Satellite-Based Long-Term Spatiotemporal Patterns of Surface Ozone Concentrations in China: 2005-2019

Although short-term ozone ($$) exposure has been associated with a series of adverse health outcomes, research on the health effects of chronic $$ exposure is still limited, especially in developing countries because of the lack of long-term exposure estimates.

The present study aimed to estimate the spatiotemporal distribution of monthly mean daily maximum 8-h average $$ concentrations in China from 2005 to 2019 at a 0.05° spatial resolution.

We developed a machine learning model with a satellite-derived boundary-layer $$ column, $$ precursors, meteorological conditions, land-use information, and proxies of anthropogenic emissions as predictors.

The random, spatial, and temporal cross-validation $$ of our model were 0.87, 0.86, and 0.76, respectively. Model-predicted spatial distribution of ground-level $$ concentrations showed significant differences across seasons. The highest summer peak of $$ occurred in the North China Plain, whereas southern regions were the most polluted in winter. Most large urban centers showed elevated $$ levels, but their surrounding suburban areas may have even higher $$ concentrations owing to nitrogen oxides titration. The annual trend of $$ concentrations fluctuated over 2005-2013, but a significant nationwide increase was observed afterward.

The present model had enhanced performance in predicting ground-level $$ concentrations in China. This national data set of $$ concentrations would facilitate epidemiological studies to investigate the long-term health effect of $$ in China. Our results also highlight the importance of controlling $$ in China's next round of the Air Pollution Prevention and Control Action Plan. https://doi.org/10.1289/EHP9406.