Effect of COVID-19 Response Policy on Air Quality: A Study in South China Context

: Mass suspension of anthropogenic activities is extremely rare, the quarantine due to the coronavirus disease 2019 (COVID-19) represents a natural experiment to investigate the impact of anthropogenic activities on air quality. The mitigation of air pollution during the COVID-19 lockdown has been reported from a global perspective; however, the air pollution levels vary in different regions. This study initiated a novel synthesis of multiple-year satellite observations, national ground measurements towards SO 2 , NO 2 and O 3 and meteorological conditions to evaluate the impact of the COVID-19 lockdown in Beihai, a speciﬁc city in a less developed area in southwest China, to reveal the potential implications of control strategies for air pollution. The levels of the major air pollutants during the COVID-19 lockdown (LP) and during the same period of previous years (SP) were compared and a series of statistical tools were applied to analyze the sources of air pollution in Beihai. The results show that air pollutant levels decreased with substantial diversity during the LP. Satellite-retrieved NO 2 and SO 2 levels during the LP decreased by 5.26% and 22.06%, while NO 2 , SO 2 , PM 2.5 and PM 10 from ground measurements during the LP were 25.6%, 2.7%, 22.2% and 22.2% lower than during SP, respectively. Ground measured SO 2 concentrations during the LP were only 2.7% lower than during the SP, which may be attributed to uninterrupted essential industrial activities, such as power plants. Polar plots analysis shows that NO 2 concentrations were strongly associated with local emission sources, such as automobiles and local industry. Additionally, the much lower levels of NO 2 concentrations during the LP and the absence of an evening peak may highlight the signiﬁcant impact of the trafﬁc sector on NO 2 . The decrease in daily mean O 3 concentrations during the LP may be associated with the reduction in NO 2 concentrations. Indications in this study could be beneﬁcial for the formulation of atmospheric protection policies.


Introduction
The occurrence of severe environmental pollution events resulting from rapid urbanization and industrialization has strengthened public awareness of environmental protection and the pursuit of better environmental quality [1][2][3]. Urban air is a major focus, as it is the closest environmental element to human health and welfare [4]; however, air spatially diverse manner. O 3 levels were decreased in Southern China and Southwestern Europe and elevated in other regions of China and Northern Europe [38,39]. PM 2.5 showed a reduction during the COVID-19 period in the continental United States [40], cities in China [41,42], and in the 50 most contaminated capitals [43], but did not change in Ontario, Canada [44]. Specifically, although up to a 90% reduction in certain emissions during the city-lockdown period can be identified, extreme particulate matter levels simultaneously occurred in northern China [45]. We may conclude that the COVID-19 pandemic has undoubtedly altered emissions, but the linkages between emissions and air quality can be highly nonlinear, and complicated by atmospheric chemistry. How the pandemic has and will continue to impact air quality in different regions is worth investigating. Additionally, most studies focused on a regional, national, or cross-continental scale, and none have considered a specific city in a less developed area and synthesized satellite observations and national ground measurements to analyze the atmospheric influence of COVID- 19. Therefore, this study aims to: (1) reveal the long-term trend and spatial distribution of major air pollutants (SO 2 , NO 2 and O 3 ) over Beihai on a 15-year basis using satellite remote sensing; (2) compare the unique variation characteristics of air pollutants under the pandemic situation in 2020 and those in the same period of previous years to identify the impact of the COVID-19 lockdown in Beihai and (3) analyze the potential sources to obtain a comprehensive understanding of air pollution status, to reveal potential implications for more effective control strategies of air pollution.

