Forest disturbances such as bark beetle outbreaks are increasing in severity and extent across western North America. Classification of remote sensing imagery is a powerful way to analyze and detect large-scale disturbances. We used a temporal sequence of four Landsat TM images (1991, 1995, 1999, and 2003) to detect the spatiotemporal change in spectral response of Engelmann spruce (Picea engelmannii Parry ex. Engelm.) killed by an unprecedented spruce beetle outbreak in southern Utah, USA. After co-registration and masking out non-vegetation the Disturbance Index (DI) was calculated for each image. DI values associated with Engelmann spruce mortality, determined by comparing each image to a no outbreak baseline image, were then used to classify the images. Dendrochronologically determined dates of spruce death collected from across the outbreak area were used to assess the ability of the DI to accurately differentiate stands of dead spruce from live conifer forest. The overall classification accuracy of the DI varied from 80 to 82% while the accuracy to detect spruce beetle-killed spruce varied from 59 to 71%. Both user's and producer's accuracy to classify beetle infested stands increased over the temporal sequence of image dates. However, confusion matrix-derived statistics varied by image date. Consistent with previous studies, the spruce beetle outbreak began building in multiple, seemingly independent locations across the study area. Over time, areas attacked earlier in the outbreak enlarged and coalesced on the landscape.