Advanced Journal of Environmental Science and Technology

ISSN 2756-3251

Advanced Journal of Environmental Science and Technology | Vol. 17 (5), pp. 397-409, May, 2026

Research Article

Low-Cost IoT Aerosol Sensor Networks and Physics-Informed Neural Networks for Hyperlocal PM2.5 Source Apportionment in Sub-Saharan Megacities

Ibrahim Musa¹*, Eleanor Vance², Sandeep Kulkarni³

¹Department of Physics, Ahmadu Bello University, Zaria, Nigeria
²School of Earth and Environment, University of Leeds, Leeds, United Kingdom
³Centre for Environmental Science and Engineering, Indian Institute of Technology Delhi, New Delhi, India

Corresponding Author: Ibrahim Musa — [email protected]

Abstract

Rapidly growing Sub-Saharan African cities face severe fine particulate matter (PM2.5) air pollution, yet sparse reference monitoring networks obscure neighborhood-level exposure disparities and emission sources. This study deployed 85 solar-powered optical particle counter nodes integrated with gas-phase sensors across Lagos, calibrated against gravimetric reference samplers using relative-humidity-corrected machine learning algorithms. Coupling multi-sensor time series with a physics-informed neural network (PINN) incorporating urban boundary-layer meteorology resolved spatial source contributions at 250-meter resolution. Source apportionment attributed 38% of ambient PM2.5 to diesel generator and vehicular exhaust, 29% to open solid waste and biomass burning, 21% to windblown crustal dust, and 12% to industrial emissions. High-temporal-resolution diurnal mapping pinpointed nocturnal waste-burning corridors, providing municipal environmental agencies with targeted enforcement metrics.

Keywords: PM2.5 source apportionment; low-cost air sensors; physics-informed neural networks; urban air quality; Sub-Saharan Africa; exposure mapping

Manuscript Timeline: Received 21 April, 2026; Revised 17 May, 2026; Accepted 28 May, 2026; Published 31 May, 2026.

Citation: Musa, I., Vance, E., and Kulkarni, S. (2026). Low-Cost IoT Aerosol Sensor Networks and Physics-Informed Neural Networks for Hyperlocal PM2.5 Source Apportionment in Sub-Saharan Megacities. Advanced Journal of Environmental Science and Technology, 17(5), pp. 397-409.