Predicting nitrogen dioxide column density from the International Wealth Index with a convolutional neural network in Africa
Abstract
In the pursuit of improving socioeconomic living conditions of their populations, many nations struggle to meet their nationally determined contributions (NDCs) under the Paris agreement, and that shortfall makes accountability for emissions reductions critical. The long atmospheric lifetimes of key gases such as carbon dioxide (CO ₂ ) and methane (CH ₄ ) hinder traditional greenhouse gas (GHG) monitoring, because what any one place emits adds only a small excess to a large, well-mixed background. Nitrogen dioxide (NO ₂ ) behaves differently: it is short-lived and highly variable in space and time, so satellite instruments resolve its patterns close to source regions. Across the African continent, mapped temporal correlations between NO ₂ and CO ₂ are moderate in magnitude but spatially structured and consistently positive. Although we do not use that association to mechanically convert NO ₂ observations into CO ₂ , we report it as contextual evidence towards NDCs. Given this context, this study asks the following question: to what extent can socioeconomic development at the neighbourhood scale predict NO ₂ column densities (the amount of NO ₂ in the air column above that neighbourhood) across Africa? We apply a convolutional neural network (CNN) to gridded International Wealth Index (IWI) estimates and to NO ₂ observations from the tropospheric Monitoring Instrument (TROPOMI), and obtain an average predictive accuracy of R2 = 0.56 across the African continent. Because IWI can be projected forward, we also present an illustrative 2030 scenario as a methodological demonstration of the relationship between socioeconomic development and NO ₂ trajectories. These findings indicate that gridded wealth estimates carry predictive information about NO ₂ column density in Africa. Such estimates offer a lightweight and scalable screening tool that complements process-based atmospheric monitoring rather than replacing it.
Related
- Independent Verification of the European Climate Law's Effect on Nitrogen Dioxide Pollution: A Satellite-Derived, Difference-in-Differences Analysis
- Simulated heatwave reduces resilience to subsequent combined high‐temperature and high‐carbon‐dioxide stress in a diatom
- Convolutional neural networks for enhanced classification of VLF rock rupture precursors
- Efficacy of seaweed-based carbon dioxide removal reduced by iron limitation and nutrient competition with phytoplankton
- The Rich Chemistry of Ionized Methanol With Sulfur Dioxide: Pathways to Prebiotic and Elusive Species in Earth's Atmosphere and Space
BibTeX
@article{grieco2026predicting,
title = {{Predicting nitrogen dioxide column density from the International Wealth Index with a convolutional neural network in Africa}},
author = {Francesco Grieco and Mohammad Kakooei and Adel Daoud},
journal = {Scientific Reports},
year = {2026},
month = oct,
doi = {10.1038/s41598-026-73707-0},
url = {https://www.nature.com/articles/s41598-026-73707-0},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Predicting nitrogen dioxide column density from the International Wealth Index with a convolutional neural network in Africa
AU - Francesco Grieco
AU - Mohammad Kakooei
AU - Adel Daoud
PY - 2026
DA - 2026/10/15
JO - Scientific Reports
DO - 10.1038/s41598-026-73707-0
UR - https://www.nature.com/articles/s41598-026-73707-0
AB - In the pursuit of improving socioeconomic living conditions of their populations, many nations struggle to meet their nationally determined contributions (NDCs) under the Paris agreement, and that shortfall makes accountability for emissions reductions critical. The long atmospheric lifetimes of key gases such as carbon dioxide (CO ₂ ) and methane (CH ₄ ) hinder traditional greenhouse gas (GHG) monitoring, because what any one place emits adds only a small excess to a large, well-mixed background. Nitrogen dioxide (NO ₂ ) behaves differently: it is short-lived and highly variable in space and time, so satellite instruments resolve its patterns close to source regions. Across the African continent, mapped temporal correlations between NO ₂ and CO ₂ are moderate in magnitude but spatially structured and consistently positive. Although we do not use that association to mechanically convert NO ₂ observations into CO ₂ , we report it as contextual evidence towards NDCs. Given this context, this study asks the following question: to what extent can socioeconomic development at the neighbourhood scale predict NO ₂ column densities (the amount of NO ₂ in the air column above that neighbourhood) across Africa? We apply a convolutional neural network (CNN) to gridded International Wealth Index (IWI) estimates and to NO ₂ observations from the tropospheric Monitoring Instrument (TROPOMI), and obtain an average predictive accuracy of R^2 = 0.56 across the African continent. Because IWI can be projected forward, we also present an illustrative 2030 scenario as a methodological demonstration of the relationship between socioeconomic development and NO ₂ trajectories. These findings indicate that gridded wealth estimates carry predictive information about NO ₂ column density in Africa. Such estimates offer a lightweight and scalable screening tool that complements process-based atmospheric monitoring rather than replacing it.
SN - 2045-2322
ER -