Using Landsat images to detect oak decline in the Mark Twain National Forest, Ozark Highlands

Abstract Following the severe drought in 1999–2000 there was a widespread outbreak of oak decline in the Ozark Highlands. Over 400,000 ha of dead and dying oak trees were observed by the USDA Forest Service in this region. Although oak forests that are dead can be easily interpreted from air photos or classified from satellite images, it is difficult to detect dying trees that are still green but will die back or recover in the following years. In this study, we applied a normalized difference water index (NDWI) to map the continuous forest dynamics related to oak decline. The Landsat TM image in 1992 and the ETM+ image in 2000 were processed to calculate the differential NDWI which revealed moisture variation primarily caused by the drought and the associated red oak borers. A simple thresholding method was used to map oak dying back, recovery and non-change areas in the study area. The died-back areas were extracted from the modified land use/land cover maps created by the Missouri Resource Assessment Partnership (MoRAP). The forest dynamics map was compared with the online FIA database in which tree species at randomly selected sites were recorded in 1989 and 2003. The overall accuracy of forest dynamics mapping with remote sensing imagery was 75.95%. The user's accuracy of dying/recovery area mapping was also high although the producer's accuracy is questionable because of the limitation in ground data collection. The continuous dying/recovery map in this study could provide valuable information on the prediction of oak decline and evaluation of damage when another period of environmental stresses occurs.