A landscape characterization procedure is described as a first step in ecosystem management. Five attributes were used to characterize patterns at different scales of ecological organization from the plot to the region: climatic regions, ecoregions, biophysical environments, floristics, and vegetation. Examples of the characterization of selected attributes are presented for three western U.S. study areas. Other aspects of the characterization process are illustrated with examples from the Northern Region of the Forest Service and from forested locations in the southwestern U.S. Location of a study area within a climatic region provided an understanding of broad-scale climatic constraints operating on the biota. Assessment at the ecoregion level was used to examine finer scale environmental constraints due to landform effects. The ecoregion also provided a context for further analysis of biotic and environment variability in a study area. Biophysical environments (combinations of environmental factors) were characterized within landscapes. The range of environmental variability in a landscape was compared to that of a larger region to determine the representation of regional environments in a landscape. Biotic variability was characterized at a relatively coarse spatial scale by examining plant species distributions among floristic types. Such an analysis provided information about the long-term ecological and evolutionary pressures exerted on species in an area. Variability in vegetation was characterized at a varietv of scales by examning the distribution of types within the levels of a hierarchical vegetation classification for the western U.S. In addition, biotic distributions along environmental gradients were described using canonical correspondence analysis to determine within-type variability. Temporal variability in vegetation was assessed within a hierarchical ecosystem-based framework. Characterization of biotic-abiotic relationships is important in determining biotic responses to historical and current landscape conditions. We constructed statistical models to predict biotic responses to environmental factors using a powerful class of regression models, generalized linear models. Models developed for a plant community in the Southwest, the ponderosa pine series, revealed complex biotic responses to environmental factors. This result suggests that biotic-abiotic relationships warrant careful characterization, and that some commonly used simple models may not accurately predict biotic responses to environmental change.