Impact of the COVID-19 on the vertical distributions of major pollutants from a tower in the Pearl River Delta.

The outbreak of the 2019 novel coronavirus (COVID-19) had a large impact on human health and socio-economics worldwide. The lockdown implemented in China beginning from January 23, 2020 led to sharp reductions in human activities and associated emissions. The declines in primary pollution provided a unique opportunity to examine the relationship between anthropogenic emissions and air quality. This study reports on air pollutant and meteorological measurements at different heights from a tall tower in the Pearl River Delta. These measurements were used to investigate the vertical scale response of pollutants to understand reductions in human activities. Compared to that in the pre-lockdown period (from December 16, 2019), the concentrations of surface layer nitric oxide (NOx), fine particulate matter (PM2.5), and daily maximum 8 h average ozone (MDA8O3) declined significantly during the lockdown by 76.8%, 49.4%, and 18.6%, respectively. Although the vertical profiles of NOx and O3 changed during the lockdown period, those of PM2.5 remained the same. During the lockdown period, there were statistically significant correlations between PM2.5 and O3 but not between PM2.5 and NOx at four heights, indicating that the main composition of PM2.5 have dramatically changed, during which the impact of NOx on PM2.5 became insignificant. Additionally, O3 concentrations were also insensitive to NOx concentrations during the lockdown, implying that O3 levels were more of a representative of regional background level. In this case, local photochemical formation is no longer a significant ozone source. This evidence suggests that it is possible to mitigation of PM2.5 and O3 levels simultaneously by significant reductions in anthropogenic emissions.

Introduction

The novel coronavirus 2019 (COVID-19) pandemic has completely changed the world and caused the considerable loss of life around the world. At present, over 200 countries and regions have been affected by the pandemic, and the number of infections and deaths from severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) and its variants continue to rise (Wang et al., 2020). Many countries have opted to implement lockdowns to curb the spread of the pandemic; this has meant reducing gatherings and instigating social distancing among individuals. Generally, these measures have reduced human activity, and either decreased or completely halted manufacturing work and the movement of people. Although lockdowns have had devastating socio-economic impacts, recent studies have demonstrated that they have been beneficial for the natural environment (Chakraborty and Maity, 2020).

The reduction in human activities due to the pandemic has greatly decreased primary pollutant emissions, with significant impacts on regional air quality (Xing et al., 2020; Salma et al., 2020; Wang et al., 2021; Kim et al., 2021) and climate (Gettelman et al., 2021), despite differences between regions. In South East Asia, the lockdown led to a notable decrease in aerosol optical depth over the region and in pollution outflow over oceanic areas. A significant decrease (27%–30%) in tropospheric nitrogen dioxide (NO2) was observed over territories unaffected by seasonal biomass burning (Kanniah et al., 2020). Srivastava (2021) noted that the aerosol optical depth had reduced by up to 50% over the Indo-Gangetic Plain during lockdown. In Italy, urban road traffic decreased by 48%–60% on average during the implemented lockdowns, greatly decreasing concentrations of NO2 and particulate matter with an aerodynamic diameter <10 μm (PM10) and <2.5 μm (PM2.5) (Gualtieri et al., 2020). Rodríguez-Urrego and Rodríguez-Urrego (2020) found that the average PM2.5 concentration of the 50 most polluted capital cities in the world had decreased by 12% on average. The analysis of the emissions data of 28 cities in the United States of America (USA) during its first round of lockdowns (March 15, 2020 to April 25, 2020) showed that two of three cities experienced substantial reductions in NO2 and carbon monoxide (CO) concentrations (with decreases up to 49% and 37%, respectively), compared with the 2017–2019 historical baseline and pre-lockdown levels. The extent of decreases in NO2 and CO concentrations were in proportion to local population density; however, the PM2.5 and PM10 concentrations only decreased significantly in north-eastern USA, California, and Nevada, which also experienced the largest decreases in NO2 concentrations (Rodríguez-Urrego and Rodríguez-Urrego, 2020).

China was the first country to report SARS-CoV-2 infections to the World Health Organization. The atmosphere of China was also significantly affected by lockdown measures during the pandemic; some studies demonstrated considerable reductions in atmospheric NO2 concentrations (Zhang et al., 2020). These reductions first occurred in Wuhan prior to occurring throughout the remainder of China (Wang and Su, 2020). The Pearl River Delta (PRD) is one of the most important economic zones in China and is also a region experiencing the most rapid urbanisation rates in the world; as such, the intensity of human activity in the PRD is amongst the highest worldwide (Li et al., 2021). The PRD was once severely affected by air pollution, manifesting as increasingly frequent hazy weather and rising particulate matter (PM) concentrations. Air quality over the PRD has improved significantly over the past decade because of optimised industrial structures and the implementation of increasingly stringent pollution control measures (Zhang et al., 2018). Despite these measures, as the PRD contains immense transportation networks and a dense distribution of factories, it has been difficult to completely eradicate pollutant emissions. Therefore, the nitric oxide (NOx), PM2.5, and ozone (O3) concentrations in the PRD often spike because of unfavourable weather conditions (Li et al., 2020).

