Deep learning-based surface O3 responses to anthropogenic and meteorological changes

The applications of deep learning (DL) technique in atmospheric environment research are expanding rapidly. Here we developed a DL framework to quantify the responses of surface ozone (O₃) to anthropogenic and meteorological changes in China. The DL-based analysis suggests volatile organic compound (VOC)-limited regimes in urban areas over northern inland China, and thus, reductions of nitrogen oxide (NO_x) emissions have resulted in increases in surface O₃ concentrations. In contrast, changes in meteorological conditions led to a dramatic decrease in surface O₃ concentrations in 2019-2021, particularly, in the North China Plain, whereas the decline in surface O₃ concentrations driven by beneficial meteorological conditions in 2019-2021 has been completely reversed due to the occurrence of long-lasting heatwave in 2022, particularly in central China. The DL framework, developed in this work, provides a novel data-driven pathway to assess the causes of surface O₃ changes, and is helpful for a comprehensive understanding of the driving factors of surface O₃ evolution in China.