Monthly Gridded Concentrations of Six Major Air Pollutants Across China (2003–2025)

This dataset provides monthly gridded concentrations of six major ambient air pollutants across China from January 2003 to December 2025 at a spatial resolution of 0.1° × 0.1°. The six pollutants include PM2.5, PM10, CO, NO2, SO2, and O3. The dataset was developed to support long-term air quality assessment, atmospheric environment studies, exposure analysis, health impact assessment, and compound air pollution research. It provides spatially and temporally continuous pollutant fields covering mainland China over a 23-year period. The monthly fields were derived from a high-resolution daily air pollution dataset generated using a deep learning framework based on a multi-task Long Short-Term Memory (MP-LSTM) model. The model jointly predicts six air pollutants by integrating ground observations, meteorological reanalysis data, satellite-derived aerosol information, vegetation indicators, population density, topographic variables, and emission inventories. Monthly means were calculated from the corresponding daily products. Dataset characteristics: • Spatial coverage: China (mainland China) • Spatial resolution: 0.1° × 0.1° • Temporal coverage: January 2003 – December 2025 • Temporal resolution: Monthly • Pollutants: PM2.5 (μg m⁻³) PM10 (μg m⁻³) CO (mg m⁻³) NO2 (μg m⁻³) SO2 (μg m⁻³) O3 (μg m⁻³) • Data format: NetCDF4 Each NetCDF file contains monthly mean concentrations for a single calendar year, including 12 monthly layers and six pollutant variables. Geographic coordinates are provided in latitude–longitude format using the WGS84 coordinate reference system. File naming convention: monthly_mean_2003.ncmonthly_mean_2004.nc...monthly_mean_2025.nc Potential applications include: • Long-term air quality trend analysis • Air pollution exposure assessment • Environmental epidemiology • Climate and atmospheric chemistry studies • Compound pollution characterization • Machine learning and Earth system modeling Users are encouraged to cite this Zenodo record when using the dataset in scientific publications, reports, or derived products.