Quantifying Air Pollutant Variations during COVID-19 Lockdown in a Capital City in Northwest China

In the context of the outbreak of coronavirus disease 2019 (COVID-19), strict lockdown policies were implemented to control nonessential human activities in Xi’an, northwest China, which greatly limited the spread of the pandemic and affected air quality. Compared with pre-lockdown, the air quality index and concentrations of PM2.5, PM10, SO2, and CO during the lockdown reduced, but the reductions were not very significant. NO2 levels exhibited the largest decrease (52%) during lockdown, owing to the remarkable decreased motor vehicle emissions. The highest K+ and lowest Ca2+ concentrations in PM2.5 samples could be attributed to the increase in household biomass fuel consumption in suburbs and rural areas around Xi’an and the decrease in human physical activities in Xi’an (e.g., human travel, vehicle emissions, construction activities), respectively, during the lockdown period. Secondary chemical reactions in the atmosphere increased in the lockdown period, as evidenced by the increased O3 level (increased by 160%) and OC/EC ratios in PM2.5 (increased by 26%), compared with pre-lockdown levels. The results, based on a natural experiment in this study, can be used as a reference for studying the formation and source of air pollution in Xi’an and provide evidence for establishing future long-term air pollution control policies.


Introduction
Since the Industrial Revolution in the 18th century, industrialization has transformed production patterns and lifestyles in society. However, industrialization and modernization have led to major environmental problems [1][2][3]. Anthropogenic emissions, including vehicle and industrial exhaust emissions, fossil fuel combustion, resident smoking, and household heating, are an important cause of deteriorating air quality [4][5][6].
Ghaffarpasend et al. [7] concluded that motor vehicle emissions accounted for an average of 45% of the air pollutants in Tehran, Iran. Industrial processes contributed 10.7% of particulate matter with aerodynamic diameter equal to or less than 2.5 µm (PM 2.5 ), and fossil fuel combustion contributed 15.8% of particulate matter with aerodynamic diameter equal to or less than 10 µm (PM 10 ) emissions in Shandong Province of China [3]. Moreover, Zhang et al. [2] reported that industrial processes were the main source of organic carbon (OC) emissions in total suspended particulate (TSP), accounting for 23.6% of the total OC emission in Henan Province, China. The emission inventory reported by Zhong et al. [4] proved that the contribution rate of power plants and industrial combustion 3.3%, and the output values of light and heavy industries decreased by 23.5% and 13.2%, respectively [24], and the number of construction projects decreased by 16.3% compared with that in the same period in 2019. Among them, industrial construction projects reduced by 31.2% [27].
The lockdown brought about drastic impacts at social and economic fronts [28][29][30][31] as well as impacts on the environment, particularly in the context of air quality, environmental, and noise reduction [32][33][34]. To date, previous data and studies have shown that the emergency measures taken by the government to prevent human activities in the lockdown period of COVID-19 had effectively and significantly reduced ambient air pollution [20,[35][36][37][38][39][40][41]. However, no comprehensive investigation has been conducted on the impact of COVID-19 control measures on gaseous and particulate air pollutants in Xi'an. In the present study, offline PM 2.5 filters and online air pollutant monitoring records were collected simultaneously in Xi'an from 1 January to 7 March 2020. The main objective of this study was to determine the variation in air pollutants, including PM 2.5 , OC, elemental carbon (EC), water-soluble ions (WSIs), PM 10 , and gaseous pollutants (SO 2 , NO 2 , CO, and O 3 ) in Xi'an in relation to the restrictive anthropogenic activities before, during, and after the COVID-19 lockdown. Additionally, this study referenced and categorized relevant domestic and international results (also covering previous years and months of the lockdown period) to summarize the characteristics of various air pollutants in several regions and to obtain a more in-depth understanding of air quality improvement and PM 2.5 compositions in Xi'an during this event. The lockdown caused by the COVID-19 pandemic provided an opportunity to perform a natural experiment for evaluating air quality responses to drastic emission reduction, and it is helpful to formulating more targeted policies in this heavily polluted area and sustainable development [33,42,43]. In line with this, future awareness campaigns should focus more on a multidisciplinary area in practitioners from all walks of life towards Penta Helix Collaboration [44][45][46] in the post-COVID-19 world [47,48].