Study Areas
Beihai is a medium-sized city with huge economic potential located in the south of the Guangxi Zhuang Autonomous Region of China and on the east coast of the most convenient seaport (Beibu Bay) in southwest China ( Figure 1). Beihai covers an area of 3337 km 2 between 108 • 50 45" to 109 • 47 28" E, 20 • 26 to 21 • 55 34" N with a population of 1.68 million, and a Gross Domestic Product (GDP) of 133 billion CNY (BSB, 2019). The urbanization rate of Beihai in 2018 was 58.6% [46]. There are significant differences in the topography of the city: the north is high, with hills; the south is low, with terraces and plains. The difference in topography may be an influencing factor for air pollution transportation. As a port and tourist city, vessel and vehicle exhaust may also play a role in air pollution. Beihai could be a typical example of a less developed city in South China. reported on a global scale [37]; however, air pollutants, such as O3 and PM2.5, responded in a spatially diverse manner. O3 levels were decreased in Southern China and Southwestern Europe and elevated in other regions of China and Northern Europe [38,39]. PM2.5 showed a reduction during the COVID-19 period in the continental United States [40], cities in China [41,42], and in the 50 most contaminated capitals [43], but did not change in Ontario, Canada [44]. Specifically, although up to a 90% reduction in certain emissions during the city-lockdown period can be identified, extreme particulate matter levels simultaneously occurred in northern China [45]. We may conclude that the COVID-19 pandemic has undoubtedly altered emissions, but the linkages between emissions and air quality can be highly nonlinear, and complicated by atmospheric chemistry. How the pandemic has and will continue to impact air quality in different regions is worth investigating. Additionally, most studies focused on a regional, national, or cross-continental scale, and none have considered a specific city in a less developed area and synthesized satellite observations and national ground measurements to analyze the atmospheric influence of COVID-19. Therefore, this study aims to: (1) reveal the long-term trend and spatial distribution of major air pollutants (SO2, NO2 and O3) over Beihai on a 15-year basis using satellite remote sensing; (2) compare the unique variation characteristics of air pollutants under the pandemic situation in 2020 and those in the same period of previous years to identify the impact of the COVID-19 lockdown in Beihai and (3) analyze the potential sources to obtain a comprehensive understanding of air pollution status, to reveal potential implications for more effective control strategies of air pollution.

Study Areas
Beihai is a medium-sized city with huge economic potential located in the south of the Guangxi Zhuang Autonomous Region of China and on the east coast of the most convenient seaport (Beibu Bay) in southwest China ( Figure 1). Beihai covers an area of 3337 km 2 between 108°50′45″ to 109°47′28″ E, 20°26′ to 21°55′34″ N with a population of 1.68 million, and a Gross Domestic Product (GDP) of 133 billion CNY (BSB, 2019). The urbanization rate of Beihai in 2018 was 58.6% [46]. There are significant differences in the topography of the city: the north is high, with hills; the south is low, with terraces and plains. The difference in topography may be an influencing factor for air pollution transportation. As a port and tourist city, vessel and vehicle exhaust may also play a role in air pollution. Beihai could be a typical example of a less developed city in South China.

OMI Data
OMI is a new generation of sensors aboard spacecraft on the Aura satellite with a spatial resolution of 13 km × 24 km, jointly developed by the National Aeronautics and Space Administration (NASA) and the Finnish Meteorological Society (FMI), whose major objectives are observations and studies of air quality and the climate change [47]. OMI covers the earth once a day and transits China during the hours of 13:40-13:50 and provides measurements of O 3 , NO 2 , SO 2 , BrO, HCHO as well as aerosols and clouds [48]. Tropospheric NO 2 and daily Level-3 SO 2 products of the boundary layer vertical column density (VCD) data were derived from OMI and employed in this study. The VCD is the amount of molecules for a specific atmospheric gas from the surface to the height indicated, such as the instrument height [49]. The spatial resolution of the NO 2 column concentration is 0.125 • × 0.125 • , while the daily Level-3 SO 2 products are gridded data with a 0.25 • × 0.25 • spatial resolution on uniform space-time, which is suitable for long-term trend analysis scales [50]. The data format in OMI inversion is HDF (Hierarchy Data Format). It is converted to txt after downloading. Fortran language programming was used to calculate the annual average value of gridding of air pollutants concentrations, and negative, outliers and zero values were removed, and the data within the scope of Beihai were extracted. The average values for 2005-2007, 2008-2010, 2011-2013, 2014-2016, 2017-2019, during the COVID-19 lockdown period (25 January 2020 to 25 February 2020) and during the same period for 2017-2019, and the change trend of the concentration of the pollutants was calculated. Thereafter, the spatial distribution map and the change trend graph of the average concentration of SO 2 , NO 2 and O 3 columns were prepared with ArcGIS software.

