Impact of the COVID-19 pandemic on air pollution in Chinese megacities from the perspective of traffic volume and meteorological factors.

During 2020, the COVID-19 pandemic resulted in a widespread lockdown in many cities in China. In this study, we assessed the impact of changes in human activities on air quality during the COVID-19 pandemic by determining the relationships between air quality, traffic volume, and meteorological conditions. The megacities of Wuhan, Beijing, Shanghai, and Guangzhou were selected as the study area, and the variation trends of air pollutants for the period January-May between 2016 and 2020 were analyzed. The passenger volume of public transportation (PVPT) and the passenger volume of taxis (PVT) along with data on precipitation, temperature, relative humidity, wind speed, and boundary layer height were used to identify and quantify the driving force of the air pollution variation. The results showed that the change rates of fine particulate matter (PM2.5), NO2, and SO2 before and during the lockdown in the four megacities ranged from -49.9% to 78.2% (average: -9.4% ± 59.3%), -55.4% to -32.3% (average: -43.0% ± 9.7%), and - 21.1% to 11.9% (average: -10.9% ± 15.4%), respectively. The response of NO2 to the lockdown was the most sensitive, while the response of PM2.5 was smaller and more delayed. During the lockdown period, haze from uninterrupted industrial emissions and fireworks under the effect of air mass transport from surrounding areas and adverse climate conditions was probably the cause of abnormally high PM2.5 concentrations in Beijing. In addition, the PVT was the most significant factor for NO2, and meteorology had a greater impact on PM2.5 than NO2 and SO2. There is a need for more national-level policies for limiting firework displays and traffic emissions, as well as further studies on the formation and transmission of secondary air pollutants.

Introduction

Over the past 40 years, China has experienced rapid economic growth and accelerated urbanization. By the end of 2018, the urbanization rate was 58.5%, which was 33.4% higher than that at the end of 1978 (Central People's Government of the People's Republic of China, 2019). Rapid development has also created severe air pollution in China. Several studies have reported the association of mortality and morbidity with air pollution (Chen et al., 2017; Lelieveld et al., 2019); therefore, air pollution has been an increasing concern and mitigating action is urgently required.

Controlling anthropogenic emissions can prevent serious air pollution events. The effect of vigorous air pollution control measures during large-scale societal events has been studied. For instance, remarkable improvements in air quality were observed in Beijing during the 2008 Olympic games (Wang et al., 2010), the 2014 Asia-Pacific Economic Cooperation (APEC) meeting, and the 2015 China Victory Day Parade (Lin et al., 2017). The coronavirus disease (COVID-19) broke out in late December 2019 in Wuhan, Hubei Province, China (Li et al., 2020). Due to the extreme infectivity of COVID-19, Wuhan was first placed under lockdown on January 23, 2020. This was subsequently extended to ~33% of Chinese cities, thus strictly curtailing personal mobility and economic activities (He et al., 2020). Megacities such as Beijing, Shanghai, and Guangzhou launched the first-level response to this major public health emergency on January 23 and 24, 2020, respectively. On January 30, 2020, the World Health Organization (WHO) declared COVID-19 as a public health emergency of international concern in light of the global threat (Sohrabi et al., 2020). Then, on March 11, 2020, the WHO announced COVID-19 as a pandemic. During the COVID-19 pandemic in China, drastic measures were adopted to curb the spread of the virus. The most important measure involved reducing human interaction by enforcing strict quarantines, forbidding private and public gatherings, restricting traffic, and shutting down enterprises. In particular, the cities of Hubei Province were under the strictest shutdown. On April 8, 2020, Wuhan lifted the lockdown restrictions, while other cities began to resume work and production after February 9, 2020.

Although many countries have suffered huge losses, air quality improvements during the COVID-19 lockdown have been reported in a number of cities across the world (Chauhan and Singh, 2020; Filonchyk et al., 2020; Rahman et al., 2020; Rodriguez-Urrego and Rodriguez-Urrego, 2020). In China, some studies on the effect of lockdown on air quality have also been reported (He et al., 2020; Le et al., 2020). However, there are limited longer-term studies on the influence of changes in human activities on air quality, including the period of resuming work and production, with the change of traffic and season taken into consideration.

