Upland oak-hickory forests in Arkansas, Missouri, and Oklahoma experienced oak decline in the late 1990s and early 2000s during an unprecedented outbreak of a native beetle, the red oak borer (ROB), Enaphalodes rufulus (Haldeman). Although remote sensing supports frequent monitoring of continuously changing forests, comparable in situ observations are critical for developing an understanding of past and potential ROB damage in the Ozark Mountains. We categorized forest change using a normalized difference water index (NDWI) applied to multitemporal Landsat TM and ETM+ imagery (1990, 2001, and 2006). Levels of decline or growth were categorized using simple statistical thresholds of change in the NDWI over time. Corresponding decline and growth areas were then observed in situ where tree diameter, age, crown condition, and species composition were measured within variable radius plots. Using a machine learning decision tree classifier, remote sensing-derived decline and growth was characterized in terms of in situ observation. Plots with tree quadratic mean diameter at breast height were categorized remotely as in severe decline. Landsat TM/ETM+-based NDWI derivatives reveal forest decline and regrowth in post-ROB outbreak surveys. Historical and future Landsat-based canopy change detection should be incorporated with existing landscape-based prediction of ROB hazard.