Ground Station Measurements and Meteorological Data
The Ministry of Environment and Ecology of China publishes real-time monitoring data of six major pollutants (PM 2.5 , PM 10 , O 3 , SO 2 , NO 2 and CO) in 74 cities since January 2013 [51]. In this study, the hourly PM 2.5 , PM 10 , O 3 , SO 2 , NO 2 and CO concentrations from 2016 to 2020 of four automatic ambient air quality monitoring stations were provided by the Beihai Ecological and Environment Bureau. The measuring stations at Beihai follow the regulations issued by the Ministry of Environment and Ecology [52,53], while quality assurance (QA) and quality control (QC) procedures follow the principles of the National Ambient Air Quality Urban Monitoring Network Management. Daily averages (24 h) and diurnal variation were calculated for the periods during the lockdown (25 January to 25 February in 2020) and before (averaged values of the corresponding period in the previous three years), to analyze the variations in the mean concentration (in µg/m 3 ) between both periods, and their relative change (in %). The names of the monitoring stations, acronyms, geographic locations, the measured species and the periods of data available are summarized in Table 1. Meteorological data including temperature, relative humidity, atmospheric pressure, wind speed and wind direction, were also measured at the four stations, and the hourly data were used in this study.  [54,55] was used to analyze the data in the statistical software R programming language version 3.6.2 [56].
Atmosphere 2022, 13, 842 5 of 22 Figure 2 depicts the spatial distributions of the averaged VCD SO 2 , NO 2 and O 3 on a three-year basis over the past fifteen years. The hotspots of VCD SO 2 are mainly centralized in areas near the southwest coast with a high population density and intensive anthropogenic activities (industrialization and urbanization). The SO 2 pollution over the area shows a decreasing trend since 2008. Between 2005 and 2007, the average VCD SO 2 in Beihai was higher than 0.32 DU (Dobson Units). However, the total areas where the VCD SO 2 was below 0.3 DU expanded to more than 2/3 of the total Beihai area by 2019. This decreasing trend is inseparable from a series of policies issued in recent years in China to control the emission of atmospheric pollutants. The major controlling strategy for the SO 2 emissions is the phasing out of inefficient coal-fired power generation units and steel plants, the installation of flue gas desulfurization systems in thermal power units, the change of fossil fuel and coal to natural gas, and the implementation of strict emissions standards for densely populated areas and industrial boilers [57,58]. Additionally, particulate matter removal through electrostatic precipitators contributes significantly to the simultaneous reduction in SO 2 emissions [59]. On the contrary, the three-year average VCD NO 2 increased slightly since 2005 and the NO 2 pollution area surrounding the industrial zone near the southeast coast, where the metal and chemical industry and coal-fired power generation are the major industries with high emissions of pollutants, was expanding and the pollution level was rising. The area where the average NO 2 column concentration was higher than 4 × 10 15 molecules cm −2 in 2005-2007 accounted for 54.3% of the total Beihai, which expanded to 77.1% by 2017-2019.

Spatiotemporal Variation from OMI Observations
With regard to the spatial distributions of O 3 during the last fifteen years, apparent high levels of O 3 concentrations can be observed in the southwest of Beihai. Tropospheric O 3 is generally formed in the atmosphere through photochemical pathways of nitrogen oxides (NO x = NO + NO 2 ) and volatile organic compounds (VOCs) in the presence of sunlight [60]. VOCs emitted from the surrounding chemical bases may also be transmitted to Beihai, influencing the process of photochemical reactions. Due to the complicated and non-linear relationship between O 3 and its precursors (NO 2 and VOCs) and the complex sources of VOCs in Beihai and surrounding areas including industrial production, vehicle exhaustion, solvent use, and straw burning [61,62], the distribution of severely polluted areas of NO 2  The COVID-19 lockdown measures came into effect in Beihai on 25 January 2020 to flatten the COVID-19 epidemic curve. The social isolation policy issued by the authorities has resulted in reduced anthropogenic activities, including a significant decline in traffic and the suspension of a number of industrial processes. To further investigate the air pollution characteristics of Beihai and to analyze the sources of air pollutants, we assessed the changes in air quality by comparing the air pollutant levels during the COVID-19 lockdown and (25 January 2020 to 25 February 2020) and the averages of those during the same period during the previous three years, as illustrated in Figure 3. During the COVID-19 lockdown, the decrease in the concentration of different air pollutants varies evidently. The averaged VCD NO 2 during the lockdown period (3.92 × 10 −15 molecules cm −2 ) decreased by 5.0% when compared with that pre-lockdown (4.12 × 10 −15 molecules cm −2 ). In the case of VCD SO 2 , the reduction was as high as 22.06% with 0.2222 DU during the lockdown and 0.2851 DU during the same period of the previous three years. In      It could be expected that reduced social activities due to the COVID-19 lockdown temporarily diminished the levels of VCD SO 2 and NO 2 . However, as also shown in Figure 3, areas with high levels of both VCD SO 2 and NO 2 were situated in the southeast of Beihai, in which the thermal power plant, several essential chemical materials and other industries were located and during the COVID-19 lockdown, these enterprises were completely or largely in operation. This may additionally emphasize the fact that reduced industry and traffic activities due to COVID-19 lockdown temporarily diminished the levels of VCD SO 2 and NO 2 . The levels of VCD O 3 saw a significant increment during the COVID-19 lockdown in Beihai. On the one hand, the low levels of nitrogen oxide (NO) around the boundary layer reduce the O 3 consumption (titration, NO + O 3 = NO 2 + O 2 ), thus leading to an increase in O 3 concentrations [63,64]; on the other hand, the possible increase of insolation and temperatures from January to February of lockdowns caused an increase in O 3 [65], as the first quarter of 2020 in Beihai was the second warmest on record. The stratosphere-troposphere exchange of O 3 may, additionally, play a role [66].

