Brazilian Air Quality Dataset – BRAQ: PM2.5 (4 km, monthly, 2014–2023)

BRAQ-PM25 v1.0 Brazilian Air Quality Dataset– PM2.5Monthly PM2.5 concentrations over Brazil at 4 km resolution (2014–2023) ---------------------------------------------------------------------------------------------------------- 1. Overview BRAQ-PM25 is a high-resolution, monthly gridded dataset of surface PM2.5 concentrations covering the entire Brazilian territory for the period 2014–2023, with a spatial resolution of about 4 km. The dataset was developed to address the limited spatial coverage and uneven distribution of Brazil’s reference air-quality monitoring network. It integrates complementary information from reference monitoring stations, low-cost sensor networks, and satellite-based PM2.5 products into a single hybrid surface concentration field. This dataset supports applications in air-quality assessment, public-health studies, environmental epidemiology, and climate and land-use interaction analyses, particularly in regions lacking continuous ground monitoring. ---------------------------------------------------------------------------------------------------------- 2. Dataset content and formats The dataset is distributed in two complementary formats, designed to serve different user communities and analytical workflows. 2.1 NetCDF files The primary format is annual NetCDF files, each containing monthly mean PM2.5 concentrations for one calendar year. File structure PM25_BR_4km_monthly_2014.nc PM25_BR_4km_monthly_2016.nc … PM25_BR_4km_monthly_2023.nc Variables and dimensions Variable: pm25 Units: µg m⁻³ Spatial dimensions: x (easting, meters) × y (northing, meters) CRS: Lambert Cylindrical Equal Area (CEA), datum WGS84; units: meters Temporal dimension: time (monthly, one timestep per month) Each NetCDF file contains a single PM2.5 variable (pm25), with time represented explicitly along the temporal dimension. 2.2 GeoTIFF files (auxiliary format) For users working primarily with GIS-based workflows or visualization tools, the dataset is also provided as monthly GeoTIFF files, organized and distributed as annual ZIP archives. GeoTIFF package naming convention Monthly PM2.5 over Brazil - 2014 - TIFF.zip Monthly PM2.5 over Brazil - 2016 - TIFF.zip … Monthly PM2.5 over Brazil - 2023 - TIFF.zip Each ZIP archive contains 12 GeoTIFF files, corresponding to the monthly mean PM2.5 concentrations for that year. The GeoTIFF format is provided for compatibility and ease of visualization in GIS software (e.g., QGIS, ArcGIS). ---------------------------------------------------------------------------------------------------------- 3. Methodological summary The BRAQ-PM25 dataset was generated using a hybrid data-fusion approach, as described in detail in the associated article. 3.1 Input data sources Reference air-quality stations (AQS) Hourly PM2.5 measurements from Brazil’s official monitoring networks, compiled primarily from the Institute for Energy and Environment (IEMA) database (2014–2023), following national data-quality criteria. Low-cost sensors Outdoor PurpleAir PM2.5 sensors subjected to rigorous quality control, including sensor screening, minimum data-availability thresholds, and exclusion of implausible values. Satellite-based PM2.5 estimates Monthly gridded PM2.5 products derived from satellite aerosol optical depth (AOD), chemical transport modeling, and statistical fusion, specifically the V6.GL.02 product from the Atmospheric Composition Analysis Group (ACAG), based on a convolutional neural network (CNN; available at: https://sites.wustl.edu/acag/surface-pm2-5/). 3.2 Fusion procedure (summary) The satellite-based PM2.5 product was resampled from 1 km to 4 km resolution using spatial averaging. Reference stations and low-cost sensors were integrated using area-weighted buffers, adjusting satellite values where ground observations were available. Reference stations were given priority over low-cost sensors where both overlapped the same grid cell. Satellite estimates were preserved in regions without ground observations, ensuring wall-to-wall national coverage. This approach improves spatial representativeness while retaining physically consistent background fields in data-sparse regions. ---------------------------------------------------------------------------------------------------------- Citation If you use this dataset, please cite: Rudke, A. P., Martins, L. D., Sobrinho, O. M., Ribeiro, A. K. C., Hoinaski, L., Hoelzemann, J. J., Ynoue, R. Y., Pedruzzi, R., Andrade, M. F., & Albuquerque, T. T. A. (2026). Nationwide portrait of air pollution in Brazil: Closing monitoring gaps with a hybrid PM2.5 product. [Journal name], [DOI].