Experimental Design
Our study focused on air quality variation during the COVID-19 lockdown in the city of Xi'an (the capital of Shaanxi Province), China. The COVID-19 lockdown period in this study was divided into three time intervals, namely pre-lockdown (1 January to 23 January 2020), during lockdown (24 January to 13 February 2020), and post-lockdown (14 February to 7 March 2020). Air quality was expected to improve because a series of policies had been implemented to control human activities ( Figure 1). In this study, offline PM 2.5 filter samples were collected, and the carbonaceous fraction and WSIs were analyzed. Simultaneously collected online and offline air quality data during the study period were used for comparison. The aforementioned information was processed to study the changes in air pollution sources and the improvement in air quality during the COVID-19 lockdown.

PM 2.5 Sample Collection
Daily PM 2.5 samples were collected on the roof top of a five-storied building (16.3 m above the ground) on the campus of Xi'an Jiaotong University (108.990 • E, 34.252 • N) that is surrounded by residential areas and other campus buildings, and is approximately 200 m away from the Xingqing Road and South Second Ring Road (Figure 2), which have heavy traffic, making it a suitable region with a mixture of mobile emission and stationary emission sources. Twenty-four-hour PM 2.5 samples (10:00 am to 10:00 am next day local time) were collected using pre-fired (780 • C, 3 h) 90 mm PALLFLEX TISSUQUARTZ filters (QM/A, PALL, Ann Arbor, MI, USA) with the HY-100SFB high-load PM sampler at a flow rate of 100 L min −1 from 1 January 2020 to 7 March 2020. A total of 67 PM 2.5 samples and 3 field blank filters (1 for each period) were collected in this study. The final data were obtained by subtracting all field blanks to avoid any artifacts induced by gas absorption.

PM2.5 Sample Collection
Daily PM2.5 samples were collected on the roof top of a five-storied building (16.3 m above the ground) on the campus of Xi'an Jiaotong University (108.990° E, 34.252° N) that is surrounded by residential areas and other campus buildings, and is approximately 200 m away from the Xingqing Road and South Second Ring Road (Figure 2), which have heavy traffic, making it a suitable region with a mixture of mobile emission and stationary emission sources. Twenty-four-hour PM2.5 samples (10:00 am to 10:00 am next day local time) were collected using pre-fired (780 ℃, 3 h) 90 mm PALLFLEX TISSUQUARTZ filters (QM/A, PALL, Ann Arbor, MI, USA) with the HY-100SFB high-load PM sampler at a flow rate of 100 L min −1 from 1 January 2020 to 7 March 2020. A total of 67 PM2.5 samples and 3 field blank filters (1 for each period) were collected in this study. The final data were obtained by subtracting all field blanks to avoid any artifacts induced by gas absorption.

Pre-lockdown Dur-lockdown Post-lockdown
Attractions and performances were closed or suspended temporarily [49].
The first level emergency response to public health emergencies was initiated [50].
The number of people going out per household per day was limited [52].

Feb. 14
Hubei Province, the epicenter of the pandemic, began to resume work. 34 large enterprises had resumed work in Xi'an [53].

Feb. 19
The resumption rate of major projects in Xi'an reaches 80% [54].

Feb. 28
The current traffic flow had recovered to 80% of the pre-holiday [55].

Gravimetric and Chemical Analyses
Gravimetric analysis: The PM2.5 samples required to be equilibrated at 20-23 ℃ and 35-45% of relative humidity for 24 h. Then, filters were weighed for mass concentration determination using a Sartorius LA 130S-F (Sartorius, Germany) electronic microbalance (sensitivity: 0.1 mg). Each filter was weighted at least four times (two times before sampling and two times after sampling), and the weight of PM2.5 was obtained by subtracting the pre-sampling weights from the post-sampling weights. The mass concentration of PM2.5 was obtained by dividing the weight mass by the sampling volume.
OC and EC analyses: A 0.5 cm 2 punch was cut from each filter and placed into the Desert Research Institute Model 2001 Thermal/Optical Carbon Analyzer (Atmoslytic Inc.,