Long-Term Trends Analysis with Ground Measurements
The concentration data of the measured air pollutants at the four ground observation sites in Beihai were coded to smooth lines to analyze the long-term trends ( Figure 4). The smooth line is essentially established on the basis of the Generalized Additive Modelling with "mgcv" package [67]. A plot of monthly concentrations and bands along with the smooth lines in the corresponding color indicate the 95% confidence intervals of the fits.  Beihai may be associated with the dramatically reduced road and non-road transportation and the industrial suspension due to recommendations on reducing social contact. Emission data provided by the Beihai Ecological and Environment Bureau (BEEB) have shown that the NO x emissions from the major industrial sources from 25 January to 25 February were 651.0 tons in 2019; this declined to 387.32 tons (40.5% reduction) in 2020. In order to complete the "13th Five-Year Plan" pollution reduction task successfully, the government of the Guangxi Zhuang Autonomous Region has promoted the implementation of ultralow emission renovation projects for coal-fired power plants vigorously, using integrated technology to remove multiple air pollutants systematically and efficiently [68]. Moreover, most of the projects were completed by 2018. Therefore, power plants in Beihai may not be the main source of air pollutant emissions, but the steel industry could be a considerable contributor. NO x emissions from the petrochemical industry and thermal power plants remained approximately unchanged before and during the COVID-19 outbreak according to the emission data provided by BEEB. Sicard et al. (2020) also observed strong reductions in both NO and NO 2 during COVID-19 lockdowns in several European cities, such as Nice, Rome and Valencia [41], which is consistent with our results (NO concentrations decreased by 15.3% and 29.2% in NO x ), as shown in Figure S1. A relatively smaller reduction in SO 2 concentrations can be explained by the uninterrupted operation of essential industrial activities, such as power plants, steel and petrochemical enterprises; this has also been reflected in the SO 2 emission data, as 12.7% less SO 2 was emitted during the LP. Although a reduction in automobile emissions was expected, the low sulfur content of automobile gasoline (China V standard, maximum sulfur level of~10 ppm) implemented since 2017 can weaken the contribution of traffic reduction to decrease in SO 2 concentrations. In summary, the reduction in human activity reduced primary air pollutant emissions and led to an improvement in air quality, as also observed by other studies [69][70][71][72]. However, based on the comparison of measurements of air pollutants and emission data of major industries in Beihai, different emission sources may contribute to the reduction in SO 2 and NO 2 . The traffic sector in Beihai may have a limited contribution to the concentration of SO 2 .
The daily mean values of CO concentrations calculated over 5 years never exceeded the Chinese daily Class I Standard of 4 mg/m 3 and varied from 0.31 mg/m 3 (IA) to 2.05 mg/m 3 (BEEB). Decreasing trends of PM2.5, PM10 and CO were observed at all the sites, while the changing trends of SO2 were essentially constant. Only IA witnessed fluctuating interannual concentrations of SO2. In summary, the tendencies of all the pollutants measured at ground stations were consistent with the satellite remote sensing observations from the past five years.   Table 1).   Table 1).

