Global Hourly Surface PM2.5 Concentrations with Quality Flags

NASA GES DISC provides global hourly surface total PM2.5 mass concentration data from 2000 to 2024. This dataset is generated by applying a convolutional neural network to bias-correct MERRA-2 reanalysis data for the Health and Air Quality Applied Sciences Team. It contains two primary parameters: PM2.5 concentration and a corresponding quality flag.

Use Cases

• Analyze temporal trends in MERRA2_CNN_Surface_PM25 concentrations to study long-term global air pollution changes from 2000 to 2024.
• Filter data using the QFLAG parameter to select only high-quality grid points (values 3 and 4) for quantitative regional air quality assessment.
• Model the relationship between surface PM2.5 levels and health outcomes using the spatially and temporally resolved concentration data.
• Validate other air quality models or satellite retrievals against this machine learning-corrected global PM2.5 dataset.
• Investigate diurnal and seasonal patterns of particulate matter by analyzing the hourly resolution of the PM2.5 concentration variable.

Strengths

• 25-year temporal coverage from 2000 to 2024.
• Global spatial coverage at the horizontal resolution of the MERRA-2 reanalysis.
• Includes a dedicated quality flag (QFLAG) to indicate data reliability at each grid point.
• Data is bias-corrected using a convolutional neural network method.

Limitations

• Specific sample size (row count) and file size are not provided.
• The underlying MERRA-2 reanalysis may have inherent biases that the CNN correction cannot fully address.
• Data quality is variable, with QFLAG values of 1 and 2 indicating lower reliability.

Provenance

Source

NASA Goddard Earth Sciences Data and Information Services Center (GES DISC).

Collection Method

Derived by applying a convolutional neural network to bias-correct the MERRA-2 atmospheric reanalysis.

Time Range

2000 to 2024.

Geography

Global.