Gravimetric and Chemical Analyses
Gravimetric analysis: The PM 2.5 samples required to be equilibrated at 20-23 • C and 35-45% of relative humidity for 24 h. Then, filters were weighed for mass concentration determination using a Sartorius LA 130S-F (Sartorius, Germany) electronic microbalance (sensitivity: 0.1 mg). Each filter was weighted at least four times (two times before sampling and two times after sampling), and the weight of PM 2.5 was obtained by subtracting the pre-sampling weights from the post-sampling weights. The mass concentration of PM 2.5 was obtained by dividing the weight mass by the sampling volume.
OC and EC analyses: A 0.5 cm 2 punch was cut from each filter and placed into the Desert Research Institute Model 2001 Thermal/Optical Carbon Analyzer (Atmoslytic Inc., Calabasas, CA, USA) for the OC and EC analyses in PM 2.5 following the IMPROVE_A (Interagency Monitoring of Protected Visual Environment) thermal/optical reflectance protocol. The filter was heated gradually and analyzed first in an oxygen-pure He atmosphere, and OC1, OC2, OC3, and OC4 were obtained at 140, 280, 480, and 580 • C, respectively. Then, OP (carbon formed during the cracking process of OC) and EC were analyzed in a He atmosphere containing 2% oxygen. EC1, EC2, and EC3 were obtained at 580, 740, and 840 • C in a step-by-step manner. The detection limits for OC and EC were 0.82 and 0.20 µg·m −2 , respectively. Details of quality assurance and quality control (QA/QC) are provided in Cao et al. [56] and Xu et al. [57].

Online Data Collection
Online air quality index (AQI), PM 2.5 , PM 10 , SO 2 , NO 2 , CO, and O 3 _8h data were obtained from the nearest air quality monitoring station; these data were downloaded from the website of China Environmental Monitoring Center [59]. Mann-Whitney U test was performed on online data to determine whether there are significant differences for AQI, and online, six national controlled air pollutants between adjacent time periods, p < 0.05 (*) is considered to be statistically significant. The meteorological data {i.e., air temperature (AT), relative humidity (RH), prevailing wind direction (PWD), and wind speed (WS)) used in this study (Table 1) were derived from the air quality monitoring station [60].