of 22
Atmosphere 2022, 13, 842 10 of 23 The inconsistency of O 3 levels between OMI observations and ground measurements may be due to the difference in observation frequency of the two monitoring techniques and the vertical ozone distribution affected by a number of factors [73]. The local scanning time of the OMI instruments is approximately 13:45 [74] when the O 3 concentrations maintain a high level. However, the ground measured O 3 concentrations were processed as daily averages. Generally, the O 3 formation depends on the VOC-NO x interaction [75]. Previous studies have shown that O 3 levels experienced a significant increase during the COVID-19 lockdown in the Yangtze River Delta region [71], Milan [76], Wuhan and a few European cities [41]. In our case, the decrease in daily O 3 mean concentrations in Beihai might be associated with reduced ground NO 2 concentrations, according to the conclusion demonstrated by Wang et al. (2019) that NO x sensitivity dominated O 3 production in the southwest of China [77]. NO serves as the limiting reagent in reactions with RO 2 and HO 2 and O 3 production can be restricted by limiting NO x emissions. During the COVID-19 lockdown, a decrease in ozone precursor emissions, such as VOCs and CO (see Figure  S1) reduction, may also have contributed to the ground surface O 3 decrease [78,79]. The difference in O 3 variations between our observation and other cities could be attributed to the differences in industrial structure and O 3 formation mechanisms, e.g., NOx-limited in Beihai [77], while VOCs-sensitive in Wuhan [80].  43.6 µg/m 3 in 2020) during the studied period was recorded. PM 10 are emitted principally from traffic, industrial processes, construction work and dust entrainment [81,82], whose decrease can presumably reflect the decrease in traffic density and industrial activities (PM 2.5 ), and construction work and dust entrainment (PM 2.5-10 ). The identical reduction rate of PM 2.5 and PM 10 may suggest a synchronous reduction in both traffic and industrial activities, and construction work and dust entrainment. In additional PM emission data of major industries provided by BEEB, the PM emissions declined by 10.9% (76.85 tons during LP in 2020 and 86.26 tons during the same period in 2019). Figure 6 further investigated the diurnal variation of the concentration levels of the six air pollutants in Beihai during the LP and SP. The diurnal distribution of NO 2 during the LP showed different distribution compared to the SP, and at a much lower level. The NO 2 during the SP began to increase at 05:00 and peaked (14.92 µg/m 3 ) at 09:00, which is consistent with the LP, while the second peak during the SP commenced at 16:00 and reached the highest level of the day (18.75 µg/m 3 ) at around 19:00, but a second peak during the LP is not prominent. Generally, increased NO 2 concentrations in the atmosphere could be attributed to primary emissions from automobiles and other combustion sources, weak dispersion conditions and the formation of NO 2 through the reaction between the primary emitted NO and O 3 [18,83]. The increased NO 2 concentration during both LP and SP from 06:00 to 09:00 in Beihai could be attributed to elevated traffic emissions and insufficient photochemical consumption of NO 2 [18]. The decreasing trend of NO 2 concentrations between 09:00 and 15:00 can be caused by the increased photochemical reaction of O 3 production from NO 2 and VOCs accompanied by high temperatures, the increase in solar radiation, and the boundary layer height [84,85]. High dispersion and high dilution conditions resulting from the high temperature during the mid-day time may consequently enhance the thermal turbulence and thus lead to a decrease in NO 2 concentrations. After intensive photochemical consumption of NO 2 , enhanced anthropogenic activities, particularly automobile emissions may be responsible for the rise and peak of the NO 2 concentrations. During evening and nighttime until early morning, the boundary layer descends and resists the mixing of NO 2 emissions with the upper layer, thereby resulting in high levels of NO 2 [86]. The much lower levels of NO 2 concentrations during the LP and the absence of an evening peak may highlight the reduced impact of emissions from the traffic sector on the NO 2 concentrations during the COVID-19 lockdown in Beihai.
