SatNO₂ V4.GL.01
Global Monthly 0.01x0.01 Surface NO₂ (2005–2023)
We estimate global monthly surface NO₂ concentrations at fine spatial resolution (0.01° x 0.01°) over 2005-2023 by developing, optimizing, and applying a highly efficient gradient boosting decision tree (LightGBM) framework underpinned with multi-source datasets, including geophysical process-based a priori information that leverages simulations with the GEOS-Chem chemical transport model, road‐network metrics, land‐cover features, and other relevant environmental predictors over 2005-2023. Evaluated against 9,913 ground monitors across 66 countries, this monthly surface NO₂ estimate achieves significant long-term performance.
Reference:
Yu, Y., Shen, S., Li, C., Zhang, Y., Martin, R., Haihui, Z., Chatterjee, D., & Singh, I. (2026). Global Monthly 0.01x0.01 Surface NO2 (2005–2023) [Dataset]. Zenodo. https://doi.org/10.5281/zenodo.19740022
Scientific Datasets:
The global NO₂ dataset is archived on Zenodo as two companion records covering January 2005 through December 2023 at 0.01° x 0.01° spatial resolution (https://doi.org/10.5281/zenodo.19740022). Each record contains 19 annual tar archives that together hold 228 monthly NetCDF-4 files. Annual archives are named Global_Surface_NO2_YYYY.tar and unpack to monthly files names NO2_surface_monthly_0p01deg_YYYYMM.nc. Each NetCDF file is self-describing in accordance with the CF-1.8 and ACDD-1.3 conventions, so it can be opened directly by standard geoscientific software (e.g. Panoply, Xarray). Files in both records cover the global domain with latitude from -59.995° to 69.995° (lat size 13,000) and longitude from -179.995° to 179.995° (lon size 36,000), with grid increments of 0.01°. Two auxiliary files accompany the archives at the same Zenodo record: global_monthly_no2_2005-2023.csv, the monthly in-situ NO₂ ground-based measurements used for training and evaluation, and unique_countries.txt, the list of countries represented in those observations. Each monthly file contains two two-dimensional data variables on an identical (lat, lon) grid together with their coordinate variables and a crs grid-mapping variable, stored with zlib level 4.
The NO₂ and NO2_uncertainty variables declare _FillValue = NaN so that missing-data pixels (e.g., permanent water bodies, regions outside the training domain) are handled correctly by CF-aware tools.
Usage note. Because both NO₂ and NO2_uncertainty are provided in the same file on the same grid and in the same physical units (ppb), relative uncertainty can be obtained simply as NO2_uncertainty / NO₂.
The dataset is available at: Global Monthly 0.01x0.01 Surface NO₂ (2005–2023)