AQI and Online Six National Controlled Air Pollutants
Comparisons of AQI and six national controlled air pollutants (PM 2.5 , PM 10 , SO 2 , CO, NO 2 , and O 3 _8h) in the pre-lockdown, during lockdown, and post-lockdown periods of COVID-19 in Xi'an in 2020 are shown in Figure 3. The AQI decreased in sequence before, during, and after the COVID-19 lockdown with statistical difference between periods of post-lockdown and during lockdown, and air quality gradually improved from moderately polluted (AQI: 151-200) to mildly polluted (AQI: 101-150) and further, to good air quality (AQI: 51-100; Figure 3a). In addition to the AQI, PM 2.5 , PM 10 , SO 2 , and CO exhibited gradually decreasing trends, but surprisingly, they did not present the lowest values during the COVID-19 lockdown (Figure 3b,c,e,f). This may be due to the adverse meteorological factors and the "delayed effect" of pollutant reduction. Different from the trends of the abovementioned air pollutants, NO 2 dropped from 56.7 µg m −3 in prelockdown to the lowest value in during lockdown (27.3 µg m −3 ), and then increased to 32.7 µg m −3 in post-lockdown ( Figure 3g) due to the most direct relationship with motor vehicle primary emissions. Among all the national controlled air pollutants, NO 2 decreased the most during the COVID-19 lockdown to 52% in a statistically significant way. Travel restrictions during the lockdown caused the most significant reduction in NO 2 , consistent with previous studies [39,40,[61][62][63].
Atmosphere 2021, 12, x FOR PEER REVIEW 7 of 17 than in pre-lockdown was mainly due to the changes in particle emissions, especially enhanced emission from anthropogenic sources and the secondary formation of PM2.5 (discussion below). The primary emission should be reduced during the lockdown (Figure 1; restriction on travel and temporary suspension of industries, factories, and construction sites); thus, the elevated PM2.5 proportion may attribute to an enhanced secondary reaction during the lockdown, that is, increased oxidation in the atmosphere. This is more evident in an explanation of O3 variations below. Contrary to the trends of other air pollutants mentioned above, the concentration of O3_8h was the highest in the lockdown period (88.0 μg m −3 ), which was 2.6 times that in the pre-lockdown period, and the difference was significant. The increase in the O3 concentration during the lockdown can be explained as follows. As mentioned earlier, NOx emissions are closely related to motor vehicle emissions enhanced by transportation activities and human travel. However, there are various sources of volatile organic compounds (VOCs), and their emissions varied during this event. For example, one of the main emission sources of VOCs is the evaporation of industrial solvents; the use of industrial solvents did not decrease as much as transportation did in the lockdown period. Therefore, the reduction in NOx was more significant than that in VOCs. This leads to a weakening of the titration of NOx, which reduces the depletion of ozone [65]; therefore, the concentration of O3 in the atmosphere increased during the lockdown period. Simultaneously, the O3 production and oxidation capacity (Ox) during the day increased in this study, which promoted the concentration of OH radicals during the day and NO3 radicals at night. As a result, the increased oxidation capacity of the atmosphere promoted the formation of secondary air pollutants during the COVID-19 lockdown in Xi'an. The atmospheric Ox was roughly represented by the sum of the concentrations of NO2 and O3 [66], with a higher Ox in the lockdown period (115.3 μg m −3 ) than before the lockdown (90.7 μg m −3 ), and Ox was maintained at relatively high levels after the lockdown (116.4 μg m −3 ). The lower O3 concentration after the lockdown was due to the more favorable weather conditions and the weakening of secondary reactions, which can also explain the variations in PM2.5 in this study. PM 2.5 was the only one pollutant in this study that exceeded the national ambient air quality standard in China [64], especially in the pre-and during lockdown periods. Compared with PM 2.5 in pre-lockdown, the reduction of PM 2.5 in the post-lockdown period was the most significant (59%) among all air pollutants in this study, and statistical difference was observed. Moreover, the proportion of PM 2.5 in PM 10 in pre-, during, and post-lockdown periods of COVID-19 in Xi'an was 85%, 95%, and 64%, respectively ( Figure 3d). The proportion of PM 2.5 in PM 10 was the highest in the during lockdown period, increasing by approximately 10% and 30% respectively from the pre-and postlockdown periods. We compared the meteorological factors among the pre-, during, and post-lockdown periods of COVID-19 to demonstrate the drastic changes. The PWD in the three periods was northeast. The weather conditions before and during the lockdown were almost the same, whereas the wind speed increased and the RH decreased significantly after the lockdown. Therefore, the significant reduction in PM 2.5 /PM 10 during post-lockdown was owing to the increase in coarse dust mainly from the earth's crust resulting from the higher wind speed and temperature. Moreover, the 10% higher proportion of PM 2.5 in PM 10 during lockdown than in pre-lockdown was mainly due to the changes in particle emissions, especially enhanced emission from anthropogenic sources and the secondary formation of PM 2.5 (discussion below). The primary emission should be reduced during the lockdown (Figure 1; restriction on travel and temporary suspension of industries, factories, and construction sites); thus, the elevated PM 2.5 proportion may attribute to an enhanced secondary reaction during the lockdown, that is, increased oxidation in the atmosphere. This is more evident in an explanation of O 3 variations below.
Contrary to the trends of other air pollutants mentioned above, the concentration of O 3 _8h was the highest in the lockdown period (88.0 µg m −3 ), which was 2.6 times that in the pre-lockdown period, and the difference was significant. The increase in the O 3 concentration during the lockdown can be explained as follows. As mentioned earlier, NO x emissions are closely related to motor vehicle emissions enhanced by transportation activities and human travel. However, there are various sources of volatile organic compounds (VOCs), and their emissions varied during this event. For example, one of the main emission sources of VOCs is the evaporation of industrial solvents; the use of industrial solvents did not decrease as much as transportation did in the lockdown period. Therefore, the reduction in NO x was more significant than that in VOCs. This leads to a weakening of the titration of NO x , which reduces the depletion of ozone [65]; therefore, the concentration of O 3 in the atmosphere increased during the lockdown period. Simultaneously, the O 3 production and oxidation capacity (O x ) during the day increased in this study, which promoted the concentration of OH radicals during the day and NO 3 radicals at night. As a result, the increased oxidation capacity of the atmosphere promoted the formation of secondary air pollutants during the COVID-19 lockdown in Xi'an. The atmospheric O x was roughly represented by the sum of the concentrations of NO 2 and O 3 [66], with a higher O x in the lockdown period (115.3 µg m −3 ) than before the lockdown (90.7 µg m −3 ), and O x was maintained at relatively high levels after the lockdown (116.4 µg m −3 ). The lower O 3 concentration after the lockdown was due to the more favorable weather conditions and the weakening of secondary reactions, which can also explain the variations in PM 2.5 in this study.

