Machine learning and causal inference for disentangling air pollution reduction during the Asian Games in megacity Hangzhou.
Journal:
Environmental pollution (Barking, Essex : 1987)
Published Date:
Jul 4, 2025
Abstract
Controlling air pollution in megacities remains challenging due to complex pollutant sources, atmospheric chemistry, and weather conditions. During the 19th Asian Games (AG) in Hangzhou, the government enforced strict traffic restrictions in 10 % of the urban area while maintaining industrial operations. The AG coincided with the China's National Day Holiday (CNDH) from September 29th to October 6th, providing a distinctive opportunity to investigate the impact of social events on urban air pollution. Using Machine-Learning-based weather normalization, we separated the effects of weather and emissions on pollutant concentrations. The Synthetic Control Method quantified pollutant reductions attributable to the AG by comparing observed data to a counterfactual scenario without the AG but with the CNDH. Results showed that NO, PM, and O concentrations changed by -4.6, -0.5, and -0.5 μg/m during the AG compared to the prior 15 days. Weather variations contributed -3.1, 0.4, and -3.1 μg/m to these changes, respectively. The AG led to a statistically significant NO reduction of -3.9 μg/m, but PM and O showed no significant changes (0.1 and 1.3 μg/m). The CNDH alone reduced NO, PM, and O by -2.1, -2.2, and -0.4 μg/m, respectively. Local traffic control during the AG accounted for only 43 % of the NO reduction, with no impact on PM or O. Overall, pollutant reductions were largely driven by the CNDH and weather variations. These findings suggest that local traffic restrictions during social events have limited effectiveness on megacity air quality, highlighting the dominance of broader factors.
Authors
Keywords
No keywords available for this article.
