Elk Distribution and Modeling in Relation to Roads

We tested performance of 3 aspects of an elk (Cervus elaphus)-road density model that has been used extensively throughout western North America but has not been sufficiently validated. First, we tested the hypothesis that elk selection of habitats increases with increasing distance away from open roads. This forms the empirical basis for the model. Second, we compared the model's predictions of relative elk habitat selection, or habitat effectiveness (HE), with observed values at varying levels of road density. And third, we examined the potentially confounding effects of different spatial patterns of roads on model predictions. We conducted our study during spring and summer, 1993-95, at the Starkey Experimental Forest and Range (Starkey), northeast Oregon. Selection ratios were calculated using >100,000 recorded locations of 89 radiocollared female elk, with locations mapped in relation to 0.1-km-wide distance bands away from open roads. Selection ratios increased with increasing distance from open roads, and varied between seasons, but not among years or individual animals. Linear regression models, using distance to open roads as a predictor, accounted for significant variation in selection ratios during spring and summer. Model predictions of HE, as measured by number of elk locations, corresponded only weakly, however, with observed values of HE. The contradictory results of these 2 analyses may be explained in part by our simulation results, which showed that potential reductions in elk HE vary strongly with the spatial pattern of roads, which is not measured by the elk-road density model. Our results suggest that (1) management of roads and related human activities during spring and summer should remain an important consideration for modeling and managing the elk resource; and (2) a spatially explicit road component is needed for elk habitat models.