Hourly Seamless Retrieval of Near-Surface PM2.5 and O3 Concentrations across China from 2018 to 2023

In recent years, China’s air pollution control efforts have significantly reduced fine particulate matter (PM_2.5) concentrations. However, ozone (O₃) pollution has emerged as a key issue, becoming the second major pollutant affecting air quality. During certain periods, simultaneous high concentrations of PM_2.5 and O₃ lead to “dual-high” compound pollution. Understanding the dynamic evolution of these pollutants is essential for precise control strategies.

Existing observational data on PM_2.5 and O₃ are insufficient for research and applications. Ground-based monitoring stations provide temporally continuous data but have limited spatial coverage, while geostationary satellites offer wide spatial coverage but suffer from data gaps due to cloud interference, retrieval algorithm limitations, and the lack of nighttime observations. These challenges highlight the need for spatiotemporally continuous, all-weather data.

This study develops a high-precision retrieval model for near-surface PM_2.5 and O₃ concentrations, integrating AI algorithms with multispectral data from Himawari-8/9, reanalysis meteorological data, and geographic parameters. By incorporating the spatiotemporal autocorrelation of pollutants as a physical constraint, the model innovatively combines Gaussian smoothing adjustment and discrete cosine transform to create a data fusion framework. This framework generates seamless, all-weather datasets, addressing data gaps and correcting systematic biases in satellite retrievals.

Independent validation shows strong model performance, with hourly R² values exceeding 0.85. Using retrieval results from 2018 to 2023, we analyzed the spatiotemporal distribution of PM_2.5 and O₃ across four major urban agglomerations in China (Beijing-Tianjin-Hebei, Yangtze River Delta, Pearl River Delta, and Central China) before, during, and after the COVID-19 outbreak. The findings reveal the effects of emission reductions and the post-pandemic pollution recovery.

This study demonstrates the potential of integrating remote sensing, AI, and mathematical modeling to achieve spatiotemporally continuous monitoring of PM_2.5 and O₃. It offers critical support for air pollution control and environmental policy evaluation.