Daily and Annual NO2 Concentrations for the Contiguous U.S.

An ensemble of three machine learning models and a generalized additive model predicts daily Nitrogen Dioxide levels at a 1-km resolution from 2000 to 2016. The modeling framework incorporates satellite column concentrations, land-use data, meteorological variables, and outputs from chemical transport models GEOS-Chem and CMAQ. This high-resolution dataset supports research into the short- and long-term health effects of air pollution across the contiguous United States.

Use Cases

• Estimating short-term NO2 exposure for epidemiological studies based on daily concentration predictions.
• Analyzing long-term trends in air pollution for environmental justice research using annual averages.
• Validating regional air quality models with high-resolution, cross-validated ground-level estimates.
• Assessing correlations between land-use variables and NO2 concentrations for urban planning.

Strengths

• High predictive accuracy with a cross-validated R-squared of 0.79 overall, and a spatial R-squared of 0.84.
• Uses a multi-source ensemble approach combining satellite data, machine learning, and chemical transport models.
• Provides both daily and annual data at a fine 1-km spatial resolution for detailed analysis.

Limitations

• Metadata completeness is low; specific column names, row counts, file size, and license information are unavailable from all sources.
• Temporal coverage ends in 2016, which may limit analysis of recent trends.
• Conflicting 'last updated' dates (2026-05-27 vs. 2016-12-31) create uncertainty about dataset maintenance.

Provenance

Source

Socioeconomic Data and Applications Center (SEDAC)

Collection Method

Predictions generated via an ensemble modeling framework combining machine learning estimates with a generalized additive model.

Time Range

2000 - 2016

Freshness

2026-05-27 17:58:13

Geography

Contiguous United States