Since the COVID-19 pandemic, numerous studies have used this unique opportunity to gain important insights into the mechanisms underpinning air pollution. However, most studies have been based on ground-level data or space-based measurements of atmospheric column concentrations. By contrast, there are no reports on the vertical distribution of air pollutants during the COVID-19 pandemic. This vertical distribution provides a crucial part of understanding how air pollution events are occurred; meteorological towers are one of the most useful platforms to investigate the vertical distribution of near-surface pollutants. Unlike tethered balloons or UAV(unmanned aerial vehicle). Previously, numerous studies have been conducted using the meteorological tower in Beijing, China (e.g., Meng et al., 2008; Sun et al., 2010; Sun et al., 2013).

Shenzhen is a highly developed city experiencing intense human activities, facing air quality problems (Li et al., 2015; Yang et al., 2020). Li et al. (2020) analysed the vertical distribution of pollutants, including PM2.5, NOx, and O3, in the PRD during peak pollution season, based on air quality and meteorological datasets obtained at the Shenzhen Meteorological Gradient Tower (SZMGT) from December 2017. This analysis provided useful insights on the vertical structure of air pollutant distribution in the PRD. As the beginning of the COVID-19 pandemic coincided with peak pollution season in the PRD, the vertical distribution of pollutants recorded by the SZMGT during this period was invaluable in terms of demonstrating how reductions in human activity may affect the vertical distribution of pollutant concentrations. This study characterises changes in the vertical distribution of atmospheric pollutants induced by the extreme emissions reduction caused by human activities and explores possible mechanisms. The results provide scientific support for air pollution mitigation, particularly regarding the coordinated control of PM2.5 and O3 concentrations.

Data and methods

Observational data used in this study were sourced from a meteorological observation base on the east side of the Pearl River estuary, spanning from December 16, 2019 to February 15, 2020; namely, the Shiyan Meteorological Observation Base (hereafter Shiyan Base), managed by the Shenzhen National Climate Observatory (Fig. 1 a). The SZMGT is the most important observational platform in Shiyan base. The base lies approximately 10 km from the coastline, and is located is a woodland area surrounding a reservoir. As the reservoir is an important source of drinking water for residents of Shenzhen, the environment within 1 km around the SZMGT is protected and rarely disturbed by human activity, ensuring that the underlying surface will remain natural for a long time.

Fig. 1

Location of the Shiyan Meteorological Observation Base and Shenzhen meteorological gradient tower (SZMGT): (a) location of the Base; (b) aerial view of the meteorological tower; and (c) layout of the air quality and meteorological instrument setup on the tower.

The entire Shenzhen area is located within a subtropical monsoon climate zone. The dominant wind direction in summer is southerly, and airflow introduces clean air from the ocean to the base. In winter, the dominant wind direction becomes northerly, and airflow carries pollutants from the inland PRD to the base; this favours the accumulation and formation of air pollutants (Li et al., 2020). The peak of the COVID-19 pandemic largely occurred during winter, when meteorological conditions are generally favourable to pollutant formation and accumulation in Shenzhen.

The SZMGT is 365 m tall (Fig. 1b), contains 13 layers of meteorological observation platforms that begin from 10 to 350 m (Fig. 1c). Four of these layers (i.e., at 60–70, 110–120, 210–220, and 325–335 m, respectively), are atmospheric observation platforms (Fig. 1c). The distance from the SZMGT to the nearest built-up area is approximately 1 km. Approximately 800 m north-east of the Shiyan Base, there is a busy highway from which vehicular pollutant emissions may influence the measurements on the tower. There is also an airport located approximately 10 km west of the base which serves an estimated 356000 flights in a normal year. As such, the departure and arrival of airplanes at the airport may also potentially influence pollutant concentration measurements by the SZMGT (Li et al., 2020). An additional atmospheric environmental observation station has been installed at the bottom of the SZMGT. As this station is located on the ground, the height of its sampling port is lower than the surrounding forest top.

The meteorological data used in this study were collected at all 13 platform heights, as shown in Fig. 1c, whereas the environmental data were collected at heights of 110–120, 210–220, and 325–335 m. The data at 60–70 m height was not included in the analysis as equipment failure had occurred at this point during the pandemic. Data obtained by the atmospheric environmental observation station at the bottom of the SZMGT were also used.

The equipment used at the SZMGT to measure wind, temperature and humidity, and visibility included the Vaisala WMT700 Ultrasonic Wind Sensor, Vaisala HMP155 Humidity and Temperature Probe, and Vaisala PWD Present Weather Visibility Sensor, respectively. PM2.5, NOx, and O3 concentration data were collected using the Thermo Scientific™ 5030i Sharp Particulate Monitoring equipment, Thermo Scientific™ 42i Gas Analyser, and the Thermo Scientific™ 49i Gas Analyser; data from these instruments were downloaded once every 5 min. Arithmetic averaging of the data was carried out for all measured parameters, with the exception of wind direction; this averaging was undertaken to obtain hourly average data. The daily average data by arithmetic averaging were obtained using the hourly average data over 24 h. To determine the wind direction, representative values were obtained by calculating the maximum wind frequency for the hour and for the day. The instruments on the tower were maintained by professional service providers; meteorological observation instruments were routinely exchanged for calibration once every three months, and air quality observation instruments were maintained once every month.

