Forest management affects the quality and availability of roost sites for forest‐dwelling bats, but information on roost selection beyond the scale of individual forest stands is limited. We evaluated effects of topography (elevation, slope, and proximity of roads and streams), forest habitat class, and landscape patch configuration on selection of summer diurnal roosts by 6 species of forest‐dwelling bats in a diverse forested landscape of the Ouachita Mountains, Arkansas, USA. Our objectives were to identify landscape attributes that potentially affect roost placement, determine whether commonalities exist among species in their response to landscape attributes, and evaluate the effects of scale. We modeled roost selection at 2 spatial scales (250‐ and 1,000‐m radius around each roost). For each species, parameters included in models differed between the 2 scales, and there were no shared parameters for 2 species. Average coefficients of determination (R2) for small‐scale models were generally higher than for large‐scale models. Abundance of certain forest habitat classes were included more often than patch configuration or topography in differentiating roost from random locations, regardless of scale, and most species were more likely to roost in areas containing abundant thinned forest. Among topographic metrics, big brown bats (Eptesicus fuscus) were more likely to roost at higher elevations; roosts of big brown bats, northern long‐eared bats (Myotis septentrionalis), and Seminole bats (Lasiurus seminolus) were influenced by slope; and big brown bats, evening bats (Nycticeius humeralis), and Seminole bats were more likely to roost closer to water than random. Northern long‐eared bats and red bats (Lasiurus borealis) were more likely to roost closer to roads, whereas eastern pipistrelles (Perimyotis subflavus) were more likely to roost further from roads than random. Common parameters in most models included 1) positive associations with group selection (5 of 6 species) and thinned mature forest (4 species) at the small scale; 2) negative associations with unmanaged mixed pine‐hardwood forest 50–99 years old at the large scale (4 species); 3) negative association with stands of immature pine 15–29 years old at the small scale (3 species); and 4) a positive association with largest patch index at the large scale (3 species). Our results suggest that, in a completely forested landscape, a variety of stand types, seral stages, and management conditions, varying in size and topographic location throughout the landscape, would likely provide the landscape components for roosting required to maintain a diverse community of forest bats in the Ouachita Mountains.
