Air pollution is a major global health concern, contributing significantly to premature deaths. In China, poor air quality, driven by emissions of NOx, CO, and SO2, leading to secondary pollutants like ozone (O3) and particulate matter (PM), is strongly influenced by meteorology. Understanding the relative roles of emission changes, atmospheric chemistry, and meteorology on air quality and health risks requires isolating these confounding factors. The COVID-19 lockdown in China, coinciding with the Chinese New Year (CNY), provided a unique opportunity to study this. CNY typically sees a decrease in NO2 due to reduced economic activity. The COVID-19 lockdown intensified these restrictions, leading to observed decreases in NO2 across China. However, pre-2020 trends in NO2 reduction, attributed to China's Clean Air Plan (CAP), complicate the attribution of changes solely to the lockdown. The CAP, implemented in phases, included measures to reduce coal-fired, industrial, vehicle, and dust emissions. Meteorological conditions, particularly atmospheric transport and Planetary Boundary Layer (PBL) height, also play a crucial role in air quality. Previous studies quantifying the impact of the lockdown on NO2 varied significantly depending on the baseline used for comparison, and few accounted for CNY, CAP, and meteorological effects comprehensively. This study aims to provide a more detailed attribution of NO2 and O3 concentration changes to different drivers at the city level across China during 2020, quantifying the contributions of the CAP, COVID-19 lockdown, and meteorological conditions.

Numerous studies have attempted to quantify the effect of COVID-19 lockdown measures on emission reductions, primarily focusing on NO2. However, these studies varied widely in their findings, primarily due to differences in baseline comparisons (using different reference periods to compare against the lockdown period), lack of consideration for the concurrent effects of CNY and CAP, and insufficient incorporation of meteorological factors. Satellite-based studies showed considerable reductions in NO2, while ground-based measurements revealed significant variability across different regions and cities. Some studies highlighted the role of meteorology, focusing on PM2.5 or specific cities. A key limitation of existing studies was the failure to account for CNY, CAP, and meteorological influences simultaneously, leading to uncertainties in attributing observed changes to the COVID-19 lockdown alone. This lack of comprehensive analysis made it challenging to understand the true magnitude of the lockdown's impact on air quality.

This study employed a combined model-measurement approach to disentangle the effects of various drivers on air quality. Data sources included in-situ measurements of PM2.5, PM10, SO2, NO2, O3, and CO from 367 Chinese cities (excluding Taiwan, Hong Kong, Macau, and Laiwu due to data limitations), provided by the National Environmental Monitoring Center (NEMC). Meteorological data (temperature, humidity, wind speed, direction, precipitation, pressure) from 367 meteorological stations near the air quality monitoring sites were obtained from the National Meteorological Science Data Center. The Copernicus Atmosphere Monitoring Service Reanalysis (CAMSRA) dataset from ECMWF provided NO2 and O3 data, serving as a counterfactual representing a scenario without emission reductions from CNY, CAP, or the COVID-19 lockdown. A Gradient Boosting Machine (GBM) machine learning model was used to predict air pollutant concentrations, separating the influence of meteorology from other factors. The model was trained on data from 2015-2019 and tested on data from the first three months of 2020. The CNY effect was quantified by comparing observed and CAMSRA NO2 and O3 during CNY periods from 2015-2019 and 2020. The CAP effect was determined by comparing observed and CAMSRA data in the period before CNY during 2018-2020. The COVID-19 lockdown effect was isolated by subtracting the CNY and CAP effects from the total anthropogenic influence. The meteorological effect was assessed by comparing GBM predictions using meteorological data from 2020 with predictions using average meteorological conditions from 2015-2019. Health risks were evaluated using the Health-based Air Quality Index (HAQI), considering WHO guidelines and the Chinese Ambient Air Quality Standard grade II (CAAQS-II). Excess risks (ER) were calculated for each pollutant to assess the health impact of air quality changes.

The analysis revealed distinct contributions of CNY, CAP, and COVID-19 lockdown to NO2 and O3 variations: * **CNY (2015-2019):** Reduced NO2 by an average of 26.7% and increased O3 by 23.3%. * **CAP (2018-2020):** Reduced NO2 by 15.7% and increased O3 by 4.9%. * **COVID-19 Lockdown (excluding CNY and CAP):** Further reduced NO2 by 11.6% during the CNY period in 2020 and by 34.7% during the Extended COVID-lockdown period; increased O3 by 21% and 22.7% respectively. * **Meteorology (2020 vs 2015-2019):** Increased NO2 by 7.8% during the Total COVID-lockdown period; O3 was almost unchanged (-0.9%). The combined effects of these drivers led to a significant reduction in multi-pollutant health risks across China. NO2 reduction was the main driver of the health benefit, offsetting the increase in O3. The WHO-based HAQI showed an average reduction of 51.4% across all cities. However, in Northeast China, unfavorable meteorological conditions partially offset the health benefits of emission reductions. The study showed that the average anthropogenic emissions of NO2 across China under the impact of the CNY and the CAP decreased by 26.7% and 15.7%, respectively. The latter highlights the effectiveness of China's new CAP regulations. The COVID-19 lockdown led to additional reductions, with the total reduction during CNY being 54% and 50.4% during the extended period. For O3, increases were noted under all factors, ranging from 21% to 49.3%. The study also analyzed the health risk changes in 31 provincial capital cities. Using WHO guidelines, the average HAQI dropped significantly (61%), while using CAAQS-II standards resulted in a 21% decrease. The study found differences between WHO-HAQI and CAAQS-HAQI calculations because more air pollutants contribute to WHO-HAQI increases, whereas CAAQS-HAQI is only sensitive to PM concentrations. Cities with increased WHO-HAQIs (7.1%) were primarily located in the Yunnan-Guizhou Plateau and Northwest China, areas often affected by transboundary pollution and local factors.

This study's findings provide crucial insights into the complex interplay of various factors influencing air quality during the COVID-19 lockdown in China. The results confirm the significant role of both short-term events (CNY) and long-term policies (CAP) in shaping air pollution levels. The isolated impact of the COVID-19 lockdown on NO2 reduction, while considerable, was smaller than often reported previously. This highlights the importance of considering confounding factors when assessing the impact of such events. The significant decrease in multi-pollutant health risk, largely driven by NO2 reductions, demonstrates the substantial public health benefits associated with even temporary emission reductions. However, the negative effect of meteorological conditions in some regions underscores the need for regional-specific strategies to maximize air quality improvements. The study's methodology, combining observations, reanalysis data, and machine learning, provides a robust framework for attributing air quality changes to multiple drivers. This approach can be adapted to analyze similar events in other regions and assess the effectiveness of environmental policies.

This research demonstrates the multifaceted influences on air quality in China during the COVID-19 lockdown. The study successfully disentangled the impacts of CNY, CAP, COVID-19 restrictions, and meteorological changes, quantifying their contributions to NO2 and O3 variations. Significant reductions in multi-pollutant health risks were observed, primarily due to NO2 reductions. However, regional disparities emerged, with unfavorable meteorological conditions affecting some areas. The findings highlight the effectiveness of the CAP and the public health benefits of even temporary emission reductions, emphasizing the need for region-specific adaptation strategies and policies to account for future meteorological changes and maintain improvements in air quality.

The study's conclusions are based on data from a specific time period and region. Generalizing the results to other contexts may require further investigation. Data limitations regarding Laiwu and the absence of data from some regions of China might have slightly affected the overall assessment. While the machine learning model effectively predicts pollutant concentrations, some uncertainties inherent to ML approaches remain. The health impact assessment relies on statistical relationships between pollutants and mortality, which may not capture the full complexity of health effects.