Results and discussion

Changes of pollutant concentrations and meteorological parameters

Fig. 2 shows the daily mean concentrations of PM2.5, O3, and NOx observed at Shiyan Base in Shenzhen, alongside the daily mean relative humidity (RH), daily mean temperature, daily mean wind speed, and daily dominant wind direction from December 16, 2019 to February 15, 2020. Two key dates were marked with blue dotted lines on the PM2.5, O3, and NOx graphs: 15th–23rd January 2020. The first case of COVID-19 in Shenzhen was reported by local news outlets on January 15, 2020. Then, Guangdong Province (where Shenzhen is located), activated its top-level emergency response on January 23, 2020; all residents in Shenzhen were stay at home unless necessary. Therefore, the intensity of human activity in Shenzhen in terms of manufacturing and traffic, had begun to decrease on January 15, 2020. By January 23, 2020, Shenzhen had been virtually shut down due to the strengthening of activity restrictions. Aside from the most vital logistics chains, very little traffic remained on the streets. Although it has not been possible to quantitatively estimate the degree to which human activity decreased in Shenzhen during this period due to a lack of data, air traffic from the airport to the west of Shiyan Base may provide some indication of the reduction extent. News reports mentioned that the number of passengers at the airport had decreased by as much as 79.5% in February 2020. In addition, a nationally popular navigation service provider announced that during lockdown, the daily traffic flow was ∼14.1% compared to that in the pre-lockdown period. This means there were approximately 282000 vehicles on the road in all of Shenzhen (spanning 2000 km2) every day during the lockdown; by contrast, the average number of vehicles under pre-lockdown levels is usually ∼2 million vehicles. As the Spring Festival (Chinese New Year) is usually in February, this decrease in passenger volume and road traffic is a testament to the magnitude by which human activity decreased in this region.

Fig. 2

Daily variations in pollutant concentrations and related meteorological factors in the surface layer from 16th December 2019 to February 15, 2020: (a) PM2.5; (b) O3; and (c) NOx concentrations averaged using data from the four different levels of the SZMGT; (d) air temperature and relative humidity observed at the Shiyan Meteorological Observation Base; and (e) wind speed and direction observed by the automatic weather station at the Shiyan Meteorological Observation Base.

Fig. 2a and c shows that the daily mean concentrations of PM2.5 and NOx closely tracked the lockdown-mediated change in human activity. As there were no cases of COVID-19 in Shenzhen prior to January 15, 2020, the local government did not impose any restrictions between December 16, 2019 and15th January 2020; maintaining relative high pollutant concentrations. Following the first report of COVID-19 on January 15, 2020, many residents began to reduce the frequency of their outdoor activities due to their awareness of the pandemic. As these reductions in human activity were voluntary and not universal, pollutant concentrations had only gradually decreased. The wide spread implementation of high-level restrictions on January 23, 2020 led to drastic and sustained reductions in pollutant concentrations. In general, the daily mean concentrations of PM2.5 and NOx remained low after this date, whereby their range of variation had become significantly restricted. Although all three pollutant emissions had been reduced by the lockdown, the change in NOx concentrations was the most significant. This is because NOx is primarily derived from traffic emissions; as the decrease in human activity also prompted a decline in traffic emissions, the concentration of NOx in the atmosphere decreased instantaneously upon the cessation of vehicular traffic. Although the daily mean concentration of O3 did not change significantly after23rdJanuary 2020(Fig. 2b), the daily range of variation in its concentration (i.e., difference between the minimum and maximum concentration within a day), did decrease significantly after this date.

The large scale surface weather maps (figures not provided here) showed that during the period of study, Shenzhen was primarily controlled by uniform pressure or a weak high pressure ridge with sparse ground isobaric lines for most of the time; this is typical for this area in winter. During the study period, three cold fronts occurred on 26th-27th December 2019, January 12, 2020, and 27th–30th January 2020. When the cold fronts passed Shenzhen, there were more dense ground isobaric lines than those in normal conditions. Fig. 2d and e presents the variations in daily mean temperature, daily mean RH, daily mean wind speed, and daily dominant wind direction during the study period. As evident in Fig. 2d, the RH and temperature were strongly correlated, indicating that the dry air in the PRD predominantly comes from cold air masses. Whenever a cold front pass over the Shiyan Base, the daily mean temperature and RH will decrease. Fig. 2e shows that the daily mean ground wind speeds during the study period were usually below 2 m/s, even when cold fronts had passed. Although there was no dramatic change in the daily mean ground wind speed over the entire study period, the most polluted periods were related to weak wind speed <1.5 m/s. During the pre-lockdown period, the peak PM2.5 and NOx concentrations occurred on December 22, 2019, and the daily mean wind speed recorded that day was ∼0.9 m/s. During the lockdown, relatively high concentrations of PM2.5 and NOx occurred during from 11th–13th February 11–13, 2020, when the daily mean wind speed was between 1.0 m/sto1.3 m/s. Additionally, the daily dominant wind direction was northerly for approximately 75% of the time. Generally, the weather during the study period was relatively typical of winters in Shenzhen, indicating that no meteorological abnormalities had occurred during this time.

