Hybrid CNN–LSTM Model for Spatiotemporal Prediction of PM2.5 Concentrations in Selected African Cities

Project Title: Hybrid CNN–LSTM for Spatiotemporal Prediction of PM2.5 Concentrations in Selected African Cities Authors: Ugochi Adaku Okengwu, Olalekan Popoola, Alfred Aminayanate Bob-Manuel Abstract Air quality has become a growing environmental concern in African cities due to rapid urbanisation and economic development. In particular, particulate matter (PM2.5), a major airborne contaminant, harms human health by causing respiratory diseases and other ailments. African cities are generally constrained by limited capacity and infrastructure for comprehensive air quality management. This research presents a data-driven framework to predict a PM2.5-based Air Quality Index (AQI) using a deep-learning, multi-model approach involving CNN, and LSTM. The data-driven framework employs PM2.5 together with different spatial and environmental precursors extracted from the air quality sensors situated in Lagos, Port Harcourt and Nairobi - three major African cities to exploit the temporal and spatial variations of air pollution. These selected locations reflect disparities in local climates, varying degrees of coverage, and pollution regimes, thereby minimising the computational workload of large multi-city datasets. This investigation uses four essential datasets: PM2.5 monitoring from OpenAQI, a Digital Elevation Model from Copernicus, Terrain characteristics from ERA5, and land-cover information from Copernicus WorldCover. Appropriate predictive evaluation metrics will be applied to determine the accuracy of models in forecasting PM2.5 and corresponding AQI in real time, so that the best model can then be selected. These metrics, include RMSE, MAE, R2, Accuracy, F1, Precision, Recall, and others. This work seeks to raise awareness of the adverse effects of PM2.5 while contributing to policymaking and prevention measures for future hazards. Also contributes to affordable, context-specific air pollution monitoring for 3 urban cities in Africa, and the developed predictive system assists public policymakers, environmental managers, and the health sector in developing sustainable environmental policies and programmes.