Data layers providing detailed information on vegetation composition and structure are critical for fire risks assessment, post-fire rehabilitation, and a range of ecological studies. For the LANDFIRE project, these data layers will be used as input to a suite of models for predicting fire behavior and fuel loadings. In this study, we report the results of a prototype study for evaluating the feasibility of deriving a number of vegetation attributes including cover type, canopy density and height for natural vegetation communities including forests, shrubs and grasses at the 30 m spatial resolution. The prototype area covered the Wasatch Range-Uinta Mountains region of central Utah, an area of about 70,000 km2. Source data included three dates Landsat ETM+ images acquired in the spring, summer and fall seasons between 1999 and 2001, and 30 m digital elevation data and derived slope, aspect and a position index. Ground reference data collected from over 6000 plots by different agencies were used. Vegetation cover type was modeled and predicted using a classification tree technique, while height and canopy density were modeled and predicted using both classification and regression tree techniques. Model results were evaluated through cross-validation. Preliminary results indicated that overall accuracies of about 60% could be achieved in separating 12 forest cover types, 10 shrub cover types, and 6 grass cover types, respectively. Correlation coefficients (r) for modeling percent cover of trees, shrubs and grasses were about 0.9, 0.6, and 0.55 respectively, and were 0.7, 0.5, and 0.2 for modeling height of the three life forms, respectively. When only two classes – high and low, were considered for the height and percent cover of shrubs and grasses, overall accuracies of about 75% were achieved for percent cover of both shrubs and grasses, and 82% and 65% for height of the two life forms, respectively. We are currently evaluating whether additional data layers describing environmental gradients and potential vegetation distributions can be used to improve mapping accuracies. A field trip is scheduled this summer to collect independent ground reference data for validating prototype results.