Atmosphere 2022, 13, 842 12 of 23 in high levels of NO2 [86]. The much lower levels of NO2 concentrations during the LP and the absence of an evening peak may highlight the reduced impact of emissions from the traffic sector on the NO2 concentrations during the COVID-19 lockdown in Beihai. It can be noticed from Figure 6 that the diurnal variation of O3 during both the LP and the SP is a typical signature of high concentrations during the daytime and low concentrations during the late night and early morning. The lowest O3 concentration during both the LP and SP appears around the morning hours (07:00-09:00), while they peaked at about 16:00, which is essentially opposite to the distribution of the ozone precursor, NO2. A similar diurnal variation pattern of O3 has been reported in numerous studies worldwide [87][88][89][90]. Solar radiation intensity increases in the daytime and favors the conditions to power the photochemical conversion to O3. The low level of O3 concentration during late night and morning is related to the consumption of O3 by deposition and or reaction with NO and NO2 and the absence of the photochemical reactions [91]. The diurnal patterns of O3 for the LP and SP show a similar pattern; however, the levels during lockdown were much lower, which may be attributed to the low levels of ozone precursors (NOx and VOCs emissions) resulting from limited traffic and various industrial activities. It can be noticed from Figure 6 that the diurnal variation of O 3 during both the LP and the SP is a typical signature of high concentrations during the daytime and low concentrations during the late night and early morning. The lowest O 3 concentration during both the LP and SP appears around the morning hours (07:00-09:00), while they peaked at about 16:00, which is essentially opposite to the distribution of the ozone precursor, NO 2 . A similar diurnal variation pattern of O 3 has been reported in numerous studies worldwide [87][88][89][90]. Solar radiation intensity increases in the daytime and favors the conditions to power the photochemical conversion to O 3 . The low level of O 3 concentration during late night and morning is related to the consumption of O 3 by deposition and or reaction with NO and NO 2 and the absence of the photochemical reactions [91]. The diurnal patterns of O 3 for the LP and SP show a similar pattern; however, the levels during lockdown were much lower, which may be attributed to the low levels of ozone precursors (NO x and VOCs emissions) resulting from limited traffic and various industrial activities.
The SO 2 concentration in Beihai during the LP and SP shows a single mode diurnal variation, and an evident lower level of SO 2 during the LP is observed, e.g., the peak SO 2 concentration during the SP was 16.7% higher than that during the SP. However, the SO 2 concentrations between 20:00 and 00:00 during the LP were slightly higher than those during the SP. This may not be a result of an increase in the SO 2 emissions, because the traffic and industrial activities saw an evident fall during the COVID-19 lockdown.
Meteorological conditions, such as the boundary layer height [6], might be responsible for the high level of SO 2 concentrations during the nighttime.
In the case of PM 2.5 and PM 10 , the diurnal variations of PM 2.5 and PM 10 during the LP and SP were highly identical, but the concentration levels of both PM 2.5 and PM 10 during the LP were evidently lower. The PM 2.5 concentrations during the daytime were perceptibly lower than the nighttime during the LP, which is the opposite of that during the SP. This may be because of the reduction in primary PM 2.5 emissions due to reduced vehicle and industrial activities and the accumulation effect during the nighttime [92]. Additionally, the reduction rate of coarse particles (PM 2.5-10 ) is approximately the same as PM 2.5 , which may suggest a proportional reduction in coarse particle emissions, such as construction [93].

Source Apportionment with Polar Plot Analysis
The potential sources of atmospheric pollutants at four ground observation sites are generally investigated preliminarily by bivariate polar plots analysis, which is a valuable statistical tool for mapping the pollutant concentrations as a continuous surface by wind speed and direction [55,94]. These polar plots provide a reference for the directional dependence of different sources and the variation of air pollutant concentrations with respect to wind speed and direction [67,95]. Polar plots were computed for four sites with hourly measured data from 1 January 2016 to 25 February 2020, as illustrated in Figure 7.
Polar plots at the BEEB station show the highest concentrations of both gaseous and particulate pollutants when the wind comes from the northwest. The prevailing air pollutant emission sources from the northwest are a fishing port zone integrating fishing boats, aquatic product processing, and fishery tourism, a cargo terminal, a major motor road, residual areas and an industrial park including rubber and plastic production plants, as well as an electronic company. Polar plots at the BP station showed very different variation characterization compared to those at the BEEB. SO 2 in the BP station exhibited high concentrations from all wind directions except for west and southwest but increased towards the northeast and northwest. A major motor road with a bus station and the Beihai Yintan tourist area with a sophisticated transportation system lies near the north of this measuring site, which may lead to the transport of SO 2 emissions toward the observatory site.
