Large-scale slum mapping in sub-Saharan Africa's major cities: Remote sensing and deep learning reveal strong slum growth in the urban periphery between 2016 and 2022

Abstract

Around half of sub-Saharan Africa's urban population lives in slums, yet data on the spatiotemporal development of slums remains scarce, impeding policies to alleviate urban poverty and inequality. We propose a solution to this problem by applying deep learning to open-access satellite imagery to map slums in 529 major cities across sub-Saharan Africa and track their spatiotemporal development. Our model produced 10m resolution ‘slum probability maps’ allowing timely and cost-effective tracking of slum growth. On this basis, we estimated that in 2022 the share of the urban population living in slums exceeded 50% in 274 cities, and in 84% of cities this share increased between 2016 and 2022, most severely in Middle and West Africa. Slum growth occurred primarily in the urban periphery, which tends to be missed in survey-based slum monitoring.