Quantitative assessments of air pollution are required to understand if the nationwide lockdown significantly improved air quality. The main objectives of this study are: (1) to compare the average concentrations of fine particulate matter (PM2.5), SO2, and NO2 in four megacities in China (Wuhan, Beijing, Shanghai, and Guangzhou) before the lockdown with those during the lockdown and in the period of resuming work and production; (2) to identify the potential sources of abnormally high PM2.5 in Beijing during the lockdown by using the Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model and Concentration Weighted Trajectory (CWT) model, and (3) to examine the effects of traffic volume and meteorological conditions on regional concentrations of PM2.5, SO2, and NO2.

Materials and method

Study area and data source

The four megacities of Wuhan, Beijing, Shanghai, and Guangzhou were selected as the study area (Fig. 1 ). Wuhan covers a total area of 8569.2 km2, and was the Chinese city most seriously affected by the COVID-19 pandemic. The permanent population in Wuhan reached 11.2 million by the end of 2019, with an urbanization rate of 80.5%. Beijing, Shanghai, and Guangzhou are the three most developed cities in mainland China, and the urbanization rate of these three cities all exceeded 86% by the end of 2019. Located in North China, Beijing is the capital city, with a total area of 16,410.5 km2 and a population of 21.5 million. Shanghai and Guangzhou are coastal cities located in East China and South China, respectively, with total areas of 6340.5 km2 and 7434.4 km2, and populations of 24.2 million and 15.4 million, respectively.

Fig. 1

Locations of Wuhan, Beijing, Shanghai, and Guangzhou.

The list of data sources in this study is presented in Table S1. Briefly, the PM2.5, NO2, and SO2 data for Wuhan, Beijing, Shanghai, and Guangzhou were obtained from the National Urban Air Quality Real-Time Release Platform, which releases hourly in-situ urban air quality data. There are 10–12 stations in each city, and the hourly data for each city were aggregated by averaging the data of all available stations in the city. The daily PM2.5, NO2, and SO2 concentrations of each city were calculated by averaging the hourly values.

Meteorological factors have been identified as the dominant transmission or dispersion drivers of air pollutants (Mamtimin and Meixner, 2011; Yang et al., 2017). In this study, precipitation (Pre), temperature (Tem), relative humidity (RH), and wind speed (WS) data for the four cities were obtained from the China Meteorological Data Network, which provides daily data from nearly 700 meteorological monitoring stations. It is known that stagnant airflow is not conducive to air pollution dispersion (Le et al., 2020; Song et al., 2017); therefore, monthly-average boundary layer heights (BLH) were also obtained from the Climate Data Store, which is a climate data platform implemented by the European Center for Medium-Range Weather Forecasts (ECMWF).

Traffic emissions are an important source of air pollutants, including NO2 (Agudelo-Castaneda et al., 2020; Ielpo et al., 2019) and particulate matter (Gertler et al., 2000). During the COVID-19 pandemic in China, the Chinese government implemented strict controls over communities, including travel. Given the availability of traffic data, two traffic indicators at the city scale were selected to explore the relationships between traffic and PM2.5, NO2, and SO2. The monthly passenger volume of public transportation (PVPT) is the total number of passengers who took buses and rail transit in each city in a month, while the monthly passenger volume of taxis (PVT) is the total number of passengers who took taxis in each city in a month. The monthly PVPT and PVT of the four cities were obtained from the Ministry of Transport of the People's Republic of China, which provides monthly traffic volumes of provinces and central cities in China.

Data processing and analysis

The study period of January 1, 2020, to May 31, 2020, was divided into six periods so that the PM2.5, NO2, and SO2 concentrations before the lockdown could be compared with those during the lockdown and in the period of work resumption. The first period (January 1–22) was considered as the period before the lockdown (hereafter “pre-LD”). Because air pollutants would probably have been influenced by human activities during the holiday period, the second period was the Chinese New Year Festival (CNYF) (January 24 to February 8), and the third period was February 9–29. March, April, and May were regarded as the fourth, fifth, and sixth periods, respectively. In addition, the data for 2016–2019 during the same period with 2020 in the Chinese lunar calendar were taken as the historical reference period. The Wilcoxon test was applied to test whether there was a significant difference in the air pollution levels before and during the lockdown.