Table 1 illustrates the decrease in PM2.5 and NOx concentrations during the lockdown, where the decrease in the latter was much more drastic than that in the former. The change in O3 was more complex than PM2.5 or NOx; the daily average O3 concentration had slightly increased during the lockdown, consistent with the findings from other studies (Gualtieri et al., 2020). However, the mean daily maximum 8 h average O3 (MDA8O3) concentration had significantly decreased during the lockdown. As the daily O3 concentration and MDA8O3 exhibited different types of changes indicates the relatively different chemical environments related to O3 prior to and during lockdown.

Table 1

Comparison of pollutants and meteorological elements during the COVID-19 lockdown and prior to the lockdown.

Time period	Before 23rd Jan.	After 23rd Jan.	Relative changes
PM2.5 (μg/m3)	38.5	19.5	−49.4%
O3 (ppbv)	26.8	29.4	+9.7%
MDA8O3 (ppbv)	51.4	42.1	−18.6%
NOx (μg/m3)	50.9	11.8	−76.8%
Temperature (°C)	18.9	16.5	−12.7%
RH (%)	75.3	77.0	+2.3%
Wind speed (m/s)	1.6	1.8	+12.5%
Wind direction	NNE	NNE	–

* NNE: north-north-east. All pollutant concentrations in the table are averages for the whole surface layer recorded by the tower.

The changes in meteorological factors during the lockdown compared with those in the pre-lockdown period may largely be attributed to the intense cold air front that developed on 27th–30th January 2020; this event decreased the mean temperature during the lockdown period. By contrast, there was little change in the RH following the implementation of the lockdown. The meteorological factors most closely related to pollutant dispersal in the Shenzhen region were wind speed and direction. Although the average wind speed increased during the lockdown, it was still weak and rarely exceeded 2.0 m/s, limiting improvement in pollutant dispersion. The dominant wind direction during the pre-lockdown and lockdown periods was north-north-east, indicating that the winds in Shenzhen largely travel from the inland regions of China. In a normal year, these winds transport large amount of air pollution from the inland parts of the PRD, causing a spike in pollutant concentrations (Li et al., 2020). Fig. 3 shows the distribution of potential PM2.5 source areas near the Shiyan Base, to clearly illustrate the possible impact of meteorological conditions on air quality; this distribution was obtained using TrajStat software (Wang et al., 2009) to make a statistic on the 72 h backward trajectories affecting the Shiyan area. Fig. 3 demonstrates that the spatial distribution of potential PM2.5 source areas for the Shiyan Base were relatively similar during the pre-lockdown and lockdown periods. The contribution to PM2.5 from the north is dominant with high probability value. During the lockdown, the source areas in the two regions became narrower, and the probability with high-value areas were both closer to the measurement site, indicating that the impact of the emissions far away from the site has become weaker than the pre-lockdown period.

Fig. 3

Potential PM2.5 source area for the Shiyan Base:(a) pre-lockdown period; and (b) during lockdown period. Different colours indicate the probability that the airflow affecting the PM2.5 concentration of Shiyan Base passes through that area. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Fig. 4 provides the average vertical air temperature profiles recorded by the SZMGT before and after January 23, 2020, which shows that there is no significant difference in air temperature stratification before and after the outbreak of the pandemic. Thus, there were no significant differences in the meteorological conditions before and during the lockdown, with the exception of the strong cold front that had begun on January 27, 2020. The weathers during both periods may be generally considered poor meteorological conditions for air quality in Shenzhen. Although an intense cold spell occurred after 27th January2020 and the average wind speed had slightly increased, this change was not sufficient to cause dramatic decreases in average PM2.5 and NOx concentrations recorded during the entire lockdown period. Overall, the drastic changes in pollutant concentrations over the study period were unlikely to be caused by changes in meteorological factors.

Fig. 4

Average vertical air temperature profiles recorded by the Shenzhen Meteorological Gradient Tower before and after January 23, 2020.

Diurnal variations at different heights

Fig. 5 shows the diurnal variations in PM2.5, NOx, and O3 concentrations on the surface (2 m) and at three different heights (120, 220, and 335 m) of the SZMGT, before and during the lockdown. The PM2.5time series curves in Fig. 5a are characterised by two trends: (1) a bimodal distribution for the ground level (2 m) and 120 m curves; and (2) a unimodal distribution for the 220 and 335 m curves. The peaks of the bimodal curves occurred at 09:00Local Sidereal Time (LST) and 20:00 LST; these peaks roughly correspond to the morning and evening rush hours. The difference between the PM2.5 curves at 0/120 m and that of 220/335 m is likely to reflect the uplift process of the top mixing layer in this area. During winter nights and early mornings, the height of the top mixing layer is frequently between 110 and 220 m (Fig. 6 ); Therefore, the diurnal curves of pollutants in the upper and lower layers are relatively disparate. Following noon, the rise of the top mixing layer causes the pollutants curves of all layers to become similar. Although the PM2.5 concentrations at 2 and 120 m followed the same qualitative trend, the ground values were generally lower than those at 120 m. There was no direct pollution source within 100 m around the Shiyan Base, as it is located in a water source protection area and surrounded by a litchi forest. Pollutants affecting the base are typically sourced from the surrounding highways, airports, and built-up urban areas. The litchi forest surrounding the SZMGT is 10 m high on average, and thus, is able to obstruct pollutants beyond the base from reaching the sensors on the ground of the base; however, it cannot prevent these pollutants from reaching the measurement point located at 110–120 m height. The peaks of the unimodal 220 and 335 m curves occurred at 17:00–19:00 LST. Therefore, diurnal variations in PM2.5 concentration differed at lower and higher heights. This is consistent with Li et al. (2020) who implied that high and low-height PM2.5 may have different sources or formation mechanisms. High-height PM2.5 is formed predominantly by secondary chemical reactions, whereas low-height PM2.5 may be derived from multiple sources, including surface-level primary emissions and secondary chemical reactions.

