the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Tracking surface ozone responses to clean air interventions under a warming climate in China
Abstract. Surface ozone, a major air pollutant with profound implications for human health, ecosystems, and climate, shows long-term trends shaped by both anthropogenic and climatic drivers. Here, we develop a machine learning-based approach – the Fixed Emission Approximation (FEA) – to disentangle the effects of meteorological variability and anthropogenic emissions on summertime ozone trends in China. We identify three distinct phases of ozone trends corresponding to clean air actions. Anthropogenic emissions drove a +23.2 ± 1.1 μg m⁻3 increase in summer maximum daily 8-hour average ozone during 2013–2017, followed by a −4.6 ± 1.5 μg m⁻3 decrease during 2018–2020. However, during 2021–2023, extreme meteorological anomalies – including heatwaves and extended monsoon rainfall – emerged as key drivers of ozone variability. Satellite-derived formaldehyde-to-nitrogen dioxide ratios reveal widespread urban volatile organic compounds-limited regimes, with a shift toward nitrogen oxides-limited sensitivity under influence of heatwaves. Finally, we assess ozone trends under sustained climate warming from 1970 to 2023 based on the FEA framework. The results indicate a significant climate-driven increase in ozone levels across China's urban agglomerations, underscoring the amplifying role of climate change in ozone pollution. Together, these findings highlight the dual influence of anthropogenic and climatic factors on ozone pollution and emphasize the need for integrated strategies that couple emission mitigation with climate adaptation to effectively manage ozone risks in a warming world.
- Preprint
(2452 KB) - Metadata XML
-
Supplement
(9777 KB) - BibTeX
- EndNote
Status: open (until 23 Oct 2025)
- RC1: 'Comment on egusphere-2025-4014', Anonymous Referee #2, 30 Sep 2025 reply
Viewed
HTML | XML | Total | Supplement | BibTeX | EndNote | |
---|---|---|---|---|---|---|
101 | 12 | 3 | 116 | 15 | 3 | 3 |
- HTML: 101
- PDF: 12
- XML: 3
- Total: 116
- Supplement: 15
- BibTeX: 3
- EndNote: 3
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1
Tropospheric ozone is a globally important air pollutant and a short-lived climate forcer, with substantial impacts on human health, climate change, and terrestrial ecosystems. Understanding the relationship between ozone concentration changes and their driving factors is essential for developing effective control strategies. This study utilizes ground-based observational data from the Chinese monitoring network during 2013-2023 and develops a machine-learning-based method to quantitatively disentangle the contributions of meteorological conditions and anthropogenic emissions. The analysis is further extended to evaluate the sensitivity of ozone to climate change. In addition, the authors employ satellite retrievals to explore the changes in precursor ratios and to diagnose the shifts in chemical regimes. The proposed analytical framework provides valuable insights and will be highly informative for future studies. Overall, the manuscript is clearly structured, well-designed, and well-written, and it fits well within the scope of ACP. I would recommend publication after the following issues are addressed:
Title suggestion: Consider revising the title to “Tracking surface ozone responses to clean air actions under a warming climate in China” for clarity and stronger alignment with the scope.
Lines 52-54: This sentence requires additional references. In particular, the 2021 IPCC report should be cited and carefully verified.
Line 81: Provide the full name of XGBoost when first introduced.
Line 89: Please remove the word “monsoon.”
Line 109: The study develops an innovative machine-learning framework for attribution analysis, including an extension to climate change. This is a key contribution, but I suggest adding more technical details, such as a conceptual diagram of the methodology, to improve clarity and accessibility for readers.
Line 114: The role of time variables requires clarification. Were the diurnal and seasonal/monthly variables included to remove short-term and seasonal variability, leaving the long-term trend for quantitative attribution? Please explain explicitly.
Line 125: Why was the modeling performed separately for each city, rather than by grouping cities into regions? Please explain the rationale.
Lines 152-161: The uncertainty analysis, particularly for the Fixed Emission Approximation (FEA) method, is highly valuable. I strongly recommend moving these results (currently Figure S3) into the main text.
Lines 202-205: The manuscript highlights several regions in China. Please explain why these regions were emphasized and include a map showing their geographic distribution for better context.
Line 227: The references here primarily address ecological impacts, yet the text mentions “both human and ecological health.” Please provide more references specific to human health. Also, revise “ecological health” to “ecosystem health.”
Line 229: The phrase “reflecting initial policy effectiveness” is unclear. Please rephrase for precision.
Lines 230-232: The conclusion drawn here seems overstated, as the evidence provided is insufficient. This section mainly discusses temporal and spatial ozone concentration trends. A more cautious interpretation is recommended: instead of attributing trends directly to policy effectiveness, the authors could note that observed trends occurred under varying emission control backgrounds, while meteorology also played an important role.
Line 243: Please define the parameters τ and p shown in Figure 1.
Line 276: Correct “Emission-driven” to “emission-driven.”
Lines 279-281: The logic here is confusing. The discussion first emphasizes the role of anthropogenic emissions, but then suggests that changes in emissions highlight the role of meteorology. Please clarify or restructure this argument.
Lines 299-300: Please rephrase the sentence. The term “near-baseline” is ambiguous and requires clarification.
Line 342: There is an editorial error that needs correction.
Line 390: I recommend revising the y-axis labels for greater accuracy. For instance, in panel (a), the label currently suggests “extreme weather,” but it actually represents only “extreme heatwave”. In contrast, panel (b) provides a more specific description. The labeling should be made consistent and precise to avoid potential misinterpretation.
Line 397: The title of this section should be revised, since the authors are not reconstructing the ozone trend per se. A more accurate option could be “Reshaping distributions of ozone controlled by a warming climate.” This section is indeed interesting and methodologically innovative. However, the manuscript should elaborate more clearly on which specific factors are included in the climate-change-driven trend, especially considering the constraints posed by the limited length and coverage of historical observational records.
Line 331 (Figure 3): Ensure the map format is consistent with that in the Supplementary Figures.
Lines 374-376: Since the SHAP interpreter is a key tool used to analyze predictor contributions, it should be briefly described in the Methods section.
Line 437: In Figure 5, the authors present both climate-change-driven and emission-driven trends. I am curious about how the results from the proposed FEA method compare with those from other widely used machine-learning approaches for trend analysis, such as de-weather. A comparison between different methods would not only be interesting but also serve as a useful validation of the robustness of the proposed framework.
Line 445 (Conclusion): The conclusion is overly lengthy. Please condense and refine this section for clarity and impact.