Global and regional PM2.5 concentrations are estimated using information from satellite-, simulation- and monitor-based sources. Aerosol optical depth from multiple satellites (MODIS, VIIRS, MISR, SeaWiFS, and VIIRS) and their respective retrievals (Dark Target, Deep Blue, MAIAC) is combined with simulation (GEOS-Chem) based upon their relative uncertainties as determined using ground-based sun photometer (AERONET) observations to produce geophysical estimates that explain most of the variance in ground-based PM2.5 measurements. A subsequent statistical fusion incorporates additional information from PM2.5 measurements.

Contents:

V6.GL.02.02 is recommended for all regions and is available for 1998-2022
V5.GL.04 is recommended for all regions and is available for 1998-2022
V5.NA.04.02 is available for compositional use over North America and is available for 2000-2022

Additions tools developed by users for accessing these datasets are available here

Previous versions are available from the Satellite-derived PM2.5 Archive


———————————————————————————

Global Estimates (V6.GL.02.02):

We estimate annual and monthly ground-level fine particulate matter (PM2.5) for 2000-2019 by combining Aerosol Optical Depth (AOD) retrievals from the NASA MODIS, MISR, SeaWIFS, and VIIRS with the GEOS-Chem chemical transport model, and subsequently calibrating to global ground-based observations using a residual Convolutional Neural Network (CNN), as detailed in the below reference for V6.GL.01. V6.GL.02.02 follows the methodology of V6.GL.01 but updates the ground-based observations used to calibrate the geophysical PM2.5 estimates for the entire time series, extends temporal coverage through 1998 – 2022, and includes retrievals from the SNPP VIIRS instrument.

Reference:
Shen, S. Li, C. van Donkelaar, A. Jacobs, N. Wang, C. Martin, R. V.: Enhancing Global Estimation of Fine Particulate Matter Concentrations by Including Geophysical a Priori Information in Deep Learning. (2024) ACS ES&T Air. DOI: 10.1021/acsestair.3c00054 [Link]

Annual and monthly datasets are provided in NetCDF [.nc] format, with naming convention V6GL02.02.CNNPM25.REGION.YYYYMM_START-YYYYMM_END.nc. REGION refers to the file region (e.g. ‘Global’). YYYYMM_START and YYYYMM_END refer to the numeric start and end date of the file (e.g. for annual mean PM2.5 for 2015, YYYYMM_START is 201501 and YYYYMM_END is 201512). Gridded files use the WGS84 projection.

Variable names within these files include:

‘lat’: Latitude coordinate centers of the PM2.5 grid
‘lon’: Longitude coordinates centers of the PM2.5 grid
‘PM25’: Gridded mean PM2.5 concentrations

Please contact Siyuan Shen (s.siyuan@wustl.edu) for further information.

Note that these estimates are primarily intended to aid in large-scale studies. Annual and coarse-resolution averages correspond to a simple mean of within-grid values. Gridded datasets are provided to allow users to agglomerate data as best meets their particular needs. High-resolution (0.01° × 0.01°) datasets are gridded at the finest resolution of the information sources that were incorporated but are unlikely to fully resolve PM2.5 gradients at the gridded resolution due to influence by information sources at coarser resolution.

Lijian Han (ljhan@rcees.ac.cn) has prepared an example R-based code that converts the below netcdf files to raster for import into GIS-based programs that lack netcdf support. It can be accessed via this link: [NC to Geotif.txt]. Please contact Lijian if you have further questions.

A tutorial that creates a time series of annual mean PM2.5 at selected locations around the world by importing this dataset into Matlab is available here.

Annual and monthly mean PM2.5 [ug/m3] at 0.01° × 0.01°:
[https://wustl.box.com/v/ACAG-V6GL0202-CNNPM25]
Annual and monthly mean PM2.5 [ug/m3] at 0.1° × 0.1°:
[https://wustl.box.com/v/ACAG-V6GL0202-CNNPM25c0p10]

Processed Datasets:
These summary files are processed from the Scientific Datasets above for ease of accessibility. Population-weighted estimates and total population describe only those people covered by the V6.GL.02.02 dataset and are provided by GPWv4. Country borders are defined following GAD3.6.

Annual Global country-level mean PM2.5.
Annual Canada provincial-level mean PM2.5.
Annual China regional-level mean PM2.5.
Annual India regional-level mean PM2.5.
Annual United States state-level mean PM2.5.

Satellite-derived PM2.5 data V6.GL.02.02 are licensed under CC BY 4.0

———————————————————————————

Global/Regional Estimates (V5.GL.04):

We estimate annual and monthly ground-level fine particulate matter (PM2.5) for 1998-2022 by combining Aerosol Optical Depth (AOD) retrievals from the NASA MODIS, MISR, SeaWIFS, and VIIRS instruments with the GEOS-Chem chemical transport model, and subsequently calibrating to global ground-based observations using a Geographically Weighted Regression (GWR), as detailed in the below reference for V5.GL.01. V5.GL.04 follows the methodology of V5.GL.01, but updates the ground-based observations used to calibrate the geophysical PM2.5 estimates for the entire time series, extends temporal coverage through 2022, and includes retrievals from the SNPP VIIRS instrument.