Air pollutants can be formed by local emissions or transmitted over long distances (Dimitriou et al., 2015; Yassin et al., 2018). In this study, the HYSPLIT model was used to identify the source of pollutants during a serious pollution period. The HYSPLIT model is commonly used to identify the origin of air mass trajectories and establish a source–receptor relationship (Yassin et al., 2018), and has been frequently used to assess long-range transport patterns of air pollutants (Benchrif et al., 2018). The NCEP/NCAR reanalysis data were used for trajectories obtained from the National Oceanic and Atmospheric Administration (NOAA). Clustering based on the Euclidean system was applied to categorize the trajectories into distinct transport patterns. The CWT model is a statistical tool used to identify the major source by weighting trajectories with associated concentrations (Hsu et al., 2003). Three-day backward trajectories arriving at Beijing were computed at 12:00 Coordinated Universal Time (UTC) at an elevation of 500 m above ground level on a 0.5° × 0.5° resolution grid. The trajectories and CWT were simulated by the Trajstat (Eq. (1)) (Wang et al., 2009) plug-in in MeteoInfo, with an arbitrary weight function (Eq. (2)) (Dimitriou et al., 2015; Hsu et al., 2003) to reduce the uncertainty of cells with few endpoints. where C is the weighted average concentration in grid cell i, j; C is the measured concentration at the sampling site during day k; and τ is the residence time of the 24-h backward trajectories corresponding to day k in grid cell i, j (Dimitriou et al., 2015).

To explore the factors affecting the change in air pollutants, Spearman correlations were determined between the traffic volume (PVPT and PVT), meteorological factors (Pre, Tem, RH, WS, and BLH), and air pollutants (PM2.5, NO2, and SO2). Air pollutants, Tem, RH, WS, and BLH for each month were taken as the monthly average values, while the traffic volume and Pre were taken as the monthly cumulative values. Furthermore, to identify and quantify the driving force of the air pollution change trend in the selected cities from January to May 2020, a stepwise regression with backward elimination method was applied to select variables from the PVT, Pre, Tem, RH, WS, and BLH variables. All the independent variables were entered into the model first, and were deleted if they did not contribute to reducing the Akaike information criterion (AIC) of the model. No more than three variables were retained to avoid overfitting. The statistical analyses were run in R.

Results and discussion

Variation trends of atmospheric PM2.5, NO2, and SO2

Overall variation trends of PM2.5, NO2, and SO2 from January to may

During the CNYF (second period, January 24–February 8), the average PM2.5 concentrations in Wuhan, Shanghai, and Guangzhou decreased by 17.2%, 32.1%, and 35.3%, respectively, with respect to the pre-LD concentrations (January 1–22) (Fig. 2 , Table S2). In addition, the average PM2.5 concentrations in Wuhan and Shanghai reduced further during the third period (February 9–29) to <35 μg/m3, thus meeting the level II standard of the Environmental Air Quality Standard of China (GB3095-2012). The average PM2.5 concentration in Wuhan subsequently remained stable, while that in Shanghai reached its lowest value in March before increasing. Moreover, the average PM2.5 concentration in Guangzhou remained lower than the level II standard from January to May 2020. In contrast to these three cities, the average PM2.5 concentration in Beijing increased considerably (71.4%) during the CNYF with respect to the pre-LD concentration. The average PM2.5 concentration remained high in the third period (February 9–29) before reducing in March to meet the level II standard.

Fig. 2

Variation trends of average PM2.5, NO2, and SO2 concentrations in historical years (2016–2019) and 2020. pre-LD: the period before lockdown; CNYF: Chinese New Year Festival; 2016–2019: the average of four year's data.

The average pre-LD NO2 concentration in each city exceeded the standard of 40 μg/m3, while the average SO2 concentration in each city satisfied the level I standard (i.e., <20 μg/m3) (Fig. 2). In the second period (CNYF), the average NO2 concentrations in Wuhan, Beijing, Shanghai, and Guangzhou decreased sharply relative to the pre-LD NO2 concentrations by 49.4%, 42.0%, 50.3%, and 66.1%, respectively. The average NO2 concentration in Wuhan decreased further during the third period (February 9–29) before increasing, while those in Shanghai and Guangzhou started to increase in the third period. The average NO2 concentration in Beijing remained relatively stable between the second and sixth periods. Excluding Wuhan, the changing trends of SO2 were similar to those of NO2 in the other three cities, but with a smaller rate of decrease. Furthermore, the changing trends of NO2 and SO2 in Shanghai were homogeneous compared to those in Guangzhou. In addition to Hubei Province, a study found that Beijing and its surrounding provinces recovered more slowly due to the extension of lockdown (Zheng et al., 2020). Therefore, air pollution in Beijing rebounded more slowly than the other cities during the study period (Fig. 2).