PM 2.5 from Offline Filter Samples
The PM 2.5 mass concentrations from the offline filter samples in pre-lockdown, during lockdown, and post-lockdown periods of COVID-19 in Xi'an in 2020 are shown in Figure 4. The PM 2.5 mass concentration in offline filters obtained using the gravimetric measurement method (x) was first compared with the online PM 2.5 data obtained through automatic monitoring by using the β ray method (y) in Section 3.1. A close correlation was found between them, with the regression equation y = 1.06x−28.6, and a correlation coefficient (R 2 ) of 0.773. The pattern of change in PM 2.5 in periods of pre-lockdown, during lockdown, and post-lockdown of COVID-19 was consistent with the variation trend mentioned in Section 3.1, showing a downward trend gradually. As shown in Figure 3, the online concentrations of PM 2.5 in the three periods were 132.8, 107.7, and 54.4, respectively, which represent lower values than those in offline PM 2.5 filter samples. In comparison, PM 2.5 in pre-lockdown was 1.2 and 1.6 times those during lockdown and post-lockdown, respectively, according to the mass weighting (Figure 4), which may be due to the relatively high wind speed (4.5 ± 3.0 m s −1 ) and low relative humidity (50 ± 5%); after lockdown provided favorable meteorological conditions for the diffusion of air pollutants compared with before and during the lockdown. Similar continuous decrease in PM 2.5 concentration in the post-lockdown period were observed in Wuhan, China [67] and Mumbai, India [68]. It is inferred that the lockdown policies have a relatively long-term and lasting effect on reducing the concentration of air pollutants. Table 2 summarizes the average concentrations (mean ± standard deviation) and percentages of TC, OC, and EC in PM 2.5 . TC accounted for 13.3% ± 2.7%, 14.4% ± 4.2%, and 11.5% ± 3.8% of PM 2.5 mass in pre-lockdown, during lockdown, and post-lockdown periods of COVID-19, respectively. The proportion of EC in PM 2.5 remained almost unchanged (2.2-2.7%) during the different research intervals, whereas the proportion of OC in PM 2.5 reached the maximum (11.9%) during the lockdown, and was 1.1 and 1.3 times of those before and after the COVID-19 lockdown in 2020. As mentioned earlier, the weather conditions in Xi'an were basically stable in pre-and during lockdown periods. The reduction in the direct primary emission of PM 2.5 sources in the during lockdown period did not lead to a decrease in the concentration and proportion of OC. The OC generated by the secondary conversion during lockdown was the main reason for the increase in OC in this case, which can be also proven by the ratios of OC and EC. The OC/EC ratio can be used to determine the characteristics of carbonaceous aerosols' emission and transformation; an OC/EC ratio exceeding 2.0 suggests the presence of secondary organic carbon (SOC) [69,70]. All OC/EC values were higher than 2.0 in this study, with the maximum  (Table 2), thus indicating elevated SOC (i.e., secondary reaction) during the lockdown. online concentrations of PM2.5 in the three periods were 132.8, 107.7, and 54.4, respectively, which represent lower values than those in offline PM2.5 filter samples. In comparison, PM2.5 in pre-lockdown was 1.2 and 1.6 times those during lockdown and post-lockdown, respectively, according to the mass weighting (Figure 4), which may be due to the relatively high wind speed (4.5 ± 3.0 m s −1 ) and low relative humidity (50 ± 5%); after lockdown provided favorable meteorological conditions for the diffusion of air pollutants compared with before and during the lockdown. Similar continuous decrease in PM2.5 concentration in the post-lockdown period were observed in Wuhan, China [67] and Mumbai, India [68]. It is inferred that the lockdown policies have a relatively long-term and lasting effect on reducing the concentration of air pollutants.  Table 2 summarizes the average concentrations (mean ± standard deviation) and percentages of TC, OC, and EC in PM2.5. TC accounted for 13.3% ± 2.7%, 14.4% ± 4.2%, and 11.5% ± 3.8% of PM2.5 mass in pre-lockdown, during lockdown, and post-lockdown periods of COVID-19, respectively. The proportion of EC in PM2.5 remained almost unchanged (2.2-2.7%) during the different research intervals, whereas the proportion of OC in PM2.5 reached the maximum (11.9%) during the lockdown, and was 1.1 and 1.3 times of those before and after the COVID-19 lockdown in 2020. As mentioned earlier, the weather conditions in Xi'an were basically stable in pre-and during lockdown periods. The reduction

