AFRISCAPE: An African Urban Settlement and Environmental Surface Benchmark

AFRISCAPE is a geospatial reference dataset created to support urban environmental research across Africa. The dataset includes 8,428 reference samples collected from 61 urban areas distributed across the nine IPCC African reference regions, capturing diverse climatic, geographic, and urban contexts across the continent. AFRISCAPE adopts a task-oriented hybrid LULC taxonomy consisting of five classes: formal built-up (2,939 samples; 34.87%), water bodies and wetlands (2,222; 26.36%), vegetation (1,579; 18.74%), informal built-up (1,172; 13.91%), and bare and exposed surfaces (516; 6.12%). Reference samples were manually annotated through visual interpretation of very-high-resolution imagery using a custom Google Earth Engine-based annotation application. Specific attention is paid to the differentiation of formal and informal built-up structures, which is rarely addressed in available continent-wide urban and LULC datasets. AFRISCAPE is intended to serve as a tool for reproducible benchmarking of geospatial artificial intelligence and foundation models, as well as a spatial reference resource for urban environment research, such as urban heat and cooling studies, environmental vulnerability, climate resilience, and urban heat inequality.