Spring Festival and COVID‐19 Lockdown: Disentangling PM Sources in Major Chinese Cities

Abstract

Wuhan was the first pandemic epicenter and imposed an unprecedented lockdown beginning January 23, 2020. China was subsequently placed into a lockdown. These restrictions interrupted a wide array of economic activities thereby reducing primary air pollutant emissions. The unexpected pollutant emission reductions created a unique opportunity to assess the responses of air quality to rapid temporary reductions in anthropogenic emissions, and inform future air quality abatement strategies.Many early studies showed substantial declines in observed concentrations of nitrogen dioxide (NO 2 ) during lockdown periods globally (

Revealing Drivers of Haze Pollution by Explainable Machine Learning

“…The COVID-19 lockdown in Tianjin started at the beginning of the CSF, leading to significant changes in emissions thereafter (Figure S1). 15,36 Distinctive Drivers of the Five Haze Events. On average, primary emissions far exceeded the recently updated World Health Organization guidance level (i.e., 15 μg m −3 for daily PM 2.5 ), contributing PM 2.5 concentrations of 91 ± 2, 89 ± 3, 86 ± 2, 97 ± 4, and 85 ± 3 μg m −3 during HE1−HE5, respectively.…”

“…In addition to the enhanced formation of secondary aerosol during HE4, 43,45 festival-related emissions also facilitated the formation of hazes. 15,36 It was estimated that festival-related emissions contributed ∼29 and ∼16 μg m −3 PM 2.5 on the first lunar day (January 25) in Beijing and Tianjin, respectively, 15 resulting in the maximum primary emissions being recorded on January 25 (Figure S1). Similar to the cases for HE3 and HE4, PM 2.5 in HE5 was also largely governed by RH (23 ± 15 μg m −3 ).…”

“…14,16,17,39 Time variables are usually considered as proxies for timerelated drivers (e.g., emission rates). Here we fed the RF model with the aforementioned meteorological variables, the cluster and length of air mass trajectories, an additional time variable (lunar date 15 ), and the total gaseous oxidant (i.e., O X = NO 2 + O 3 , a proxy representing atmospheric photooxidation conditions 16,40 ). Details of the RF modeling are provided in Text S1 of the Supporting Information.…”

“…Traditionally, chemical transport models (CTMs) are the most popular tools for performing such analysis via the simulation of scenarios and process analysis, − and the model performance heavily relies on the accuracy and up-to-date availability of an emission inventory. Recently, a machine learning (ML)-based meteorological normalization technique (also called “deweathering”) has been intensively used in atmospheric research. − It is a promising alternative to account for the effects of meteorology on air pollutants. The model performance of ML is generally better than that of traditional statistical analysis (e.g., linear models) and CTMs in predicting, for example, the PM 2.5 level; , however, these results are less robust in physical interpretability due to the “black box” nature of most ML models.…”

“…Unexpected severe haze episodes occurred in northern China during the COVID-19 lockdowns in early 2020, , which attracted a great deal of public attention . Existing studies based on field measurements and conventional CTMs suggested that the pandemic hazes were formed with extremely unfavorable meteorological conditions compared to a year before, , together with enhanced secondary nitrate chemistry , as well as increased festival emissions and residential burning .…”