WSIs in PM 2.5 Filter Samples
The average total concentrations of nine WSIs were 46.8 ± 19.4, 38.9 ± 19.0, and 21.0 ± 14.5 µg m −3 , accounting for 32.1% ± 8.3%, 31.3% ± 9.5%, and 20.6% ± 8.8% of PM 2.5 mass in the periods of pre-, during, and post-lockdown of COVID-19, respectively. SO 4 2− , NO 3 − , and NH 4 + were the most abundant ions, accounting for 90-94% of total measured ions and 20-30% of PM 2.5 mass concentration. The total ion concentrations were consistent with the change pattern of TC, OC, and EC in the three time intervals. However, the concentration variations of K + and Ca 2+ were not consistent with those of the other WSIs. As a good marker for biomass burning [71,72], K + exhibited the highest concentration during the lockdown, 1.8 and 2.9 times those in the pre-and post-lockdown periods, which is due to the fact that almost all people stayed at home during the lockdown, and going out was restricted, and then the consumption of household heating and cooking biomass fuels (e.g., corn stalks, wheat stalks, and branches) in rural areas around Xi'an increased [73][74][75]. Ca 2+ , an indicator of fugitive dust from the earth's crust and construction, displayed the lowest value during the lockdown [76], 0.8 and 0.4 times those in the pre-and post-lockdown periods, proving that the reduction of going out and construction activities during the pandemic lockdown had a greater impact on the concentration of Ca 2+ .
From the perspective of the percentage of individual to the total ion concentration, only the percentages of NO 3 − and SO 4 2− exhibited greater changes in the periods of pre-, during, and post-lockdown. The proportion of NO 3 − in total WSIs was the lowest during the lockdown, with the value of only 36.7%, which was 15.8% and 24.3% lower than that before and after the lockdown. The proportion of SO 4 2− was the highest during the lockdown, reaching 31.6%, which was 17.5% and 47.6% higher than that before and after the lockdown. NO 3 − /SO 4 2− ratio has been usually used as a relative measure of the importance of motor mobile sources versus stationary emission sources (such as emissions from industrial combustion and residential fuel combustion) in many studies [10,77].   Table 3 summarizes previous studies that examined the impact of COVID-19 lockdown measures on local air quality in various regions of the world. The results suggest the positive effects of lockdown policies on air pollutant levels around the world. Compared with the same period in the previous year or the period before the lockdown, almost all pollutants studied exhibited significant declines during the lockdown, except for O3.   Table 3 summarizes previous studies that examined the impact of COVID-19 lockdown measures on local air quality in various regions of the world. The results suggest the positive effects of lockdown policies on air pollutant levels around the world. Compared with the same period in the previous year or the period before the lockdown, almost all pollutants studied exhibited significant declines during the lockdown, except for O 3 .