Evident high SO 2 concentrations in the IA station (12-13 µg/m 3 ) were also observed with increases from the north under a range of wind speeds (0 to 8 m/s). The location of IA is near numerous emission sources including the power plant and oil refinery facilities on the north side. Similarly, elevated SO 2 concentrations in the NR site were observed from the west and south due to the stack emissions from the industrial park (~4.7 km) in the near west, and the industrial area and main motor roads on the south side. Ozone precursors, such as NO x and VOCs produced from an industrial area in the north and the highway extending northeast from the site may be the reason for the higher O 3 concentration when the station is exposed to the northeast wind.
Bivariate polar plots for PM 2.5 and PM 10 are largely identical at the IA and NR stations, demonstrating similar prevailing sources and that the majority of PM 10 is mainly made up of PM 2.5 . PM 2.5 /PM 10 was 0.69 at the IA station and 0.80 at the NR station, indicating the dominant contribution of fine particles to PM episodes. The expressway in the northeast might have caused higher PM 2.5 and PM 10 concentrations near the NR station during the northeast wind. Vehicle emissions on the main road in the northeast and residential areas in the southeast may lead to higher PM 2.5 and PM 10 concentrations when the IA station is exposed to the northeast and southeast winds. The PM 10 concentration was highest under high wind speed (>8 m/s) at both the BEEB and BP stations, which are located in urban areas. This may indicate that high PM 10 concentrations at the BEEB and BP were dominantly transported from northwest and southwest, and northeast, respectively. generally investigated preliminarily by bivariate polar plots analysis, which is a valuable statistical tool for mapping the pollutant concentrations as a continuous surface by wind speed and direction [55,94]. These polar plots provide a reference for the directional dependence of different sources and the variation of air pollutant concentrations with respect to wind speed and direction [95,67]. Polar plots were computed for four sites with hourly measured data from 1 January 2016 to 25 February 2020, as illustrated in Figure 7.  Table 1).
Polar plots at the BEEB station show the highest concentrations of both gaseous and particulate pollutants when the wind comes from the northwest. The prevailing air pollutant emission sources from the northwest are a fishing port zone integrating fishing boats, aquatic product processing, and fishery tourism, a cargo terminal, a major motor road, residual areas and an industrial park including rubber and plastic production plants, as well as an electronic company. Polar plots at the BP station showed very different variation characterization compared to those at the BEEB. SO2 in the BP station exhibited high  Table 1).
Uria-Tellaetxe and Carslaw [96] pointed out that when non-buoyant ground-level sources, such as road traffic emissions and domestic heating are the major emissions sources, high NO 2 concentrations are likely to occur under stable atmospheric conditions. The highest NO 2 concentrations occur under very low wind speed conditions and show little directional dependence at all four stations, which indicates non-buoyant ground-level sources are the main contributors to NO 2 emissions rather than long transportation, such as local traffic. Additionally, the evident distinction between the two main peak concentration levels of NO 2 during the LP and SP may also emphasize the highlighted impact of traffic on the NO 2 levels. With regard to O 3 , the polar plots of O 3 show an approximately opposite pattern to that of NO x ( Figure S2 and NO 2 at all sites, due to the titration by NO x [97]. The O 3 levels are lower in stable atmospheric conditions (at very low wind speed) and show the highest concentration when the wind is blowing from a certain direction, which is an opposite trend to that of NO x and NO 2 .
