Work

Hidden and fragmented patterns of morphological informality across African cities revealed by open geospatial data

Year: 2026
Type: article
Abstract: Informal settlements are a major yet poorly measured form of urbanization, particularly where small or fragmented clusters are embedded within otherwise formal urban fabric. Here, we develop a scalable geospatial framework for mapping morphological informality across African cities using open satellite imagery, global building data, and eight city-specific machine-learning models combined by equal averaging. Applied to all 84 African cities with projected populations exceeding one million in 2025, the framework generates 10 m probability surfaces. Mean leave-one-city-out F1 at training-selected thresholds was 0.876; independent validation using 1910 reference points yielded AUC of 0.972. Building-presence profiles were the leading SHAP contributor in all eight held-out cities, with height and spectral context providing additional signals. These results support retaining neighborhood building distributions alongside vertical structure and vegetation when mapping across diverse urban contexts. The share of selected area in patches smaller than 1 km 2 increased from 38.0% at p ≥ 0.2 to 72.4% at p ≥ 0.8. WorldPop 2023 allocated fewer people than community-reported counts in 89.0% of 308 settlement polygons, with a median WorldPop/community ratio of 0.263. Community records span 2013–2020, so the comparison diagnoses allocation discrepancies and motivates assessing settlement-specific population densities using contemporaneous evidence. The probability surfaces describe physical morphology, not tenure or household deprivation. Together, the findings identify shared morphological signals and population-allocation discrepancies that can guide local validation, surveys and future population modelling.
Language: English
Subfield: Urban Studies
Open Access status: hybrid
Estimated APC paid: $3,740
OpenAlex ID: W7163177772
FWCI:
Cites: 34
Cited by:
Related to: 10