The average SO2 concentration in Wuhan was the same in the pre-LD and CNYF periods (6.8 μg/m3), and then increased slightly to 9.5 μg/m3 in April 2020. It has been reported that SO2 is not only affected by human emissions and meteorological factors within a city, but also by the SO2 levels of adjacent cities (Yang et al., 2017). Fig. S1 shows the spatial characteristics of the SO2 concentration across mainland China from January to March 2020 (i.e., before and during lockdown), indicating lower SO2 levels in South China, although the decrease in SO2 levels in North China was greater. The changing trend of the atmospheric SO2 concentration in Wuhan was similar to that in its surrounding cities, with adjacent cities such as Ezhou and Huangshi always exhibiting higher SO2 concentrations than those in Wuhan. Huangshi is an industrial city in Hubei Province; thus, it is likely that the increased SO2 concentration in Wuhan was caused by SO2 emissions in the surrounding cities.

In comparison to NO2 and SO2, the composition of PM2.5 is much more complicated. Le et al. (2020) simulated organic aerosol, nitrate (NO3 −), and sulfate (SO4 2−), and found them to be the predominant surface aerosol species in the Beijing–Tianjin–Hebei region during the lockdown period. The major sources of PM2.5 can be summarized as secondary aerosols, coal combustion, traffic emissions, biomass burning, industry emissions, and soil dust (Tao et al., 2014; Wang et al., 2008; Wang et al., 2005). Secondary aerosol particles are the main source of PM2.5 (Wang et al., 2008). Huang et al. (2014) found that severe haze pollution events were driven to a large extent by secondary aerosol formation. Thus, limited by the more complex chemical composition and formation process, the response of PM2.5 to the lockdown period and the resumption of work and production was more delayed than the responses of NO2 and SO2.

The differences in the average concentrations of PM2.5, NO2, and SO2 between the study period in 2020 and the same period during 2016–2019 exhibited descending trends overall. Air pollutants vary seasonally because of variations in meteorology (Mallik and Lal, 2014; Wang et al., 2005; Xiang et al., 2019; Yang et al., 2017). The peak concentrations of air pollutants usually occur in winter (Duo et al., 2018; Ma et al., 2019) as a result of enhanced coal/biomass consumption for residential heating and adverse climate conditions (e.g., low wind speed, precipitation, and BLH), which mean that pollutants are poorly dispersed in winter (Ma et al., 2019). In April and May, though most of PM2.5, NO2 and SO2 were lower than pre-LD, the differences between current and historical data were much lower (Fig. 2), thus implying that favorable climate conditions played an important role in cleaning air pollutants in the process of work resumption.

Comparing the changes of air pollution concentrations during lockdown in different areas

In this study, in order to remove the influence of festival activities on data interpretation, the third period (February 9–29) was taken as the main period under lockdown. Compared with the pre-LD period, the relative changes of the concentrations of PM2.5, NO2, and SO2 during the third period in the four cities ranged from −49.9% to 78.2% (average: −9.4% ± 59.3%), −55.4% to −32.3% (average: −43.0% ± 9.7%), and − 21.1% to 11.9% (average: −10.9% ± 15.4%), respectively. Except for PM2.5 in Beijing and SO2 in Wuhan and Beijing, the air pollutant concentrations decreased significantly (p < 0.05) between the pre-LD period and third period.

In addition, the relative changes between the historical average concentrations of PM2.5, NO2, and SO2 and those during the third period ranged from −53.8% to −5.1% (average: −31.9% ± 20.2%), −65.4% to −41.3% (average: −48.5% ± 11.3%), and − 64.2% to −44.8% (average: −53.4% ± 8.0%), respectively. In the third period, except for PM2.5 in Beijing, the average air pollutant concentrations decreased significantly compared with the historical period. However, in comparison to January–May 2019, the reductions during the study period were not significant for PM2.5 in Beijing and Guangzhou, and SO2 in Wuhan and Guangzhou. China has implemented numerous actions to reduce air pollution in recent years, and corresponding reductions have been reported in several studies (Lu et al., 2020; Zhai et al., 2019), especially for SO2 and PM2.5 (Li et al., 2018; Tian et al., 2020), while the observed decrease in NO2 has been comparatively smaller. During the lockdown, most of the reduction compared with historical years in atmospheric PM2.5, NO2, and SO2 were greater than that during pre-LD (Table S2), and most of the average air pollutant concentrations in 2020 were the lowest (Fig. 3 ). These findings suggested that the lockdown restrictions had a positive effect on air pollution in the four cities, especially on NO2 emissions.

