| Summary: | Includes bibliographical references (leaves 122-124). === It is well established that land cover information is an essential component in the creation of spatial information systems. Lack of current land cover information constitutes a weakness in land resource management especially in developing countries like Uganda. In response to this need, the thesis reports on a case study on tropical forest mapping in Uganda. The geographic area of study is the Bwindi Impenetrable National Park located in the southwest of the country. Digital image processing techniques were applied to SPOT and Landsat Imagery using Erdas Imagine (8.4) and Arc View GIS software. A combination of original and derived bands (Principal Components, Tasseled Cap and Texture Analysis) was used for the image analysis. Both supervised and unsupervised classification approaches were used. The optimal combination of bands was selected on the basis of secondary correlation analysis of the derived as well as original bands. Further identification of the best bands was based on separability indices. With the band combination selected, four main land cover classes were identified in the forest i.e. dense evergreen forest, evergreen forest, mixed rangeland and brush land. In addition to this, three other land use types were extracted from the imagery within the neighborhood of the forest and these were subsistence farmlands, plantation farmlands (tea plantation) and woodland. The results were confirmed by post processing field inspection.
|
|---|
