View article

Understanding the geographic distributions of animals is central to ecological inquiry and conservation planning. Movement-based habitat selection models, like discrete-time step-selection functions, are useful for identifying key environmental attributes which animals select upon and can nearly perfectly capture preference patterns between environmental conditions. However, predictions based only on selection can often fail to accurately describe true geographic distributions. We show that by integrating local selection patterns, movement models, and explicit landscape constraints within the same framework, we can better predict simulated distributions than occurrence-based frameworks of selection alone. Using three case studies, we show that this framework can better predict distributions of organisms across increasing scales of out-of-sample prediction: within individuals, between individuals, and even between regional contexts. Movement modeling is a powerful tool to describe both the selection paradigm that organisms apply to the environment and the movement patterns that enable them to apply those selection paradigms. By understanding movement and selection at the same time, habitats that species prefer may not be ultimately occupied, as many preferred patches may be unreachable given movement strategies and landscape context, often leading to smaller geographic distributions than estimated by selection or occurrence alone.