Resumen
Data on the extension of urban areas are important for analyzing growth dynamics and to support the planning of transport and service provision. Satellite-based remote sensing has proven extremely useful, especially in cities that experience fast spatial growth. Different approaches to satellite-based mapping may, however, produce different results concerning urban categorization and delineation, often making direct comparison misleading. This study analyses four different satellite-based studies of urban land cover in Accra, Ghana and presents a new land cover map based on visual interpretation of segmented Sentinel-2 imagery. The methods and results, as well as the underlying definition of ?urban?, are compared and discussed. One method identifies exclusively areas with man-made, impervious surfaces, such as roads and buildings, as proxies for urban extent. Other methods aim to identify a broader set of land cover types, including green spaces, which are treated as part of the mixed urban fabric. Further differences are found in the way urban fringe areas under development are classified depending on their degree of urbanization, and in the distance threshold values used for defining the urban agglomeration. For the most recent maps we identify a difference in the measured size of the Accra agglomeration of almost 100%.