Scholarlink

<p>Over the past decade, ground-level ozone (O<sub>3</sub>) has emerged as a critical air pollutant in China. This study investigates the driving factors behind O<sub>3</sub> pollution in the Pearl River Delta (PRD) urban agglomeration from 2015 to 2024. We analyzed spatiotemporal trends of the daily maximum 8-hour average (O<sub>3</sub>-8&#xa0;h) and applied a Light Gradient Boosting Machine (LightGBM) model (CV-R<sup>2</sup> = 0.85, RMSE = 9.79 ppb) integrated with meteorological normalization to quantify contributions. Results show that O<sub>3</sub>-8&#xa0;h concentrations increased from 39.1 ppb in 2015 to a peak of 50.5 ppb in 2019 (annual growth of ~ 2.89 ppb), sharply dropped by 19% in 2020, and then decreased at a rate of 1.16 ppb per year (R<sup>2</sup> = 0.94) post-2021. Emissions dominated interannual variations (average contribution: 85.7%), while meteorology accounted for 14.3%. Beyond SHAP analysis identifying near-surface temperature (t2m, 20.6%) and dew point temperature (d2m, 20.3%) as key factors, the controlled variable method quantified their isolated impacts: a 10% increase in t2m and d2m respectively increased and decreased O<sub>3</sub>-8&#xa0;h by 16.7 ppb and 14.4 ppb, confirming their strong opposing effects. This quantification validates the non-linear relationships captured by SHAP, demonstrating that high temperature promotes while high humidity strongly suppresses ozone formation. The combined methodology provides robust attribution of meteorological drivers, emphasizing that effective O<sub>3</sub> control requires integrating emission strategies with meteorological insights.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Quantifying contributions of anthropogenic emissions and meteorology to ozone trends in the Pearl River Delta (2015–2024): insights from interpretable machine learning

  • 期刊论文

  • 发表日期：

    2026年6月3日

查看全文
  • Liuming Wang,
  • Guangzong Li,
  • Xiaoyong Liu,
  • Yinhao Luo,
  • Mingjin Zhan

摘要

Over the past decade, ground-level ozone (O3) has emerged as a critical air pollutant in China. This study investigates the driving factors behind O3 pollution in the Pearl River Delta (PRD) urban agglomeration from 2015 to 2024. We analyzed spatiotemporal trends of the daily maximum 8-hour average (O3-8 h) and applied a Light Gradient Boosting Machine (LightGBM) model (CV-R2 = 0.85, RMSE = 9.79 ppb) integrated with meteorological normalization to quantify contributions. Results show that O3-8 h concentrations increased from 39.1 ppb in 2015 to a peak of 50.5 ppb in 2019 (annual growth of ~ 2.89 ppb), sharply dropped by 19% in 2020, and then decreased at a rate of 1.16 ppb per year (R2 = 0.94) post-2021. Emissions dominated interannual variations (average contribution: 85.7%), while meteorology accounted for 14.3%. Beyond SHAP analysis identifying near-surface temperature (t2m, 20.6%) and dew point temperature (d2m, 20.3%) as key factors, the controlled variable method quantified their isolated impacts: a 10% increase in t2m and d2m respectively increased and decreased O3-8 h by 16.7 ppb and 14.4 ppb, confirming their strong opposing effects. This quantification validates the non-linear relationships captured by SHAP, demonstrating that high temperature promotes while high humidity strongly suppresses ozone formation. The combined methodology provides robust attribution of meteorological drivers, emphasizing that effective O3 control requires integrating emission strategies with meteorological insights.