Monitoring of Savanna degradation in Namibia using landsat TM/ETM+ data

Abstract

Detection and quantification of degradation processes as a first step for understanding and development of prevention and mitigation actions is of international interest as the described processes are observed in semi-arid landscapes globally. However, the accurate assessment of savanna degradation using remote sensing techniques has turned out difficult, as the described processes are frequently overlaid by ecosystem-inherent variability in response to the usually highly variable rain falls (Wessels et al. 2007). Furthermore, local vegetation decrease frequently turns out to be a result of recent livestock grazing impact at the time of image acquisition, rather then true vegetation degradation. In the presented work, researchers aimed to develop a bitemporal change detection approach that takes into account these difficulties. The test site was a landscape mosaic of different types of thorn bush savanna in central Namibia. Using a set of 7 Landsat TM and ETM+ images covering the study area in +/- 5 year intervals from 1984 - 2003, researchers developed a spectral decision tree classificator sensitive to increase or decrease independent from the vegetation type.