First retrieval of 24-hourly 1-km-resolution gapless surface ozone (O3) from space in China using artificial intelligence: Diurnal variations and implications for air quality and phytotoxicity

Abstract

Surface ozone (O 3 ) is a critical ambient pollutant that poses significant risks to both human health and ecosystems. However, there is a scarcity of high-spatial-resolution hourly surface O 3 data, which is crucial for understanding its diurnal variations. In this study, we employed a best-performing spatiotemporal artificial intelligence (AI) model to estimate 24-hourly 1-km-resolution surface O 3 concentrations across China, incorporating key photochemical processes responsible for O 3 formation. Our model effectively captured diurnal O 3 patterns, achieving average sample-based cross-validated coefficients of determination (root-mean-square errors) of 0.89 (16.35 μg/m 3 ) for the full day (00:00–23:00 LT), 0.92 (15.72 μg/m 3 ) during daytime (08:00–20:00 LT), and 0.82 (16.97 μg/m 3 ) at nighttime (20:00–08:00 LT). Typically, surface O 3 levels increase after sunrise, peak around 15:00 LT, and decrease overnight, with a diurnal variation magnitude of 62 % relative to the mean level. During the daytime, we found that solar radiation (in the ultraviolet and shortwave spectra) and surface temperature explained over 42 % of the diurnal variation, while nighttime O 3 levels were mainly influenced by tropospheric nitrogen dioxide (16 %), temperature (13 %), and relative humidity (12 %). In 2019, approximately 61 %, 98 %, and 100 % of populated areas in China experienced O 3 exposure risks for at least one day, with maximum daily 8-h average (MDA8) O 3 levels exceeding 160, 120, and 100 μg/m 3 , respectively. Additionally, around 70 %, 82 %, and 100 % of vegetated areas exceeded the three minimum critical thresholds for cumulative hourly O₃ exposure, as indicated by the SUM06, W126, and AOT40 indices, respectively. Notably, gross primary productivity (GPP) was the most sensitive indicator of O 3 pollution across various vegetation types, showing a strong negative correlation with AOT0 ( R = −0.43 to −0.59, p < 0.001). • A 24-hourly 1-km gapless surface O 3 dataset is generated using AI in China. • Strong diurnal variation is observed at 62 % of its mean level. • All populated areas were exposure to unhealthy air for at least one day in 2019. • GPP responds the most to surface O 3 pollution across various vegetated types.

Topics & Concepts