Comparison of Air Quality during the COVID-19 Lockdown among Studies
To better understand the variation characteristics of air quality in this study, the during lockdown/pre-lockdown air pollutant levels and PM 2.5 chemical composition were compared between Xi'an and other cities/regions, as shown in Figure 6. During the pandemic period, studies have revealed that the domestic and international lockdown measures implemented had a positive impact on the levels of PM 2.5 , PM 10 , SO 2 , NO 2 , and CO, with the ratios of during lockdown/pre-lockdown less than 1.0. The concentrations of the abovementioned five air pollutants had been reduced to varying degrees (except for SO 2 in Suzhou, China), whereas O 3 concentration showed an upward trend in all cities (Figure 6a,b), consistent with the results in the current study. compared between Xi'an and other cities/regions, as shown in Figure 6. During the pandemic period, studies have revealed that the domestic and international lockdown measures implemented had a positive impact on the levels of PM2.5, PM10, SO2, NO2, and CO, with the ratios of during lockdown/pre-lockdown less than 1.0. The concentrations of the abovementioned five air pollutants had been reduced to varying degrees (except for SO2 in Suzhou, China), whereas O3 concentration showed an upward trend in all cities (Figure 4a,b), consistent with the results in the current study. The reduction rate of PM 2.5 concentration ranged from 9.4% (four megacities in China) to 45% (Victoria, Mexico), which may be related to the specific local lockdown policies and meteorological factors. The maximum drop rates in total average concentrations of PM 10 , NO 2 , and CO were observed in Delhi and Mumbai (50%), Suzhou (64.5%), and Santiago (54%), indicating the notable environmental effect of lockdown measures. The minimum declines in PM 10 , NO 2 , and CO occurred in Xi'an (27%) in this study, Santiago (13%), and Wuhan (6%), respectively. In comparison, the SO 2 concentration exhibited a limited decrease (Figure 6a,b). Emissions from stationary sources, such as coal-fired power plants, were not considerably reduced compared with emissions from mobile sources such as traffic transportation [41,83]. For example, in Xi'an, sufficient electricity and heat were still provided as usual by the thermal power plants to ensure normal supply during the lockdown [91]. The increase in O 3 demonstrated a huge discrepancy in each city; the O 3 concentration showed a slight increase in Shanghai (5.7%), Tianjin (3.0%), Delhi and Mumbai (2.0%), and 380 cities across the globe (5.4%), but increased more than 50% in Wuhan (58%), Santiago (63%), Suzhou (104.7%), and Xi'an (160%). Specifically, the maximum O 3 increase was noticed in Xi'an, which is attributed not only to the emission sources but also directly to the relative high air temperature and low wind speed during the COVID-19 lockdown in 2020 ( Table 1). The ozone pollution in Xi'an must be further explored in a future study. Figure 6c illustrates an irregular discrepancy of the changes in the concentrations of chemical components in PM 2.5 among Chinese cities. The most obvious variation of WSIs in Xi'an was the significant increase in K + . The increased biomass combustion in the lockdown period in suburban regions and rural areas of Xi'an may be attributable to this phenomenon. Unlike in other cities, the proportions of Cl − , Na + , and Mg 2+ in PM 2.5 in the lockdown period in Xi'an remained almost unchanged from the pre-lockdown period. However, NO 3 − , SO 4 2− , NO 3 − /SO 4 2− , and NH 4 + in Tianjin, China, exhibited the opposite trend of an increase compared with other cities. This may be attributed to the weather conditions, local emissions, and lockdown policies. The highest declines in NO 3 − /SO 4 2− observed in Xi'an may be attributed to the following reasons: (1) as a well-known tourist attraction in China, suspension of tourism and strict traffic control in Xi'an during the pandemic (Figure 1) resulted in a substantial reduction in the flow rate of travelers in Xi'an; (2) except for government designation and pandemic prevention and control needs, the operation of interprovincial and municipal long-distance passenger transport lines and tour chartered buses into and out of Xi'an were suspended, and (3) cruising taxis and online car-hailing operations across provinces and cities were suspended [50]. These measures implemented in Xi'an had effectively reduced emissions from motor vehicle sources; relatively, the restrictions on welfare-related civilian industries, such as thermal power plants, were limited, which explains the maximum NO 3 − /SO 4 2− reduction in Xi'an in this study. Regarding the carbon components, both OC and EC emissions were reduced in all the four cities, with the largest OC reduction occurring in Tianjin, and the highest EC reduction in Wuhan. The elevated OC/EC during the lockdown period was observed in both Wuhan (47%) and Xi'an (26%), indicating a distinct increase in the secondary formation of organic compounds to PM 2.5 .

Conclusions
In this study, the online data of AQI, six national controlled air pollutants, and daily PM 2.5 and its bounded chemicals on the filter samples from 1 January to 7 March 2020, were used to investigate the changes in air quality in response to the control measures for the COVID-19 lockdown (pre-lockdown, during lockdown, and post-lockdown) in Xi'an, China. In this study, we found that restricting nonessential human activities could reduce several gaseous pollutants, especially NO 2 , and the specific components (e.g., Ca 2+ ) in PM 2.5 related to the particular anthropogenic sources. The lockdown policies in this study also led to an increase in primary emissions from household heating sources and an increase in secondary formation reactions in the atmosphere. Moreover, air pollution was closely influenced by meteorological factors and atmospheric oxidation. In this case, although the management and control of traffic and some non-livelihood industries improved air quality to a certain extent, the reduction of air pollutants was not significant. Therefore, readjusting the industrial and energy structure is necessary for the fundamental improvement of air quality in Xi'an.