Fig. 3

Variation trends of average PM2.5, NO2 and SO2 concentrations for the period January–May (2016–2020).

Changes in air pollutants during the lockdown have been reported for many cities in several countries (Table S3). Otmani et al. (2020) showed that the difference between the PM10, SO2, and NO2 concentrations before and during the lockdown in Salé City (Morocco) was 75%, 49%, and 96%, respectively. In the National Capital Territory (NCT) of Delhi (India), the average concentrations of PM2.5, NO2 and SO2 decreased by 53.11%, 52.68% and 17.97%, respectively, during the lockdown compared to before the lockdown (Mahato et al., 2020). In addition, there was a 35% reduction in NO2 concentrations in Almaty, Kazakhstan, during the lockdown (Kerimray et al., 2020). The rate of change in Dhaka City also showed 26.0%, 20.4%, 17.5%, 9.7%, and 8.8% declines in PM2.5, NO2, SO2, O3, and CO concentrations, respectively (Rahman et al., 2020). The regions under partial lockdown, such as São Paulo state, New York, Los Angeles (Chauhan and Singh, 2020), and Victoria (Mexico) (Tello-Leal and Macías-Hernández, 2020) also witnessed a reduction in air pollution.

Furthermore, differences between historical and recent data have also been reported. For example, the average PM2.5 concentration during the lockdown was reduced by 32.62% compared to the same period in 2017–2019 in NCT Delhi (India) (Mahato et al., 2020). The results for Almaty in Kazakhstan showed that the PM2.5 concentration during the lockdown was reduced by 21% compared to the average for the same days in 2018–2019 (Kerimray et al., 2020). In addition, a 25.5% reduction in the county level NO2 concentration was observed in the United States during the COVID-19 pandemic compared to the historical average for 2017–2019. There was also an overall (not significant) decline in PM2.5 (p = 0.07) (Berman and Ebisu, 2020). Similarly, cities under partial lockdown, including São Paulo state, New York, Los Angeles, Dubai, Delhi, and Mumba, also experienced a reduction in air pollution compared to previous years (Chauhan and Singh, 2020; Nakada and Urban, 2020).

Compared to the period before the lockdown in 2020, the NO2 concentrations in Salé (Morocco), NCT Delhi (India), and Dhaka (Bangladesh) also showed a greater reduction than SO2, which was consistent with Chinese megacities. Different from cities in other countries, the PM2.5 concentration in Beijing increased during the lockdown. Moreover, the reductions in the average PM2.5, SO2, and NO2 concentrations during the period January–May 2020 relative to the average historical values in China were greater than those reported for other countries, which may have been partially due to the implementation of emission control plans in China.

Abnormal increase of PM2.5 concentrations in Beijing during lockdown

An abnormal increase in the average atmospheric PM2.5 concentration occurred in Beijing during the second (CNYF) and third periods, which also occurred during the same periods in the historical data (Fig. 2). However, the abnormally high PM2.5 concentrations in Beijing during the lockdown exceeded the historical average values for the same period. During the CNYF, the relative humidity was much higher than that in the historical period (Table S4). During the lockdown period in North China, anomalously high humidity, wind conditions, a decreased BLH, and increasing ozone concentrations all contributed to the formation of secondary aerosols (Le et al., 2020), which has been reported as the main source of PM2.5 (Wang et al., 2008). Although the emissions from vehicles, processing, and light industries have decreased after the lockdown, the main source of industrial air pollutant emissions is resource-based industries with heavy energy consumption and emission. Industries such as power plants, steel, and coking always have uninterrupted production processes and need to operate all year round (China Environmental News, 2020; Ministry of Ecology and Environment of the People's Republic of China, 2020). In addition, firework displays are a tradition during the Spring Festival, especially on the Chinese New Year's Eve and during the Lantern Festival. During these holidays, fireworks can cause brief, but extraordinarily high levels of pollution (Lai and Brimblecombe, 2020). Pang et al. (2020) pointed out that fireworks were the largest source of PM2.5 during fireworks periods in the Spring Festival. Although firework displays in Beijing's Fifth Ring Road have been banned since 2018 (Hu et al., 2019), firework displays within the suburbs and rural areas are not restricted. In our study, the two sharpest increases of PM2.5 in Beijing occurred after the Chinese New Year's Eve and Lantern Festival (Fig. S2), reaching 171.1 μg/m3 and 206.4 μg/m3, respectively. In addition, the PM2.5/CO ratio can be used to quantify the contribution of fireworks to PM2.5 (Pang et al., 2021; Wang et al., 2014). The average PM2.5/CO in Beijing increased from 4.9% in the pre-LD period to 8.1% during January 24 to February 15 (Fig. S3), indicating that firework displays had a significant impact on PM2.5 pollution in Beijing.

