Forest Cover Change in the Ohangwena Region, Northern Namibia: a Case Study Based on Multitemporal Landsat Images and Aerial Photography

Abstract

Forest cover in northern Namibia was estimated by analysing Landsat MSS images from 1973, 1981 and 1989 as well as Landsat TM images from 1992 and 1996. Aerial photography from 1970 and 1992 was used as reference data for the classification. The spectral separation of sample areas was best for the Landsat images of 1981 and 1992, and lowest for the Landsat image of 1973. The accuracy of classifying training-set pixels was substantial when only non-forest and forest areas were classified, and moderate when the classification involved six different biomass densities. In the eastern part of the test area, which is dominated by Tree Savanna and Woodland, there was a clear decrease in forest area between 1981 and 1992, the annual change being −0,5%. The factors which may have affected the classification of satellite images include problems in reference data, seasonal differences in vegetation, differences in weather conditions, abundance of mixed pixels as well Colophospermum mopane, a tree species with a dist...