Anthropogenic environmental changes are leading to habitat loss and degradation, driving many species to extinction. In this context, habitat models become increasingly important for effective species management and conservation. However, most habitat studies lack replicated study areas and do not properly address the role of nonstationarity and spatial scales in determining factors that limit species occurrence under different environmental settings. Here we provide an optimized multi-scale framework for analyzing habitat selection of the threatened Mexican Spotted Owl (Strix occidentalis lucida) between 2 meta-replicated study areas: the Sacramento Mountains, New Mexico, and the Mogollon Plateau, Arizona. The optimized scales of habitat variables strongly differed between the 2 study areas. Percent cover of mixed-conifer was more strongly associated with the relative likelihood of Mexican Spotted Owl occurrence in the Sacramento Mountains than in the Mogollon Plateau. Topographic covariates strongly explained variance in the habitat model in the Mogollon Plateau, but not in the Sacramento Mountains. Topographically constrained habitat availability may be affecting the relative likelihood of owl occurrence in the Mogollon Plateau, but not in the Sacramento Mountains. In the Sacramento Mountains, suitable habitat and owl distributions show dissimilar spatial autocorrelation patterns, indicating that the relative likelihood of occurrence may be influenced by factors in addition to habitat. Owl distribution shows a periodic spatial pattern, suggesting that the relative likelihood of owl occurrence in the Sacramento Mountains might be influenced by territoriality. Differences in habitat relationships between the 2 study areas suggest that management strategies should be tailored to local conditions. This study underscores the advantage of scale optimization and replicated studies in analyzing nonstationary habitat selection.