Many wildfire events have burned thousands of hectares across the western United States, such as the Bitterroot (Montana), Rodeo-Chediski (Arizona), Hayman (Colorado), and Biscuit (Oregon) fires. These events led to Congress enacting the Healthy Forest Restoration Act of 2003, which, with other policies, encourages federal and state agencies to decrease wildfire risks by evaluating, prioritizing, and implementing vegetation treatments across large landscapes. Land management agencies, and society, have high expectations that vegetation (fuel) treatments and forest restoration activities will moderate fire behavior (intensity) and its effects, resulting in the enrichment of forest values. However, the uncertainty of these relations is unknown, preventing forest managers from communicating their confidence in the effectiveness of fuel treatments in reducing risk of wildfires. To address this uncertainty, we observed the relation between pre-wildfire forest structure and burn severity across cold, moist, and dry forest types. We used a combination of collaborative studies and field data from 73 wildfire events in Idaho, Oregon, Montana, Colorado, Arizona, and Utah (which burned between 2000 and 2003) to obtain over 900 observations. We used a multiple spatial scale approach to provide insight into how physical setting, weather, and site-specific forest structures relate to tree burn severity, with conditional probabilities that provide an estimate of uncertainty. The burn severity classification we developed integrates fire intensity, fire severity, and the forest's response to wildfire. Forest and wildfire characteristics that determine tree burn severity are: a particular wildfire group, tree canopy base height, total forest cover, surface fuel amount, forest type, tree crown ratio, and tree diameter. Because of the study's wide breadth, results from it are applicable throughout the Rocky Mountains.