Articles | Volume 25, issue 20
https://doi.org/10.5194/acp-25-13585-2025
https://doi.org/10.5194/acp-25-13585-2025
Research article
 | 
23 Oct 2025
Research article |  | 23 Oct 2025

Rethinking machine learning weather normalisation: a refined strategy for short-term air pollution policies

Yuqing Dai, Bowen Liu, Chengxu Tong, David C. Carslaw, A. Robert MacKenzie, and Zongbo Shi

Related authors

Enhancing Accuracy of Indoor Air Quality Sensors via Automated Machine Learning Calibration
Juncheng Qian, Thomas Wynn, Bowen Liu, Yuli Shan, Suzanne E. Bartington, Francis D. Pope, Yuqing Dai, and Zongbo Shi
EGUsphere, https://doi.org/10.5194/egusphere-2025-3839,https://doi.org/10.5194/egusphere-2025-3839, 2025
This preprint is open for discussion and under review for Atmospheric Measurement Techniques (AMT).
Short summary

Cited articles

Abdi, H. and Williams, L. J.: Principal component analysis, Wiley Interdiscip. Rev. Comput. Stat., 2, 433–459, https://doi.org/10.1002/wics.101, 2010. 
Acosta-Ramírez, C. and Higham, J. E.: Effects of meteorology and human-mobility on UK's air quality during COVID-19, Meteorological Applications, 29, 2061, https://doi.org/10.1002/met.2061, 2022. 
Andrews, S. Q.: Inconsistencies in air quality metrics: `Blue Sky' days and PM10 concentrations in Beijing, Environ. Res. Lett., 3, 034009, https://doi.org/10.1088/1748-9326/3/3/034009, 2008. 
Bannister, E. J., Cai, X., Zhong, J., and MacKenzie, A. R.: Neighbourhood-scale flow regimes and pollution transport in cities, Bound.-Lay. Meteorol., 179, 259–289, https://doi.org/10.1007/s10546-020-00593-y, 2021. 
Ben-Michael, E., Feller, A., and Rothstein, J.: The augmented synthetic control method, J. Am. Stat. Assoc., 116, 1789–1803, https://doi.org/10.1080/01621459.2021.1929245, 2021. 
Download
Short summary
Air pollution causes millions of deaths annually, driving policies to improve air quality. However, assessing these policies is challenging because weather changes can hide their true impact. We created a logical evaluation framework and found that a widely applied machine learning approach that adjusts for weather effects could underestimate the effectiveness of short-term policies, like emergency traffic controls. We proposed a refined approach that could largely reduce such underestimation.
Share
Altmetrics
Final-revised paper
Preprint