Species distribution and abundance depend on a balance between local and landscape-scale processes. To successfully manage populations in regions with anthropogenic disturbances and habitat fragmentation, an understanding of important processes at each of these spatial scales is important. We used a model selection approach to identify an effective spatial scale to manage the Idaho Giant Salamander, Dicamptodon aterrimus. We used data from field surveys to compare support for local and landscape-scale models that explain D. aterrimus occurrence and density in 40 streams distributed throughout the Lochsa River basin, Idaho. Local-scale models included covariates that reflect patch quality. Landscape-scale models included variables that reflect predictions from metapopulation theory about the importance of patch size, connectivity, and fragmentation. Our results suggest that landscape-scale processes are important controls on D. aterrimus occurrence and that this species has broad habitat requirements within streams. Specifically, we found that probability of D. aterrimus occurrence was highest in roadless drainages and lowest in spatially isolated streams and in drainages with high old-growth forest density. Surprisingly, we found that D. aterrimus density was greatest in streams with a high proportion of embedded substrate and fine sediment. The positive association with embedded substrate may reflect adaptation to a high frequency of natural disturbances, such as landslides, in our study area. We suggest that management and conservation efforts for this species focus on protecting roadless areas and restoring stream connectivity in human-impacted areas, rather than on only improving habitat quality within streams.