A framework designed to guide the effective use of remote sensing in large-area, multi-jurisdictional habitat mapping studies has been developed. Based on hierarchy theory and the remote sensing scene model, the approach advocates (i) identifying the key physical attributes operating on the landscape; (ii) selecting a series of suitable remote sensing data whose spatial, spectral, radiometric, and temporal characteristics correspond to the attributes of interest; and (iii) applying an intelligent succession of scale-sensitive data processing techniques that are capable of delivering the desired information. The approach differs substantially from the single-map, classification-based strategies that have largely dominated the wildlife literature, and is designed to deliver a sophisticated, multi-layer information base that is capable of supporting a variety of management objectives. The framework was implemented in the creation of a multi-layer database composed of land cover, crown closure, species composition, and leaf area index (LAI) phenology over more than 100,000 km 2 in west-central Alberta. Generated through a combination of object-oriented classification, conventional regression, and generalized linear models, the products represent a high-quality, flexible information base constructed over an exceptionally challenging multi-jurisdictional environment. A quantitative comparison with two alternative large-area information sources---the Alberta Vegetation Inventory and a conventional classification-based land-cover map---showed that the thesis database had the highest map quality and was best capable of explaining both individual- and population-level resource selection by grizzly bears.