PARENT SESSION
Contributed Oral Session 48: Landscape Ecology: Animal Dynamics
Tuesday, August 9, 8:00 AM - 11:30 AM, Meeting Room 520 B, Level 5, Palais des congrès de Montréal

Landscape composition and structure influence winter white-tailed deer density across a forest-dominated landscape.

LeBouton, Joseph^*,1, Laurent, Edward¹, Walters, Michael¹, Friedman, Steven¹, Liu, Jianguo¹, ¹ Michigan State University, East Lansing, Michigan

ABSTRACT- Landscape composition and structure affect the movement and distribution of wildlife at a range of spatial scales. Consequently, the grain and extent relevant to studies of any particular species depend upon how that species perceives and responds to resource availability. We used spatially explicit data on winter white-tailed deer density (from 1645 spring fecal pellet count transects in deciduous forest) and vegetation patch characteristics (from classified multi-temporal Landsat ETM+ images) to quantify how landscape characteristics affect spatial variation in winter deer densities in a 400,000 ha study area in the central Upper Peninsula of Michigan, USA. We hypothesized that local deer distribution in winter is driven by: 1) landscape composition (ratio of coniferous forest providing winter thermal cover to deciduous forest providing winter food); 2) landscape structure (density of shared perimeter (m/ha) between patches of cover vs. food); and 3) proximity of (1) and (2) (thermal cover and edge) at each sampled point. Both ordinary least-squares (OLS) and geographically weighted regression (GWR) were used to examine spatial variation in relationships between deer density and the three variables. Winter deer density responded strongly to all three, and to interaction terms among them. In the global OLS model, maximum observed deer density increased from 5 to 25 deer/km2 with decreasing distance to thermal cover (800 to 0m). Deer density also increased with shared perimeter density. Highest observed deer densities corresponded to ratios of coniferous to deciduous land cover between 0.4 and 0.6. GWR provided a ten-fold increase over ordinary least squares in model fit (R2), removed spatial structure from model residuals, and indicated how deer response to landscape characteristics varies spatially. The explicit pattern of spatial variation in model coefficients generated with GWR aid in identifying how landscape characteristics interact to cause hot- and cold-spots in white-tailed deer distribution.

Key words: landscape structure and composition, white-tailed deer, geographically weighted regression