Machine learning-guided integration of fixed and mobile sensors for high resolution urban PM2.5 mapping

Abstract Urban areas exhibit significant gradients in Fine Particulate Matter (PM2.5) concentration variability. Understanding the spatiotemporal distribution and formation mechanisms of PM2.5 is crucial for public health, environmental justice, and air pollution mitigation strategies. Here, we utilized machine learning and integrated air quality sensor monitoring networks consisting of 200 mobile cruising vehicles and 614 fixed micro–stations to reconstruct PM2.5 pollution maps for Jinan’s urban area with a high spatiotemporal resolution of 500 m and 1 h. Our study demonstrated that pollution mapping can effectively capture spatiotemporal variations at the urban microscale. By optimizing the spatial design of monitoring networks, we developed a cost-effective air quality monitoring strategy that reduces expenses by nearly 70% while maintaining high precision. The results of multi-model coupling indicated that secondary inorganic aerosols were the primary driving factors for PM2.5 pollution in Jinan. Our work offers a unique perspective on urban air quality monitoring and pollution attribution.