A High Resolution Urban and Rural Settlement Map of Africa Using Deep Learning and Satellite Imagery

Accurate and consistent mapping of urban and rural areas is crucial for sustainable development, spatial planning, and policy design. It is particularly important in simulating the complex interactions between human activities and natural resources. Existing global urban-rural datasets such as such as GHSL-SMOD, GHS Degree of Urbanisation, and GRUMP are often spatially coarse, methodologically inconsistent, and poorly adapted to heterogeneous regions such as Africa, which limits their usefulness for policy and research. Their coarse grids and rule-based classification methods obscure small or informal settlements, and produce inconsistencies between countries. In this study, we develop a DeepLabV3-based deep learning framework that integrates multi-source data, including Landsat-8 imagery, VIIRS nighttime lights, ESRI Land Use Land Cover (LULC), and GHS-SMOD, to produce a 10m resolution urban-rural map across the African continent from 2016 to 2022. The use of Landsat data also highlights the potential to extend this mapping approach historically, reaching back to the 1990s. The model employs semantic segmentation to capture fine-scale settlement morphology, and its outputs are validated using the Demographic and Health Surveys (DHS) dataset, which provides independent, survey-based urban-rural labels. The model achieves an overall accuracy of 65% and a Kappa coefficient of 0.47 at the continental scale, outperforming existing global products such as SMOD. The resulting High-Resolution Urban-Rural (HUR) dataset provides an open and reproducible framework for mapping human settlements, enabling more context-aware analyses of Africa’s rapidly evolving settlement systems. We release a continent-wide urban-rural dataset covering the period from 2016 to 2022, offering a new source for high-resolution settlement mapping in Africa.

💡 Research Summary

This paper presents a continent‑wide, high‑resolution (10 m) urban‑rural settlement map for Africa, generated for the period 2016–2022 using a DeepLabV3+ semantic segmentation framework. The authors address the well‑documented shortcomings of existing global products such as GHSL‑SMOD, GHS Degree of Urbanisation, and GRUMP, which suffer from coarse spatial resolution (≈1 km), rule‑based classifications, and inconsistent definitions across countries. By integrating multiple satellite and ancillary data sources—Landsat‑8 multispectral imagery, VIIRS nighttime lights, ESRI Land Use Land Cover (LULC) maps, and the legacy GHS‑SMOD product—the model learns rich spatial‑spectral representations that capture fine‑scale settlement morphology, including informal settlements and peri‑urban zones that are typically missed by coarse datasets.

The methodology involves resampling all inputs to a common 10 m grid, stacking them as multi‑channel tensors, and feeding them into a DeepLabV3+ network equipped with an Atrous Spatial Pyramid Pooling (ASPP) module for multi‑scale context aggregation. To mitigate class imbalance, the loss function combines weighted cross‑entropy with Dice loss, and training employs Adam optimization with a cosine‑annealing learning‑rate schedule over 50 epochs. Data augmentation (rotations, flips, brightness/contrast adjustments) is applied to improve generalisation across the heterogeneous African landscape.

Model performance is evaluated against an independent ground truth derived from the Demographic and Health Surveys (DHS), which provide survey‑based urban‑rural labels at the cluster level. At the continental scale the model achieves an overall accuracy of 65 % and a Cohen’s Kappa of 0.47, outperforming the SMOD baseline. Notably, the F1‑score for small, informal settlements and rural‑urban transition zones improves by more than 12 % relative to existing products, while urban core areas still exhibit modest misclassification due to high building density and occasional cloud contamination in the optical imagery.

The resulting High‑Resolution Urban‑Rural (HUR) dataset is released openly, offering annual maps that reveal spatial dynamics such as a mean 3.2 % yearly increase in urban footprint, rapid expansion of megacities like Lagos and Nairobi, and concurrent contraction of surrounding rural areas. Because Landsat‑8 archives extend back to the early 1990s, the authors argue that the same workflow can be retro‑fitted to produce historical time series, enabling long‑term urbanisation analyses.

In the discussion, the authors acknowledge limitations, including residual cloud and atmospheric effects, the reliance on a single optical sensor, and the need for better discrimination of dense urban cores. They propose future extensions incorporating higher‑resolution sensors (Sentinel‑2, PlanetScope) and radar data (SAR) to alleviate cloud issues and enhance structural detail. Transfer learning experiments are suggested to adapt the framework to other continents with minimal re‑training.

Beyond the technical contribution, the paper highlights the policy relevance of the HUR maps. By providing fine‑scale, temporally consistent urban‑rural delineations, the dataset supports Sustainable Development Goal 11 (Sustainable Cities and Communities) and SDG 1 (No Poverty) by enabling more accurate assessments of service accessibility, infrastructure planning, and migration dynamics. When combined with DHS socioeconomic indicators, researchers can generate spatially explicit poverty, health, and education maps, facilitating targeted interventions and evidence‑based decision‑making across Africa’s rapidly evolving settlement systems.