The final mean trajectories resulting from the clustering process were examined for the CNYF and the period from January 24 to February 15. The results revealed that 72.7% of air masses were from the northwest sectors. The CWT model results in Fig. 4 suggest that long-range transport was the main contributor to PM2.5 in Beijing. The most polluted air masses were mainly from Shanxi and Shaanxi provinces, while the trajectories from the southwest only accounted for 14.0%. Inner Mongolia, Hebei, Gansu, and Liaoning provinces were also potential sources of PM2.5 in Beijing. Thus, the probable cause of the abnormally high PM2.5 concentrations in Beijing during the lockdown was inferred to be the haze associated with uninterrupted industrial emissions and firework displays under the effect of air mass transport from surrounding areas, along with adverse climate conditions.

Fig. 4

Mean trajectories arriving at Beijing from January 24 to February 15, 2020. Top: trajectories are classified into six clusters. Bottom: concentration-weighted trajectories.

Relationships between air pollutants, traffic volume, and meteorological factors

The Spearman correlations between traffic volume, meteorological factors, and air pollutants are presented in Fig. 5 . The traffic volume was positively correlated with the three air pollutants, while most of the meteorological conditions were negatively correlated with the three air pollutants. In particular, temperature had a stronger negative influence on PM2.5 in comparison to NO2 and SO2 in all four cities. Precipitation and BLH also had stronger negative relationships with PM2.5 in all cities except for Shanghai. There was a positive relationship between relative humidity and the PM2.5 concentration in all cities except for Guangzhou, which further proved that a high relative humidity is an adverse meteorological factor for reducing PM2.5. In general, the selected meteorological factors had a greater influence on PM2.5 than NO2 and SO2. In addition, the PVPT and PVT both exhibited stronger correlations with NO2 compared with PM2.5 and SO2. This finding is consistent with the results of a previous study (Ielpo et al., 2019), whereby diffusive sources of NO2 were linked mainly to vehicular traffic, while SO2 was mainly associated with industrial sources. Moreover, our results also revealed a stronger correlation between NO2 and PVT than between NO2 and PVPT, which suggests that advocating public transportation could reduce NO2 pollution.

Fig. 5

Spearman correlation coefficients between three air pollutants, traffic volume, and meteorological conditions in four Chinese cities from January to May in 2019 and 2020. PVT: passenger volume of taxis; PVPT: passenger volume of public transportation; Pre: precipitation; Tem: temperature; BLH: boundary layer height; RH: relative humidity; WS: wind speed.

The monthly PVPT and PVT values from January to May in 2020 are shown in Fig. 6 . During the lockdown period, traffic reduced drastically, especially in Wuhan, and was almost shut down from February to March. The lockdown in Wuhan was lifted on April 8; thus, the passenger capacity subsequently began to increase, although it was still quite low compared to the same period in 2019. As the capital of China, the absolute decrease in public traffic in Beijing was the greatest, with the PVPT decreasing by 307–498 million passengers per month from February to May. The relative change in both public traffic and taxis was ranked second after Wuhan. Beijing, Shanghai, and Guangzhou had similar trends in the PVPT and PVT values from January to May 2020, with sharper decreases from January to February compared with the same period in 2019 before they began to increase. Despite the gradual resumption of work and production, the volume of traffic was much lower than that of the same period in 2019. By May, the PVPT and PVT values were 25.1%–65.4% and 47.0%–67.6% of the values during the same period in 2019, respectively.

