ObjectiveThe loss and degradation of aquatic habitat through fragmentation, water extraction, and climate warming contribute to declining native stream fish diversity. In response to these declines, a large‐scale repatriation program was established in Arizona and New Mexico to expand the distribution and abundance of native fishes. This program has had variable success, with imperiled fish populations establishing and persisting in some streams, while other populations failed to establish.MethodsTo identify factors contributing to the success of repatriation, we conducted a study investigating how biotic and abiotic variables affect native fish presence or absence at differing spatial scales throughout the Gila River basin. We conducted fish and habitat surveys in 29 river reaches at three spatial scales: the 1‐km, 0.1‐km, and mesohabitat scale. We used generalized linear models to identify associations among species presence, habitat characteristics, and nonnative species abundance across spatial scales. Logistic regression models were also used to test associations with heterogeneity of habitat characteristics and species presence.ResultModels indicated strong species‐ and scale‐specific responses to these variables. Nonnative abundance was less influential than other habitat variables in predicting the presence of native fish, although the majority of associations that emerged were negative. Further, we were able to identify that water velocity, depth, and substrate size were the most common variables associated with species' presence across all scales and age‐classes, with most species responding positively to elevated velocities and variable responses for depth and substrate size. Response to variation in habitat was species‐specific, with some associations positive and others negative.ConclusionCollectively, the information from this study will help identify locations with a reasonably high likelihood of supporting native fishes and broaden our understanding of species responses to habitat and nonnative species across multiple scales.