Daily and Annual PM2.5 Concentrations for the Contiguous U.S., 1-km Grids

1-km resolution predictions of PM2.5 concentrations across the contiguous United States from 2000 to 2016. An ensemble model combining neural network, random forest, and gradient boosting methods achieved cross-validated R-squared values of 0.86 for daily and 0.89 for annual predictions. The data integrates satellite observations, meteorological variables, land-use data, and chemical transport model outputs.

Use Cases

• Estimating long-term population exposure to fine particulate matter based on annual concentration averages.
• Analyzing short-term pollution events and their drivers using daily predicted concentrations.
• Validating regional air quality models with high-resolution, machine-learning-derived estimates.
• Studying correlations between PM2.5 levels and geographic predictors like land use and elevation.

Strengths

• High predictive performance with a 10-fold cross-validated R-squared of 0.89 for annual predictions.
• Spatially detailed coverage at a 1-km grid resolution for the entire contiguous United States.
• Long temporal coverage spanning 17 years from 2000 to 2016.
• Model ensemble incorporates multiple data sources including satellite, meteorological, and land-use variables.

Limitations

• Metadata completeness is low; specific column names, row counts, file size, and license information are not provided by any source.
• Conflicting 'last updated' dates exist between platforms (2026-07-21 on Dataverse vs. 2016-12-31 on NASA Earthdata).
• Data consists of model predictions, not direct ground measurements, which may introduce bias.

Provenance

Source

Di, Q., Y. Wei, A. Shtein, C. Hultquist, X. Xing, H. Amini, L. Shi, I. Kloog, R. Silvern, J. Kelly, M. B. Sabath, C. Choirat, P. Koutrakis, A. Lyapustin, Y. Wang, L. J. Mickley, and J. Schwartz

Collection Method

Predictions from a generalized additive model ensembling three machine learning algorithms (neural network, random forest, gradient boosting).

Time Range

2000 - 2016

Freshness

2026-07-21

Geography

Contiguous United States