Fig. 5

Diurnal variations in pollutants observed at different heights of the meteorological tower: PM2.5 concentrations (a) before lockdown; and (b) during lockdown. O3 concentrations (c) before lockdown; and (d) during lockdown. NOx concentrations (e) before lockdown; and (f) during lockdown.

Fig. 6

Mixing layer height diurnal variations for the Shiyan Base retrieved from LiDAR data in December 2017 using the method in Morille et al. (2007): (a) horizontal lines indicating average and vertical lines indicating range of variation; and (b) probability of mixing layer height between 110 and 220 m.

Fig. 5b presents the diurnal variations in PM2.5 concentrations at 2, 120, 220, and 335 m during the COVID-19 lockdown. It was obvious that PM2.5 concentrations had decreased significantly at all heights following January 23, 2020. The highest PM2.5 concentration still occur at 110–120 m and retained the bimodal structure of its pre-lockdown counterpart. The PM2.5 concentrations at 220 and 335 m were still unimodal, where the peak occurred at a similar time. The largest lockdown-mediated change in PM2.5 concentration occurred at 2 m, where the diurnal profile did not have a morning peak.

In terms of O3, it was evident that the diurnal variation in its concentration was unimodal and peaked at approximately 15:00–16:00 LST, when photochemical O3 formation is most active, before and during the lockdown (Fig. 5c and d, respectively). These diurnal variations were also qualitatively invariant with altitude; this means only the average concentration varied from one altitude to the other. However, the shape of the O3 diurnal profile did become significantly flatter during the lockdown, indicating a narrower range in the diurnal variations. The flattening of the peaks and valleys of the O3 curve implies that chemical reactions related to anthropogenic emissions (e.g., NOx, VOCs) generating O3 during the day and consuming O3 at night-time became more inactive during lockdown (Liu et al., 2021, Liu et al., 2021; Qi et al., 2021). Despite this, biological volatile organic compound (VOC) emissions may contribute to the generation of a very weak peak during the afternoon. Under these conditions, the O3 concentration appeared to be primarily determined by the background O3 concentration (Xu et al., 2020). Notably, a flatter O3 curve means a decrease in MDA8O3, implying that the mitigation of O3 and PM2.5 pollution may theoretically be realized concurrently.

The diurnal variations of NOx concentrations were bimodal before the lockdown (Fig. 5e); the 2 and 120 m curves showed a peak at 09:00 LST, coinciding with the morning rush hour. The second peak, beginning at 18:00 LST and continuing until 21:00 LST, was likely caused by the evening rush hour and night-time decreases in the altitude of the mixed layer. The first peak in the 220 and 335 m curves showed a 1 h lag from the first peak of the lower altitude curves; however, the second peak occurred at roughly the same time in both sets of curves. Although the mean NOx concentrations had decreased significantly during lockdown, its diurnal variations were still bimodal (Fig. 5f). The inter-altitude differences in NOx concentrations did become much lower during the lockdown and the timings of NOx peaks at each altitude were also closer to each other. During the lockdown, the first peak was delayed by 1 h, whereas the second peak occurred at 17:00–19:00 LST. The other significant way in which NOx concentrations had changed was that they were lower at 2 and 120 m than that at 220 and 335 m. The absolute decrease in the NOx concentration at the surface layer was much larger than that at 220 and 335 m, demonstrating that NOx in the surface layer was much more easily affected by emissions change.

To further analyse changes in pollutants during the lockdown, the concentration ratios of pollutants before and during the lockdown were calculated; this diurnal variation of the ratios is illustrated in Fig. 7 . The diurnal variation curves of different pollutants exhibited various characteristics. For NOx, curves at different heights were relatively consistent being roughly flat and maintained around 0.3; this indicates that there was an even and significant decrease in NOx concentration decreased in the boundary layer. The curves for PM2.5differed from NOx; ratio curves were relatively flat and maintained at ∼0.5 all day at heights >110 m, although there were relatively large ground level fluctuations. The ratio on the ground increased significantly between 7:00 and 18:00 LST as opposed to maintaining a relatively flat pattern. This indicates that the decrease in ground level PM2.5 concentrations (∼−30%) during the day over the lockdown period was not as drastic as that of the average data of the whole boundary layer (∼−50%). However, it was still difficult for PM2.5 generated on the ground to affect the air mass at >100 m height. The fluctuations in the ratio diurnal curves for O3 were clearer than those in the ratio diurnal curves for other two pollutants. The ratios were generally greater than 1.0 at night, clearly demonstrating less effective NOx titration at this time leading to relatively higher ozone concentrations at night than those observed in the pre-lockdown period. The ratios were <1.0 in the afternoon, which suggests that during lockdown, the O3 concentration decreased in the afternoon.