Fig. 6

Monthly variations of the passenger volume of public transportation (PVPT) (left) and the passenger volume of taxis (PVT) (right) in four megacities of China from January to May in 2019 and 2020.

The change trends of atmospheric NO2 concentrations in Wuhan, Shanghai, and Guangzhou agreed with their PVPT and PVT values, except for NO2 in May, which only slightly increased in Wuhan and decreased in Shanghai and Guangzhou. Unlike other cities, the change in NO2 in Beijing was inconsistent with the change in traffic. Compared with the volume of passengers in February, traffic in Beijing under the strictest control after Wuhan and increased slightly during the next months. Thus, coupled with gradually favorable climatic conditions, NO2 remained stable from February to May 2020 in Beijing.

Furthermore, PM2.5 and SO2 concentrations also exhibited a decrease or slight increase in the four cities in May. This may have been because meteorology was more conducive to the dispersion of air pollutants. For example, monthly precipitation in Guangzhou increased by 420 mm from April to May 2020 (Table S4), which could have removed pollutants from the air (Gao et al., 2019; Kwaka et al., 2017). The variation characteristics of PM2.5, NO2, and SO2 under the comprehensive effect of traffic and meteorology (temperature and precipitation) are shown in Fig. 7 . During the process of work resumption, the increase in anthropogenic emissions was accompanied by gradually favorable meteorology, which would have reduced the effect of anthropogenic emissions on air pollution.

Fig. 7

Combined effect of the passenger volume of taxis (PVT) and meteorology on PM2.5, NO2, and SO2 in four cities from January to May 2020.

Stepwise linear regression with variables including PVT, precipitation, temperature, RH, wind speed, and BLH was applied to identify the driving force of the air pollution change trend in the four cities in 2020 (Table 1 ). The PVT and wind speed were significantly correlated with NO2, while temperature and wind speed were significantly correlated with PM2.5. For every degree increase in temperature, the PM2.5 concentration could reduce by 1.28 μg/m3, while an increase of 1 m/s in wind speed reduced PM2.5 and NO2 concentrations by 11.31 μg/m3 and 7.75 μg/m3, respectively. The PVT was the most significant variable for NO2, further confirming that traffic emissions contributed more to NO2. In addition, a 10 million increase in the PVT could increase PM2.5 and NO2 concentrations by 3.51 μg/m3 (p = 0.059) and 6.93 μg/m3, respectively. There were no significant correlations between SO2 and all other variables, indicating unforeseen factors influencing SO2.

Table 1

Stepwise regression results of PM2.5, NO2, and SO2 from January to May 2020 in four cities.

Variables	PM2.5	NO2	SO2
PVT (10 million passengers)	3.51	6.93⁎⁎⁎	–
Precipitation (mm)	–	–	–
Temperature (°C)	−1.28⁎⁎⁎	–	7.56 × 10−2
BLH (m)	–	–	4.66 × 10−3
Relative humidity (%)	–	–	4.93 × 10−2
Wind speed (m/s)	−11.31⁎⁎	−7.75⁎	–
Intercept	70.63⁎⁎⁎	36.38⁎⁎⁎	4.58
R2	0.73	0.75	0.53
Adjusted R2	0.68	0.72	0.41
AIC	82.49	60.93	13.99
p	< 0.001	< 0.001	< 0.001

p < 0.05.

p < 0.01.

p < 0.001.

From February to March 2020, Wuhan was under full lockdown, while Guangzhou started resuming work. However, the monthly average PM2.5 concentrations in Wuhan in January (38.1 μg/m3) and February (34.5 μg/m3) were both higher than those in Guangzhou (23.8 μg/m3 and 21.2 μg/m3), and the monthly average NO2 concentrations in Guangzhou were both higher than those in Wuhan. In contrast to Wuhan, Guangzhou is a coastal city in South China, where the temperature, wind speed, and traffic emissions are higher. The linear regression model indicated that meteorology had a greater effect on PM2.5 than NO2, while traffic emissions affected NO2 more than PM2.5. For example, from the coefficients of the model, in March 2020, the simulated effect of temperature and wind speed on reducing PM2.5 in Guangzhou was 14.0 μg/m3 larger than that in Wuhan. In addition, in contrast to the pre-LD period, the NO2 concentration in Wuhan decreased by 24.3 μg/m3 and 22.5 μg/m3 during 2/9–2/29 and March, respectively. The contribution of wind speed to these reductions in NO2 was 1.6 μg/m3 (2/9–2/29) and 0.8 μg/m3 (March), while the change of human activities reduced the NO2 concentration by 22.8 μg/m3 (2/9–2/29) and 21.7 μg/m3 (March).

