Air pollution poses an increasing public health risk in many African cities, where systematic monitoring is limited by the high cost of reference-grade instrumentation. Satellite-based sensors such as TROPOMI on Sentinel-5P provide global coverage of atmospheric pollutants, including nitrogen dioxide (NO<sub>2</sub>), but their limited vertical resolution complicates the estimation of surface-level concentrations.</p> <p>Palmes diffusion tubes offer a low-cost, low-tech alternative for measuring NO<sub>2</sub> at the ground level. Combining satellite observations with data from ground-based diffusion tubes presents a promising approach for generating regional air quality maps, particularly in resource-limited settings. This study reports on the implementation of a pilot network of Palmes tubes in Kumasi, Ghana. To ensure sustainability and scalability, a local air quality laboratory was established to prepare and analyse the tubes using locally sourced materials and equipment.</p> <p>Validation against tubes prepared and analysed by an accredited external laboratory demonstrated satisfactory agreement. Measurements from March 2025 revealed a wide range of NO<sub>2</sub> concentrations, from an average of 7 µg m<sup>-3</sup> in residential areas away from traffic and industry to peak values of 88 µg m<sup>-3</sup> at heavily trafficked intersections – substantially exceeding the WHO 2021 guideline limits of 25 µg m<sup>-3</sup> for 24-hour exposure and 10 µg m<sup>-3</sup> for annual mean concentrations.</p> <p>Averaged satellite data showed the accumulation of NO₂ plumes downwind of the city, though the relationship with surface-level measurements remains complex and needs further investigation. Column-to-surface ratios in Kumasi are significantly lower than those typically observed in European cities. TROPOMI underestimates the tropospheric NO<sub>2</sub> column densities by a factor of roughly 2.5 in March 2025, as the higher aerosol loading over Ghana reduces retrieval sensitivity in the troposphere, increasing the reliance on unrepresentative <em>a priori</em> profiles.