Object-oriented approaches to mapping channel habitat on the lower Congo River

Information about large river channels is required for assessments of ecologically and economically important fish assemblages, but field mapping can be an expensive and time-consuming process. Classification of fluvial fish habitats using remotely sensed images is a potential alternative. This study examines whether widely available satellite images can be used to map habitat on the Lower Congo River in Western Africa using an object-oriented approach. One Landsat 7 image and three ASTER images of the river were classified using object-oriented techniques and the results were compared to existing topographic maps. Although insufficient ground truth data were available for a numerical accuracy assessment, the habitat maps produced from the ASTER images correctly identified all major turbulent areas, islands, and rock formations. This study concludes that an object-oriented approach provides procedural benefits over pixel-based approaches, and that the ability to incorporate spatial context is a major advantage in habitat classification.