To further analyze the sources of atmospheric pollutants, we generated polar plots for the LP and SP, as shown in Figure 8. Evident discrepancies in the concentration distribution depending on wind speed and wind direction during the LP and SP are observed. The highest NO 2 concentrations occur under very low wind speed at all four sites during the lockdown, while higher NO 2 levels occurred from certain wind directions during the SP, such as the higher concentrations when the wind comes from the northwest at the BEEB and the southeast at the BP, which is consistent with the long-term polar plot characteristics described above. This may highlight a more significant impact of local sources (traffic particularly) on NO 2 concentrations than transportation during the LP. SO 2 exhibited high concentrations from the same wind direction during the LP and SP. The closure of nonessential businesses, such as tourist sites, motorized transport, residential emissions and bus traffic might lead to the high SO 2 levels at the BP, and stack emissions from industrial plants and main traffic roads might be responsible for high SO 2 levels at the IA and NR. In short, high concentrations of SO 2 were observed at lower wind speed during the LP, suggesting the contribution of local emission sources to SO 2 during the LP. Similarly, high concentrations of PM 2.5 and PM 10 were observed at lower wind speeds during the LP, and under relatively high wind speeds during the SP. Low wind speeds are beneficial for the PM 2.5 accumulation from traffic around monitoring sites [98]. PM 10 concentrations were low when high winds occur, which was probably due to resuspension [98]. Moreover, the wind direction leading to the higher concentration was also different between the LP and SP, e.g., the highest PM 10 concentrations were observed when the wind was blowing from the northeast under a speed range of 8 to 10 m/s during the SP at BP, while PM 10 concentrations were highest under wind speeds of 2-4 m/s from the northwest during the LP, all of which indicate that the sources of particulate matter were different during the LP compared with those during the SP and there were probably more local emissions. Bivariate polar plots for O 3 also showed the different sources of O 3 between the LP and SP. At the IA station, high O 3 concentrations were observed when the wind came from the west during the SP, while high O 3 concentrations were observed when the wind came from the west during the LP; the O 3 peak values during the LP were much higher. The west of the IA station is a major traffic road in Beihai, which may indicate that the decrease in NO x caused by traffic limitation might be the cause of the decrease in O 3 concentrations.

Conclusions
Mass suspension of human activities for a long period is extremely rare, the COVID-19 lockdown serves as a natural experiment to investigate the impact of anthropogenic activities on air quality. This study initiated a novel synthesis of multiple-year satellite

Conclusions
Mass suspension of human activities for a long period is extremely rare, the COVID-19 lockdown serves as a natural experiment to investigate the impact of anthropogenic activities on air quality. This study initiated a novel synthesis of multiple-year satellite observations; national-station ground measurements of SO 2 , NO 2 and O 3, and meteorological conditions to address the impact of the COVID-19 lockdown in Beihai, a specific city in a less developed area in southwest China, to reveal the potential implications for the control strategies of air pollution. The changing trends in pollution levels (SO 2 , NO 2 and O 3 ) and spatial distribution on a 15-year basis over Beihai were characterized, while satellite retrieved and ground measurements of major air pollutants during the lockdown period and those during the same period of previous years were compared to identify the impact of the COVID-19 lockdown in Beihai; a series of statistical tools were applied to analyze the potential sources and to obtain a comprehensive understanding of air pollution status, to reveal potential implications for more effective control strategies of air pollution.
Results show that satellite observed SO 2 pollution saw a decrease since 2005, while NO 2 and O 3 showed an increasing trend during the last 15 years, which is consistent with the trend of ground measurements. A comparison between the NO 2 and SO 2 levels during the COVID-19 lockdown period and those during the same period over 2017-2019 shows that VCD NO 2 decreased by 5.26%, while VCD SO 2 saw a reduction of 22.06%. With regard to the ground measurements of air pollutants, the daily mean concentrations of NO 2 SO 2 , PM 2.5 and PM 10 during the LP were 25.6%, 2.7%, 22.2% and 22.2%, lower, respectively, than those during the SP. Averaged VCD O 3 during the LP increased by 4.26%; however, ground measured daily O 3 mean concentrations saw a decrease of 13.7%. The discrepancies in O 3 concentrations between satellite observations. Ground measured SO 2 concentrations during the LP were only 2.7% lower than those during the SP, which may be attributed to uninterrupted essential industrial activities, such as power plants. The traffic sector in Beihai may have a limited contribution to the concentration of SO 2 . The highest NO 2 concentrations occurred under very low wind speed and showed little directional dependence at all four stations, indicating non-buoyant ground-level sources could be the main contributor to NO 2 concentration. Additionally, the much lower levels of NO 2 during the LP and the absence of an evening peak may highlight the significant impact of the traffic sector on the NO 2 concentration. The decrease in the daily O 3 mean concentration during the LP might be associated with the reduction in NO 2 concentrations, as NO x sensitivity is found to be dominated in O 3 formation in the northwest part of China, which is different from other cities, e.g., Shanghai. Hence, the reduction of O 3 in Beihai could be coordinated with NO x reduction in terms of traffic controls. This study assessed the impact of the COVID-19 lockdown on air quality in Beihai and the implications of this study could be beneficial to the formulation of atmospheric protection policies. Institutional Review Board Statement: Not applicable.

Informed Consent Statement: Not applicable.
Data Availability Statement: All data generated or analyzed during this study are included in this published article and its supplementary information files.