Geostatistics reveal the scale of habitat selection

We developed a new model of habitat selection to detect the extent at which organisms respond to habitat. Based on the premise that selection of a resource entails a change in its variance, the model compares two variograms: sites used by animals versus the general environment. With an artificial landscape, we explored this model by varying the extent of perception by animals, which was revealed as the approximate point of maximum divergence in variance. Selection for greater resource abundance produced variograms with lower variance than the general environment, whereas selection for extremes of resource abundance generated the converse. We applied the model to winter resource selection of muskoxen. Variograms in forage abundance diverged sharply at 200 m, consistent with a conventional analysis that demonstrated the strongest selection at that scale. The model appears widely applicable and capable of revealing the spatial scales at which organisms react to their environment.