Assessing the influence of resource covariates at multiple spatial scales: an application to forest-dwelling caribou faced with intensive human activity

Efforts in isolating the relative effects of resources and disturbances on animal-distribution patterns remain hindered by the difficulty of accounting for multiple scales of resource selection by animals with seasonally dynamic drivers. We developed multi-scale, seasonal models to explore how local resource selection by the threatened forest-dwelling woodland caribou (Rangifer tarandus caribou) was influenced by both broad-scale landscape context and local resource heterogeneity in the intensively managed region of Charlevoix, Québec, Canada, located on the southern border of the North American caribou range. We estimated resource selection functions using 23 GPS-collared caribou monitored from 2004 to 2006 and landscape data on vegetation classes, terrain conditions, and roads. We found evidence of thresholds in road “proximity” effects (up to 1.25 km), which underscores the importance of including landscape context variables in addition to locally measured variables, and of fitting seasonal-specific models given temporal variation in the magnitude of selection and optimal scale of measurement. Open lichen woodlands were an important cover type for caribou during winter and spring, whereas deciduous forests, wetlands, and even young disturbed stands became important during calving and summer. Caribou consistently avoided roads and rugged terrain conditions at both local and landscape levels. Landscape context fundamentally constrains the choices available to animals, and we showed that failing to consider landscape context, or arbitrarily choosing an inappropriate scale for measuring covariates, may provide biased inferences with respect to habitat selection patterns. Effective habitat management for rare or declining species should carefully consider the hierarchical nature of habitat selection.