Fig. 7

Diurnal variations in during lockdown/pre-lockdown ratios of pollutants observed at different heights of the meteorological tower: (a) PM2.5, (b) O3 and (c) NOx.

Vertical distribution of pollutants

Fig. 8 presents changes in the vertical distribution of the three pollutants and Ox (= O3 + NO2), measured at 2, 120, 220, and 335 m of the SZMGT before and during the lockdown. In terms of the all-day averages (Fig. 8a–d), it was apparent that the PM2.5, NOx concentrations were lower across all altitudes during the lockdown; the differences passed the significance test at p < 0.01. By contrast, O3 concentrations did not decrease significantly, although their vertical gradations were less pronounced; as such, O3 concentrations became more uniform in the vertical direction during the lockdown. The daytime and night time average Ox concentrations (Fig. 8h–i) were generally lower during the lockdown period than during the pre-lockdown period, where the difference also passed the significance test at p < 0.01, indicating a weakened oxidation capacity for the whole boundary layer during the lockdown. The nitrate radical production rate (= kNO2+O3[NO2][O3]) was also examined during the nighttime as it is an indicator of nighttime oxidation reactions. This rate experienced a large decline, averaging ∼70%, suggesting a weakened NO3 oxidation capacity. The decrease in nighttime oxidation is mainly attributed to the dramatic reductions in NOx levels. Overall, the vertical observations showed that atmospheric oxidation processes, including photochemistry and nighttime chemistry, had largely been reduced due to the lockdown.

Fig. 8

Vertical distribution of three pollutants and Ox (i.e., NO2 + O3) observed at the Shenzhen Meteorological Gradient Tower:(a–d) whole-of-day data; (d–h) day-time data; and (i–l) night-time data.

Fig. 8a shows that the PM2.5 concentrations initially decreased with increasing altitude from 120 m, before increasing slightly with further altitude increases; this occurred before and during the lockdown period. The PM2.5 concentration was the highest at 120 m, whereas the concentration at 335 m was between those recorded at 120 and 220 m. This is an interesting result as it contradicts the expectation that PM2.5 concentrations should decrease monotonically with increasing altitude (Sun et al., 2010). This is understandable when considering the results of previous studies on possible PM2.5 sources at each altitude. At the lowest height (120 m), PM2.5 may have been sourced from ground photochemical reactions and primary pollution sources. At mid-and higher altitudes, PM2.5 is mainly formed by photochemical reactions; as such, the efficiency of PM2.5 generation at these heights may be affected by the oxidative potential of the atmosphere. Based on the observations from the SZMGT, the O3 concentration generally increased with height, and the Ox was higher at 335 m than at 220 m. As such, it is likely that the oxidation capacity of the atmosphere increased at the highest level, elevating the efficiency of PM2.5 formation at this altitude. Although the concentrations of VOCs were not measured from the SZMGT, measurements in Kaohsiung, Taiwan have shown that these compounds also increase with altitude, up to a peak of 300–400 m. As such, the VOCs provide an ample supply of reactants for photochemical reactions at high altitudes (Vo et al., 2018). Although there is a distance of ∼660 km between Kaohsiung and Shenzhen, both are located in the subtropical monsoon climate zone and with developed industry and transportation; this means the observations in Kaohsiung offer a point of reference for Shenzhen. At mid-altitudes (220 m), the PM2.5 concentration was not significantly affected by primary pollutant sources, and PM2.5-forming photochemical reactions were also less efficient at this point than higher altitudes. Consequently, the PM2.5 concentrations were lower at mid-altitudes than at higher altitudes.

O3 concentrations increased monotonically with altitude (Fig. 8b); this occurred even during the lockdown, where the average O3concentration remained high without showing any significant change. By contrast, vehicular emissions had plummeted to a very low level during the lockdown, as clearly evidenced in Fig. 8c. This figure shows that NOx concentrations had decreased considerably at near-ground altitudes, particularly at 120 m, where concentrations had decreased by >75% compared with pre-lockdown levels. Given the significant decrease in NOx concentrations (as much as −78.2% in this study), the concentrations of VOCs did not experience the same change as observed in NOx. In recent studies, Qi et al. (2021) reported that the decrease in VOCs in the PRD during the lockdown was much lower than that in NOx, whereas Liu et al., 2021, Liu et al., 2021 reported that formaldehyde (HCHO) abundance in the PRD area experienced an even slight increase during lockdown based on TROPOspheric Monitoring Instrument (TROPOMI) satellite observations. The different reduction degrees of NOx and VOCs during lockdown may largely shift ozone formation sensitivity to NOx and VOC.

