Enhancing ATMO-Street model accuracy through emission source analysis using a dense sensor network: a Warsaw case study

Abstract

Urban air quality models are essential for managing particulate matter (PM) pollution, yet their accuracy is often limited by sparse monitoring networks and outdated emission inventories. This study presents a scalable framework for improving PM10 and PM2.5 modelling through the use of high-resolution emissions inventories and enhanced validation based on calibrated low-cost sensor networks. Using Warsaw city in Poland as a representative case study, we demonstrate that incorporating high-resolution residential heating emissions from the Central Register of Emissions from Buildings (CEEB) and calibrating road dust resuspension parameters led to concentration reductions of up to 20% in urban hotspots and reduced the prediction bias for PM2.5 by 57% at key locations. Notably, the Revised scenario resolved substantial overestimations in districts where incorrect fuel classifications had previously caused overestimations. However, persistent winter overestimations and the inability to fully capture extreme PM10 peaks in dry months highlight ongoing challenges, particularly in modelling resuspension dynamics under dry conditions. Our findings reveal that low-cost sensors, when rigorously calibrated, can extend spatial coverage and improve model validation, though they may underestimate extreme pollution events. The methodological advances presented here are broadly applicable to cities worldwide, particularly those facing similar challenges of diverse emission sources and limited regulatory monitoring. This integrated approach supports more accurate forecasting and targeted mitigation strategies, offering a scalable solution for urban environments seeking to achieve international air quality standards.