KNUST Researchers have developed a scalable method to predict nitrogen dioxide (NO₂) pollution across Ghana using satellite data and machine learning, offering a new tool for managing air quality in regions with limited ground-based monitoring.

The study, published in npj Clean Air, was led by Dr. Prince Junior Asilevi, from the Department of Meteorology and Climate Science.

It combined data from the TROPOspheric Monitoring Instrument (TROPOMI) and NASA’s POWER meteorological datasets to examine seasonal and climatic influences on NO₂ concentrations.

Using Random Forest and XGBoost algorithms, the models achieved correlation scores of up to 0.92 and agreement indices of 0.96.

The researchers found that NO₂ levels peaked during Ghana’s dry season, underscoring the strong influence of seasonal variability.

They reported that wind speed, humidity and rainfall helped lower NO₂ concentrations, while higher temperatures were linked to increased levels of the pollutant in the country’s southern forest zone.

Predictive accuracy varied across climatic zones, with mean percentage differences ranging from 9.87% to 37.76%.

The team said the approach improves spatial resolution of NO₂ monitoring and provides a cost-effective framework for forecasting pollution in data-scarce environments.

They added that the findings are particularly relevant for Ghana, where urbanization and industrial activity are intensifying air quality challenges.

Other authors include: Enock Nyameasem Dzidzorm, Patrick Boakye, and Emmanuel Quansah.