.25 km2. Eleven of the 30 cameras had at least one image of a moose, 12 had mule deer, and 21 had white-tailed deer. The density estimates, without including visual obstruction in the model, were 11.98 / km2 for mule deer, 15.77 / km2 for white-tailed deer, and 0.68 / km2 for moose (Table 1). The inclusion of visual obstruction in the model slightly improved the fit of the model (i.e., reduced AIC) and increased the density estimates (Table 2). Our next step is to sample deer and moose across a larger area and variety of habitat types within the Colville National Forest and compare our new camera-based method of density estimate with density estimated by traditional techniques (i.e., pellet group counts and line transect sampling).