This spatial data product, the North American Forest Dynamics Attribution (NAFD-ATT) dataset, provides four 30-m resolution data files detailing the attribution of forest canopy cover loss events to forested areas across the conterminous United States (CONUS) as derived from Landsat images and random forest modeling for the period 1986-2010. The attribution data identifies the type of canopy cover loss event (i.e., stable/no loss, removals (harvest), fire, stress, wind, conversion, and other) predicted to have occurred in each forested pixel for the entire period. Also provided is data which identifies the year of onset of the canopy loss or forest disturbance event. Two additional files are derivatives of the random forest modeling process used to characterize confidence in the modeling process. Qualitative confidence metrics for each pixel's event prediction are available as well as a diversity index which is a measure of evenness of the distribution of the predicted canopy cover loss events in each pixel. Also provided is a file which represents the dominance of the selected type of event, that is, the difference between the number of votes for the "winning" class and "runner-up" class in the random forest modeling context.