Public lands in Wyoming are under intense pressure for energy development. However, vast areas lack soil survey data necessary for evaluating environmental impacts. We demonstrate the usefulness of classification tree analysis for rapidly and accurately predicting the distribution of soil classes in the Green River Basin of Wyoming, which is under rapid expansion of natural gas extraction. Topographic data derived from Digital Elevation Models (DEMs) and various band combinations and ratios of Landsat 7 remotely sensed spectral data were selected to represent soil forming factors. Unsupervised classification techniques were used to recognize existing soil-landscape patterns and to develop an initial sampling plan. In 2004, more than 1000 data points were collected in the field and were assigned to a developing set of soil map units. The digital data served as independent variables whereas the soil map units were dependent variables for classification tree analysis to predict soil map unit distribution on the landscape. The output images generated reasonably represented soil map unit concepts developed by the soil survey team. Prediction errors decreased as more field data were added. Boosting and cross-validation methods further decreased prediction errors. Classification trees are objective, apply soil map unit concepts consistently across the survey area, and are less time intensive than traditional soil survey methods. Models produced by classification tree analysis are transparent and could be further refined as additional data become available or as land-use needs change.