Integrating traffic pollution dispersion into spatiotemporal NO₂ prediction

Area of Science:

• Environmental Science
• Public Health
• Atmospheric Chemistry

Background:

• Traffic-related air pollution, particularly nitrogen dioxide (NO2), is a significant public health concern in urban environments.
• Accurate prediction of ambient NO2 concentrations is crucial for assessing and mitigating exposure risks.

Purpose of the Study:

• To develop and evaluate a novel fine-scale spatiotemporal model for predicting ambient NO2 concentrations.
• To incorporate traffic-related data using the scalable dispersion model, Research LINE source dispersion model (RLINE), into NO2 prediction.

Main Methods:

• Utilized national traffic estimate datasets and meteorological data within the RLINE model.
• Employed national-level spatial regression models with nearest-neighbor Gaussian processes (spNNGP) to predict road-type-specific annual average daily traffic (AADT).
• Integrated RLINE estimates as space-only and spatiotemporal covariates into a validated spatiotemporal NO2 modeling approach using data from the Multi-Ethnic Study of Atherosclerosis and Air Pollution.

Main Results:

• Integrating RLINE estimates as a space-only covariate improved overall cross-validation R² from 0.83 to 0.84 and reduced RMSE from 3.58 to 3.48 ppb.
• Model performance showed more significant improvement at roadside monitors near highways, with R² increasing from 0.56 to 0.66 and RMSE decreasing from 3.52 to 3.11 ppb.
• The enhanced model demonstrated improved predictive capabilities for roadside NO2 concentration gradients.

Conclusions:

• The novel approach effectively enhances the prediction of ambient NO2 concentrations by integrating traffic data.
• The RLINE-enhanced spatiotemporal model offers improved accuracy, particularly in areas with high traffic density like highways.
• This generalized modeling framework can advance high-resolution NO2 exposure prediction across the U.S.