Reference:
Aaron van Donkelaar, Melanie S. Hammer, Liam Bindle, Michael Brauer, Jeffery R. Brook, Michael J. Garay, N. Christina Hsu, Olga V. Kalashnikova, Ralph A. Kahn, Colin Lee, Robert C. Levy, Alexei Lyapustin, Andrew M. Sayer and Randall V. Martin (2021). Monthly Global Estimates of Fine Particulate Matter and Their Uncertainty Environmental Science & Technology, 2021, doi:10.1021/acs.est.1c05309. [Link]

Scientific Datasets:
Annual and monthly datasets are provided in NetCDF [.nc] format. Gridded files use the WGS84 projection. Please contact Aaron van Donkelaar (aaron.vandonkelaar@wustl.edu) for further information.

Note that these estimates are primarily intended to aid in large-scale studies. Annual and coarse-resolution averages correspond to a simple mean of within-grid values. Gridded datasets are provided to allow users to agglomerate data as best meets their particular needs. High resolution (0.01° × 0.01°) datasets are gridded at the finest resolution of the information sources that were incorporated, but are unlikely to fully resolve PM2.5 gradients at the gridded resolution due to influence by information sources at coarser resolution.

Lijian Han (ljhan@rcees.ac.cn) has prepared an example R-based code that converts the below netcdf files to raster for import into GIS-based programs that lack netcdf support. It can be accessed via this link: [NC to Geotif.txt]. Please contact Lijian if you have further questions.

Annual and monthly mean PM2.5 [ug/m3] at 0.01° × 0.01°:
[https://wustl.box.com/v/ACAG-V5GL04-GWRPM25]
Annual and monthly mean PM2.5 [ug/m3] at 0.1° × 0.1°:
[https://wustl.box.com/v/ACAG-V5GL04-GWRPM25c0p10]

Annual and monthly mean PM2.5 Uncertainty [ug/m3] at 0.01° × 0.01°:
[https://wustl.box.com/v/ACAG-V5GL04-GWRPM25SIGMA]

Processed Datasets:
These summary files are processed from the Scientific Datasets above for ease of accessibility. Population-weighted estimates and total population describe only those people covered by the V5.GL.04 dataset and are provided by GPWv4. Country borders are defined following GAD3.6.

Annual Global country-level mean PM2.5.
Annual Canada provincial-level mean PM2.5.
Annual China regional-level mean PM2.5.
Annual India regional-level mean PM2.5.
Annual United States state-level mean PM2.5.


———————————————————————————

North American Regional Estimates with Composition (V5.NA.04.02):

We estimate ground-level fine particulate matter (PM2.5) total and compositional mass concentrations over North America by combining Aerosol Optical Depth (AOD) retrievals from the NASA MODIS, MISR, SeaWIFS, and VIIRS instruments with the GEOS-Chem chemical transport model, and subsequently calibrated to regional ground-based observations of both total and compositional mass using Geographically Weighted Regression (GWR) for 2000-2022.

Reference:
van Donkelaar, A., R. V. Martin, B. Ford, C. Li, A. J. Pappin, S. Shen, and D. Zhang, North American Fine Particulate Matter Chemical Composition for 2000–2022 from Satellites, Models, and Monitors: The Changing Contribution of Wildfires., ACS ES&T Air, doi: 10.1021/acs.est.0c01764, 2024. [Link]

Scientific Datasets:
Annual and Biweekly datasets are provided in NetCDF [.nc] or a zipped ArcGIS-compatible ASCII [.asc.zip] file. Note that the unzipped ASCII files can be cumbersome. Gridded files use the WGS84 projection. Contact Aaron van Donkelaar (aaron.vandonkelaar@wustl.edu) for further information.

Annual, monthly, and biweekly mean total and component PM2.5 [ug/m3] at 0.01° × 0.01°:
[ASCII]
[NetCDF]

The contribution of biomass burning to annual, monthly, and biweekly mean total and component PM2.5 [ug/m3] at 0.01° × 0.01°, as described in the above publication, is additionally available from:
[ASCII]
[NetCDF]

Note that these estimates are primarily intended to aid in large-scale studies. Gridded datasets are provided to allow users to agglomerate data as best meets their particular needs. Datasets are gridded at the finest resolution of the information sources that were incorporated (0.01° × 0.01°), but do not fully resolve PM2.5 gradients at the gridded resolution due to influence by information sources at coarser resolution.


———————————————————————————
We sincerely appreciate the efforts by our users to develop tools that facilitate access of our PM2.5 estimates. While we don’t directly service these codes, we gratefully make them available for others to use. Please contact the developer indicated if you have any questions.

– Lijian Han (ljhan@rcees.ac.cn) has prepared an example R-based code that converts the netcdf files we provide into raster for import into GIS-based programs that lack netcdf support. It can be accessed via this link: [NC to Geotif.txt].

– MITRE’s Health Equity MIP has developed an open-source R package for particulate matter 2.5 data called [ACAGPM]. This package aligns ACAG’s particulate matter 2.5 data with 2019 census geographies, allowing users to efficiently analyze PM2.5 as it relates to social determinants of health and other variables. For more information, please reach out to the lead contributors indicated via github.