Fig. 8e–h and 8i–l display the vertical distributions of pollutants and Ox during the day and night, respectively; day-time and night-time distributions of PM2.5, NOx, and Ox did not significantly differ. By contrast, the vertical distribution of O3 varied significantly between day and night. At all altitudes, day-time O3 concentrations were generally lower during the lockdown, whereas night-time concentrations were higher. This paradoxical trend may be attributed to the weakening of atmospheric chemical activity during the lockdown. As reduced human activity also decreased primary pollutant emissions, there was substantially lower availability of precursors for photochemical O3 generation, resulting in decreased day-time O3 concentrations. At night, dark chemical reactions that consume O3 also became less active during the lockdown, resulting in significantly higher night-time O3 concentrations at the near-surface atmosphere; these changes are consistent with the diurnal variations in O3 (Fig. 5d).

Regional transport is another important factor affecting the atmospheric of the PRD (Qu et al., 2021). Fig. 9 presents the distribution of potential source areas for different altitudes obtained through statistical analysis based on 72 h backward trajectory data during the study period. The distribution of potential source areas of the three pollutants were similar, where all sources were located in the northeast of the SZMGT. These areas span the east part of the Pearl River estuary, the urban belt along the southeast coast of China, and the mountains of Nanling in northern Guangdong. These potential source areas may supply industrial and traffic emissions from economically developed areas, alongside biological VOCs from mountainous areas to the PRD. As the distribution of potential source area is usually consistent at all altitudes, it was difficult to confirm that regional transportation had a clear impact on the vertical distribution of pollutants. In addition, during the lockdown period, the anthropogenic emission in potential sources were also very weak like Shenzhen due to the lockdown, and could have very insignificant impact through regional transport on the observation site.

Fig. 9

Potential PM2.5 source areas at different altitudes of the Shiyan Base:(a) 110 m; (b) 220 m; and (c) 320 m. Different colours indicate the probability that airflow affecting the PM2.5 concentration of the Shiyan Base passes through that area. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Correlations at different altitudes

Fig. 10 depicts scatterplots and fit lines of O3 and its relationship with PM2.5 at each level of the SZMGT, before and during the lockdown. Prior to the lockdown, there was a weak correlation between the O3 and PM2.5 concentrations, where none of the correlation coefficients (R) passed the significance test. During the lockdown, there was a significantly stronger correlation between PM2.5 and O3, where R values for the 0, 120, 220, and 335 m scatter plots were significant at p < 0.1. In the PRD region, VOCs significantly contributed to the formation of fine particles (Liu et al., 2008; Zheng et al., 2009), particularly secondary organic aerosols (Huang et al., 2006; Chang et al., 2019; Zhang et al., 2019). Although the advent of COVID-19 did trigger reductions in the concentrations of NOx, SO2, and other primary pollutants, VOC emissions may not have experienced such dramatic changes (Liu et al., 2021, Liu et al., 2021). The VOCs may act as important precursors for O3 and the secondary organic aerosols during the lockdown, strengthening the correlation between the PM2.5 and O3 concentrations.

Fig. 10

Scatter plots of PM2.5 and O3 concentrations at different heights of the meteorological tower before and during the lockdown: (a) and (b) ground level; (c) and (d) low level; (e) and (f) middle level; and (g) and (h) high level. The fit lines of plots were produced as per Cantrell (2008).

Li et al. (2020) analysed the correlation coefficients of O3 and PM2.5 at different heights of the SZMGT in December 2017, formulating different conclusions from this study. In December 2017, the correlation coefficient of O3 and PM2.5 increased significantly with altitude; they stated that this is because PM2.5 was also mainly generated by photochemical reaction at high altitudes, thus having a strong correlation with O3. At lower altitudes, there may be a greater contribution from the primary source to PM2.5, resulting in a weaker correlation with O3. In this study, the correlation between PM2.5 and O3 was weak at all altitudes during the pre-lockdown period. This may have occurred because Shenzhen had conducted a large number of pollution emission control strategies over the past two years, resulting in a significant decrease in the levels of primary pollutants. When compared with the measured data in the winter of 2017 (Li et al., 2020), the average concentration of PM2.5 in the whole surface layer during the pre-lockdown period decreased by 18.1%, whereas the average O3 concentration had declined by 36.2%. The decrease in the O3 concentration in the surface layer was twice that of PM2.5, indicating that there were much fewer products from photochemical reactions and secondary aerosol formation; as such, the correlation coefficient between O3 and PM2.5 concentrations was no longer high even in greater altitudes. During the lockdown, the average PM2.5 concentration had once again decreased due to drastic reductions in primary emissions. During this process, PM2.5 from primary emissions had become insignificant, and photochemical oxidation became an important source of PM. This strengthened the correlation coefficients between O3 and PM2.5 compared to those observed in pre-lockdown.

