Assessing NO₂ Concentration and Model Uncertainty with High Spatiotemporal Resolution across the Contiguous United States Using Ensemble Model Averaging

Abstract

NO 2 is a combustion byproduct that has been associated with multiple adverse health outcomes. To assess NO 2 levels with high accuracy, we propose the use of an ensemble model to integrate multiple machine learning algorithms, including neural network, random forest, and gradient boosting, with a variety of predictor variables, including chemical transport models. This NO 2 model covers the entire contiguous U.S. with daily predictions on 1km-level grid cells from 2000 to 2016. The ensemble produced a cross-validated R 2 of 0.788 overall, a spatial R 2 of 0.844, and a temporal R 2 of 0.729. The relationship between daily monitored and predicted NO 2 is almost linear. We also estimated the associated monthly uncertainty level for the predictions and address-specific NO 2 levels. This NO 2 estimation has a very high spatiotemporal resolution and allows the examination of the health effects of NO 2 in unmonitored areas. We found the highest NO 2 levels along highways and in cities. We also observed that nationwide NO 2 levels declined in early years and stagnated after 2007, in contrast to the trend at monitoring sites in urban areas, where the decline continued. Our research indicates that the integration of different predictor variables and fitting algorithms can achieve an improved air pollution modeling framework.

Long-term exposure to air-pollution and COVID-19 mortality in England: a hierarchical spatial analysis

“…For the exposure, they calculated the mean of daily concentrations during 2010–2016 as modelled by a previously described ensemble machine learning model (R 2 =0·79). 25 They reported a 7·1% (1·2%–13·4%) increase in mortality per 4·5ppb (1ppb=1·25μg/m 3 ) increase in NO 2 after adjusting for confounders and spatial autocorrelation 9 (that is approximately 1·3% increase per 1 μg/m 3 ). The study in England, with partly overlapping data as in our analysis, also reported a significant association between NO 2 and COVID-19 mortality (p<0·05).…”

Long-term exposure to air-pollution and COVID-19 mortality in England: A hierarchical spatial analysis

“…The study in the US focused on deaths reported by April 29, 2020, using 3122 counties. For the exposure, they calculated the mean of daily concentrations during 2010–2016 as modelled by a previously described ensemble machine learning model (R 2 = 0.79) ( Di et al, 2019a ). They reported a 7.1% (95% Confidence Interval: 1.2%, 13.4%) increase in mortality per 4.5 ppb (1 ppb = 1.25 μg/m 3 ) increase in NO 2 after adjusting for confounders and spatial autocorrelation( Liang et al 2020 )(that is approximately 1.3% increase per 1 μg/m 3 ).…”

Assessing the Distribution of Air Pollution Health Risks within Cities: A Neighborhood-Scale Analysis Leveraging High-Resolution Data Sets in the Bay Area, California

“…Future work can make use of more spatially refined estimates of PM 2:5 (Di et al 2019). In addition, NO 2 satellite-derived models have now been developed (Di et al 2020) that can be compared with results from LUR models.…”