MERRA2 CNN HAQAST: Bias-Corrected Global Hourly PM2.5 Concentrations, 2000-2024

2000 to 2024 hourly global surface PM2.5 mass concentrations, bias-corrected using a convolutional neural network (CNN) method. The dataset is derived from MERRA-2 data and developed for the NASA Health and Air Quality Applied Sciences Team (HAQAST). It includes two parameters: PM2.5 concentration in µg/m³ and a quality flag for each grid point.

Use Cases

• Analyzing global PM2.5 pollution trends based on hourly concentration data from 2000 to 2024
• Assessing air quality for public health studies based on bias-corrected satellite-derived PM2.5 concentrations
• Validating and improving atmospheric models based on machine learning-corrected MERRA-2 data
• Filtering data for quantitative analysis based on the QFLAG quality indicator mentioned in the description

Strengths

• Hourly temporal resolution covering a 24-year period from 2000 to 2024
• Includes a quality flag (QFLAG) parameter with values from 1 to 4 to guide quantitative analysis
• Bias correction performed using a convolutional neural network (CNN) machine learning approach

Limitations

• Column-level documentation is absent; field semantics must be inferred after download
• Row count is unknown, which may limit suitability assessment

Provenance

Source

National Aeronautics and Space Administration

Collection Method

Derived from MERRA-2 data using a convolutional neural network (CNN) machine learning approach for bias correction.

Time Range

2000 to 2024

Freshness

Last updated 2026-03-13 00:27:47.988642; freshness should be verified

Geography

Global