Winter selection of landscapes by woodland caribou: behavioural response to geographical gradients in habitat attributes.
Understanding animal-habitat relationships is central to the development of strategies for wildlife management and conservation. The availability of habitat attributes often changes along latitudinal and longitudinal axes, and animals may respond to those changes by adjusting their selection. We evaluated whether landscape selection by forest-dwelling woodland caribou Rangifer tarandus caribou varied along geographical gradients in habitat attributes. Centroids (n=422) of track networks made by caribou in winter were recorded during aerial surveys conducted over 161 920 km2 of boreal forest in Québec, Canada. Autologistic models were estimated by comparing the characteristics of landscapes (201 km2) centred on each centroid to an equal number of randomly located landscapes, with an autocovariate controlling for the non-independence among caribou locations. The availability of habitat attributes varied along longitudinal and latitudinal gradients, and caribou altered their landscape selection with respect to those gradients. Information Theory provided substantial support for only one model. The model revealed that the probability of occurrence of caribou increased with the abundance of conifer forests over most of the study region, but this positive response gradually became negative towards the southern portion of the region. The association between caribou and lichens changed from being negative west of the study region to being positive in the eastern part. Availability of landscapes dominated by lichen decreased from west to east. Finally, caribou generally displayed an aversion to areas with high road density, a negative association that became positive in the southern part of the study region. Synthesis and applications. Under current legislation in Canada, the critical habitat of woodland caribou must be defined, and then protected. Our autoregressive models can help to identify landscapes to prioritize conservation efforts. The probability of occurrence of caribou was related to different landscape characteristics across their range, which implies that the typical habitat of woodland caribou differs spatially. Such behavioural plasticity could be problematic for defining critical habitat, but we showed that spatial variation in landscape selection was organized along geographical gradients. Our study illustrates how geographical trends in habitat selection can guide management and conservation decisions.