Scholarlink

<p>COVID-19 lockdowns unprecedentedly alterted anthorpogenic activities, proving an opportunity to explore atmospheric pollution in economically significant Yangtz River Delta (YRD), yet lacking comprehensive assessment before and after COVID-19. This study presents a comparative analysis of atmospheric pollution in YRD during pre-COVID-19 period (2015–2019, Phase 1), COVID-19 period (2020, Phase 2), and post-lockdown recovery period (2021–2025, Phase 3) by using ground based monitirng data on criteria pollutants from China National Environmental Monitoring Center (CNEMC), and meteorological data from MERRA-2 (Modern-Era Retrospective Analysis for Research and Applications, Version 2). Air Quality Index (AQI) improved by 12.5% and 0.43% during Phase 3 relative to Phase 1 and Phase 2, respectively, with strongest seasonal improvement in winter (22%), followed by autumn (14.2%), spring (6.1%), and summer (4.5%). In Phase 3, SO<sub>2</sub>, PM<sub>2.5</sub>, NO<sub>2</sub>, PM<sub>10</sub>, and CO experienced substantial reductions relative to both Phase 1 and Phase 2, while O<sub>3</sub> experienced slght increases.All pollutants and AQI exhibited same temporal variability, with the lowest pollution levels in summer and the highest in winter, except for O<sub>3</sub>. PM<sub>2.5</sub>, PM<sub>10</sub>, and NO<sub>2</sub> experienced reduction in total number of days as primary pollutant over time, while O<sub>3</sub> experienced an increase. Similarly, percentage of non-attainment days decreased during Phase 2 and then slightly rebound in Phase 3. All machine learning models e.g., Random Forest (RF), Gradient Boosting Decision Tree (GBDT), and Extreme Gradient Boosting (XGBoost) showed high predictive accuracy (R² 0.980–0.999), with tuned GBDT outperformed other XGBoost and RF (R² &gt;0.996). Hyperparameter tunning enhanced GBDT and XGBoost performance, but reduced RF accuracy. SHapley Additive exPlanations (SHAP) feature importance of NO<sub>2</sub> decreased while O<sub>3</sub> emerged as influential pollutant. The above-mentioned results represents significant influence of COVID-19 lockdowns on air quality, and underscore the need for targeted emission reduction strategies in YRD.</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

Decadal air quality dynamics in the Yangtze River Delta: Evidence from machine learning–driven analysis

  • 期刊论文

  • 发表日期：

    2026年3月13日

查看全文
  • Shah Zaib,
  • Muhammad Zeeshaan Shahid,
  • Imran Shahid

摘要

COVID-19 lockdowns unprecedentedly alterted anthorpogenic activities, proving an opportunity to explore atmospheric pollution in economically significant Yangtz River Delta (YRD), yet lacking comprehensive assessment before and after COVID-19. This study presents a comparative analysis of atmospheric pollution in YRD during pre-COVID-19 period (2015–2019, Phase 1), COVID-19 period (2020, Phase 2), and post-lockdown recovery period (2021–2025, Phase 3) by using ground based monitirng data on criteria pollutants from China National Environmental Monitoring Center (CNEMC), and meteorological data from MERRA-2 (Modern-Era Retrospective Analysis for Research and Applications, Version 2). Air Quality Index (AQI) improved by 12.5% and 0.43% during Phase 3 relative to Phase 1 and Phase 2, respectively, with strongest seasonal improvement in winter (22%), followed by autumn (14.2%), spring (6.1%), and summer (4.5%). In Phase 3, SO2, PM2.5, NO2, PM10, and CO experienced substantial reductions relative to both Phase 1 and Phase 2, while O3 experienced slght increases.All pollutants and AQI exhibited same temporal variability, with the lowest pollution levels in summer and the highest in winter, except for O3. PM2.5, PM10, and NO2 experienced reduction in total number of days as primary pollutant over time, while O3 experienced an increase. Similarly, percentage of non-attainment days decreased during Phase 2 and then slightly rebound in Phase 3. All machine learning models e.g., Random Forest (RF), Gradient Boosting Decision Tree (GBDT), and Extreme Gradient Boosting (XGBoost) showed high predictive accuracy (R² 0.980–0.999), with tuned GBDT outperformed other XGBoost and RF (R² >0.996). Hyperparameter tunning enhanced GBDT and XGBoost performance, but reduced RF accuracy. SHapley Additive exPlanations (SHAP) feature importance of NO2 decreased while O3 emerged as influential pollutant. The above-mentioned results represents significant influence of COVID-19 lockdowns on air quality, and underscore the need for targeted emission reduction strategies in YRD.