In recent years, PM2.5 has been the most prominent air pollution issue (Lu et al., 2020). Studies have reported that airborne particles might be possible routes of COVID-19 diffusion (Liu et al., 2020; Zoran et al., 2020), and that particulate matter is directly related to COVID-19 mortality (Magazzino et al., 2020; Mele and Magazzino, 2020; Zhu et al., 2020). Under the effect of targeted air pollution control, PM2.5 and SO2 have reduced remarkably over the past years; however, the PM2.5 pollution in many Chinese cities, especially those in or around Hebei Province, is still serious. The geographical distribution of industrial structures can be optimized from the perspective of local climate characteristics. Moreover, the decreases in NO2 and NOX have been found to be comparatively smaller in recent years (Li et al., 2018; Liu et al., 2017; Tian et al., 2020). Air pollution has gradually transitioned from the traditional coal-fired form to a coal-fired/motor vehicle hybrid emission form (Kan et al., 2009), and the previous control of vehicle emissions failed to reduce NOX effectively (Wu et al., 2016). Therefore, more effective national-level control actions for reducing NO2 emissions need to be implemented in the future, especially with respect to traffic emissions.

Conclusions

From the overall variation trend of air quality, the air quality in four megacities (Wuhan, Beijing, Shanghai, and Guangzhou) in China after the outbreak of COVID-19 improved. The exceptions were SO2 in Wuhan, which presented an ascending trend, and PM2.5 in Beijing, which exhibited a delayed decrease. After the lockdown, the three air pollutants in these four cities largely satisfied the standards of the Environmental Air Quality Standard of China; however, this was at the expense of the economy and normal population movement. In general, the largest reductions of PM2.5 and NO2 were in Wuhan, and the response of the NO2 concentration to the changes in human activities during the COVID-19 pandemic was the most sensitive of the three air pollutants.

The abnormal increase of PM2.5 in Beijing was probably caused by transported pollutants produced by uninterrupted industrial emissions and fireworks during New Year's Eve and the Lantern Festival, as well as the influence of adverse weather conditions. The contribution of fireworks to PM2.5 could not be ignored, even when the cities were under lockdown. Many large cities have banned or limited fireworks during festivals, whereas this is not the case in the suburbs and rural areas. More national-level policies for limiting firework displays along with further studies on the formation and transmission of secondary air pollutants are needed.

Although these four cities were gradually returning to work and production from February to May, the traffic volume was still lower than that in the same period in 2019, especially in Wuhan and Beijing. The traffic volume had a stronger correlation with NO2, while meteorological factors had a greater influence on PM2.5 than NO2 and SO2. Seasonal changes (from winter to summer) along with gradually favorable meteorological factors contributed to reducing the effect of anthropogenic emissions on air pollution. Controlling PM2.5 pollution is still the most challenging issue in Chinese cities, and the geographical distribution of industry could be optimized from the perspective of local climate characteristics. In addition, national-level control actions on reducing NO2 emissions need to be implemented in the future, especially with respect to traffic emissions.

This study contributes to our understanding of the variation trend during the COVID-19 pandemic. Although the impact of festival activities on atmospheric pollution during the lockdown was assessed by comparing the pre-LD data with the data of February 9–29, Beijing, Shanghai, and Guangzhou started to resume work and production at a small scale during that period; therefore, our results may underestimate the reduction of air pollutants. In addition, the PVPT and PVT are imperfect metrics due to the lack of data on the usage of private cars. Finally, limited by the availability of data, the effect of anthropogenic emission from industrial and residential activities on air pollution was not examined in this study. The effect of anthropogenic emissions on air pollution should be assessed with more detailed information.

CRediT authorship contribution statement

Chanchan Gao: Conceptualization, Methodology, Visualization, Writing – original draft, Writing – review & editing. Shuhui Li: Writing – original draft. Min Liu: Conceptualization, Methodology, Writing – original draft, Writing – review & editing, Supervision. Fengying Zhang: Resources. V. Achal: Writing – review & editing. Yue Tu: Visualization. Shiqing Zhang: Formal analysis. Chaolin Cai: Visualization.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.