Assessing land cover change in Namibia's Kavango East region: a multi-date object approach

Land cover change is a global issue but its effects can be particularly severe in developing countries such as Namibia, by affecting the ecological functions of ecosystems and hence the sustainability of its development. Namibia’s arid conditions, due to low rainfall and high evapotranspiration rates, coupled with annual savannah fires, have resulted in a heterogenous landscape composed of a mixture of trees, shrubs and herbaceous plants. As a result, land cover maps are often inaccurate at the pixel level. Despite their relatively high accuracy, object-based image analyses are yet to be exhaustively applied to the dry tropical forests of Southern Africa.  The purpose of this study was to apply a multi-date object-based approach to land cover change, in order to determine its extent and dynamics in the heterogenous landscape of Kavango East, one of the regions with the highest forest cover in Namibia. Multi-date segmentation, mean band values and image differentiation were used to detect land cover changes in four periods (1990, 2000, 2009 and 2016). The most common land conversion for all the periods was from forest to cropland. In 1990, forests covered 58% of the land but by 2016, this had dropped to 55%. Meanwhile, cropland covered 3% of the study area in 1990 and had doubled to 6% by 2016. The novel approach used in this study has produced promising results compared to traditional methods, which are prone to errors in detecting post-classification changes. The method can therefore be recommended for long term monitoring of land cover and land use change in areas with similar environmental and biophysical conditions.