Natural and anthropogenic agents can accelerate the spontaneous evolution of terrestrial landscapes. Forest disturbances such as tree harvesting, fires, and insect infestation greatly impact ecosystems and are a major input to carbon cycle simulation models. Remote sensing has been instrumental in detecting forest change; however, remotely sensed data and resulting classifications are not error free. In the present study, an algorithm implemented for the detection of forest disturbance in the western United States for the decadal period between 1990 and 2000 is validated. This is accomplished through a pixel-by-pixel assessment of the algorithm's ability to detect forest disturbance associated with fire events classified by burn severity. It was found that the forest disturbance algorithm has a total omission rate of 48% for low severity fires and 31% for high severity fires, and that the fire detection rate tends to drop with time, from 78% in 2000 to 33% in 1989 for high severity fires and from 63% in 2000 to 14% in 1989 for low severity fires.