Assessing anthropogenic contributions and uncovering inter-regional periodic patterns of ground ozone with high-resolution predictions in 2015–2019 across China

Abstract

Owing to the strong spatiotemporal variability of ozone and the complexity of its photochemical reactions, it is urgent but difficult to accurately predict the high-resolution distribution of ozone and quantify the effects of anthropogenic drivers. In this study, we employed a random forest model to predict maximum daily 8-hour average ozone concentrations (MDA8 O₃) at a high resolution of 1 km × 1 km across China from 2015 to 2019. The model's performance was validated using three approaches: sample-based, site-based, and year-based, yielding R-squared values of 0.87, 0.85, and 0.81, respectively, and demonstrating superior accuracy compared to previous studies. Our predictions revealed that Central China experienced the most rapid increase in ozone, with some areas exceeding 6 μg/m³/year, surpassing even the economically developed regions of Eastern China, as identified by Sen's slope and the seasonal Mann-Kendall test. Through high-resolution predictions, we uncovered stable inter-regional periodic patterns of high ozone concentrations across four seasons. By controlling for meteorological variables, we also quantified anthropogenic contributions to the changes in ground-level ozone in 2015–2019, which ranged from −12.18 to 43.71 μg/m³ annually, thereby driving the rapid increase in ozone concentrations over Central China. The high-resolution ozone datasets and the identification of inter-regional periodic patterns offer valuable insights for large-scale ozone studies and provide cost-effective strategies for ozone monitoring and control.