Fig. 11 compares the correlation between PM2.5 and NOx concentrations before and during the lockdown. The correlation trend between PM2.5 and NOx was the exact opposite of that between PM2.5 and O3; this means it was strong pre-lockdown (R = ∼0.5 at all altitudes), and much weaker after the lockdown (R = ∼0.2 at all altitudes). This indicates that there may be significant differences between PM2.5 sources before and during the lockdown. It was not possible to carry out composition analysis to determine the underpinning reason for the close correlation between PM2.5 and NOx emissions prior to the lockdown and not during the lockdown, due to a lack of PM2.5 compounds observation. Some recent studies have identified particulate nitrate is an important component of water-soluble aerosols in this region (Wu et al., 2020; Yang et al., 20210), which is conducive to the results shown in Fig. 11. One possible reason for this is that the primary emissions of PM2.5 in the PRD may have contributed a large proportion of the total PM2.5emissions before the lockdown; as NOx may be treated as an indicator of anthropogenic emissions, the primary emissions of PM2.5 decreased significantly during the lockdown. The other possible explanation is that the nitrate content of PM2.5 in the PRD decreased significantly during the lockdown. This is because a previous study reported that nitrate accounts for a large percentage of PM2.5, prior to the year 2020 (Yang et al., 2020).

Fig. 11

Scatterplots of the PM2.5 and NOx concentrations at different heights of the meteorological tower before and during the lockdown: (a) and (b) ground level; (c) and (d) low level; (e) and (f) middle level; and (g) and (h) high level.

Fig. 12 displays the correlation between the O3 and NOx concentrations before and during the lockdown. Prior to the pandemic, O3 and NOx were negatively correlated with each other. The relationship between O3 and NOx concentrations was fitted with an exponential function. During the lockdown, there was significant weakening of the (negative) correlation between O3 and NOx; this means at very low NOx concentrations, variations in the concentration of this pollutant appeared to have no clear effect on O3 concentrations.

Fig. 12

Scatter plots of O3 and NOx concentrations at different heights of the meteorological tower before and during the lockdown: (a) and (b) ground level; (c) and (d) low level; (e) and (f) middle level; and (g) and (h) high level.

A comparison of scatter plots before and during the lockdown showed that PM2.5 was poorly correlated to O3, albeit closely correlated to NOx before the lockdown. This indicates that a relatively large proportion of PM2.5 may originate from primary emissions or nitrate aerosol, although secondary aerosols may still account for a major part of PM2.5. Following the implementation of lockdown, PM2.5 became closely correlated to O3, and not to NOx. This indicates that the formation of PM2.5 and O3 may be highly regulated by one precursor (e.g., VOC), or the formation of PM2.5 during lockdown may primarily be limited by atmospheric oxidants, such as O3, where the fraction of primary PM2.5 may have been nearly eradicated.

Conclusions and implications

This study investigated changes in NOx, O3, and PM2.5 concentrations over the PRD caused by local COVID-19 lockdown. These changes were examined through the analysis of the vertical distribution of pollutants (NOx, O3 and PM2.5) before and during the lockdown by using data from the SZMGT. The conclusions of this study are as follows:

The advent of the COVID-19 pandemic forced a dramatic decrease in human activity. This greatly reduced the emission of primary pollutants, such as NOx, changing the chemical environment of the near-surface atmosphere. The PM2.5 concentration had also reduced significantly due to the decrease in precursor availability.

The reduction in primary pollutant emissions during the COVID-19 lockdown significantly decreased MDA8O3, whereas it did not decrease the daily average O3 concentration. The diurnal O3 concentration patterns were changed by the lockdown, where day-time concentrations were lower and night-time concentrations higher than pre-pandemic concentrations at all levels.

The correlation between PM2.5 and O3 concentrations was insignificant before the lockdown, and strengthened following the lockdown (p < 0.05) regardless of altitude. By contrast, the correlation between PM2.5 and NOx was much weaker during the lockdown. The results imply the PM2.5 composition may have changed from being predominantly from primary emissions or nitrate aerosols before the lockdown, to being predominantly a secondary organic aerosol, but the validation of this hypothesis required in further studies.

Prior to the COVID-19 pandemic, O3 and NOx concentrations were significantly negatively correlated. This correlation virtually disappeared following the outbreak of the pandemic. It may be concluded that at very low NOx concentrations, variations to its concentration have nearly no effect on the O3 concentration.

Overall, the advent of COVID-19 has devastated economies and societies around the world. However, the dramatic reduction in human activity from the lockdown measures provides unique opportunity for check the response of the atmosphere to human activities. The data indicate that the atmospheric chemical environment of the PRD has changed during the pandemic, leading to a drastic change in pollutants concentrations. These results provide a clear indication of the outcomes of the pollution mitigation policy. In the past, a number of environmental policy studies cast doubt as to whether it was necessary to further reduce traffic emissions. This was because in some areas, decreasing NOx concentration led to an increase in the O3 concentration. While this study shows that the continuous reduction in NOx emissions may reduce the peak O3 and MDA8O3, but not further reduce the daily average O3 concentration.

CRediT authorship contribution statement

Lei Li: Conceptualization, Methodology, Software, Writing – review & editing, and, Project administration. Chao Lu: Methodology, Software, Writing – review & editing. Pak-Wai Chan: Data curation. Zijuan Lan: Data curation. Wenhai Zhang: Data curation. Honglong Yang: Data curation. Haichao Wang: Conceptualization, Methodology, Software, Writing – review & editing, Supervision, and, Project administration.

Declaration of competing interest

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