Can AI-powered urban green space monitoring help African cities build climate resilience while addressing environmental inequities

Can AI-powered urban green space monitoring help African cities build climate resilience while addressing environmental inequities Keywords: Green spaces - Deep Learning (DL) - Earth Observation (EO) - Climate resilience - Climate justice Because they regulate the temperature, maintain biodiversity, and enhance human well-being, urban green areas are essential to sustainable cities [1]. Rapid urbanization, however, is causing green spaces to disappear and become fragmented in many African cities such as Lagos-Nigeria, Nairobi-Kenya or Johannesburg-South Africa, increasing the risk of flooding, air pollution, and urban heat islands [2]. While state-of-the-art approaches [3] have successfully demonstrated the use of robust DL architectures for urban green space monitoring using optical imagery and data augmentation techniques, this proposal investigates a new approach that combines data from Sentinel2, Synthetic Aperture Radar (SAR) images but also population density variables in a unified framework. The goal is to create an equity-centered urban vegetation monitoring system that informs climate adaptation policy. We propose moving beyond vegetation detection to vegetation functionality, integrating AI segmentation with temporal EO data and population density indicators. Our approach will not only map green cover but also assess its climate resilience and accessibility for urban communities over the current year. This represents a shift from simple mapping (“green or not green”) towards actionable sustainability for cities in Africa. The proposed methodology is presented in four main sections: the combination of SAR and optical data, the data augmentation module, the integration of socio-environmental indicators and the actionable outcomes as shown on the flowchart below: In terms of models, PSPNet has been selected for the delineation because it works by combining global and local context through pyramid pooling of feature maps at multiple scales, making it suitable for green space segmentation because it can accurately capture both large park regions and small scattered vegetation within complex urban environments. The model (to be trained on several cities within the same country and tested on a new city in another country) outputs will be pixel-wise semantic segmentation maps which will undergo post-processing to provide green space fraction and accessibility radius as per World Health Organization (WHO) recommendation. By considering green spaces as essential infrastructure for sustainable cities, this research highlights the role of AI-driven EO in guiding evidence-based urban policies. We propose a pathway for African cities to leverage technology in building urban environments that are ecologically sustainable. Potential Issues: Combining different data sources can create challenges such as mismatched timing between SAR and optical images and varying spatial resolutions. Using population density as a weighting factor may also introduce socioeconomic biases, limiting how well the findings apply across diverse African urban contexts. [1] Godoi, N. M. I., Gomes, R. C., & Longo, R. M. (2025). Contributions of urban green spaces to cities: A literature review. Sustainable Environment, 11(1). https://doi.org/10.1080/27658511.2025.2464418 [2] Innocent Chirisa. (2008). Population growth and rapid urbanization in Africa: Implications for sustainability. ResearchGate, 10. https://www.researchgate.net/publication/43090985PopulationgrowthandrapidurbanizationinAfricaImplicationsforsustainability [3] Miao Zhang, Hajra Arshad, Manzar Abbas, Hamzah Jehanzeb, Izza Tahir, Javerya Hassan, Zainab Samad, Rumi Chunara. Quantifying greenspace with satellite images in Karachi, Pakistan using a new data augmentation paradigm. ACM Journal on Computing and Sustainable Societies, 2025; DOI: 10.1145/3716370