Air pollutant variations in Suzhou during the 2019 novel coronavirus (COVID-19) lockdown of 2020: High time-resolution measurements of aerosol chemical compositions and source apportionment.

To control the spread of the 2019 novel coronavirus (COVID-19), China imposed rigorous restrictions, which resulted in great reductions in pollutant emissions. This study examines the characteristics of air pollutants, including PM2.5 (particles with aerodynamic diameters < 2.5 μm), gas pollutants, water-soluble ions (WSIs), black carbon (BC) and elements, as well as the source apportionment of PM2.5 in Suzhou before, during and after the Chinese New Year (CNY) holiday of 2020 (when China was under an unprecedented state of lockdown to restrict the COVID-19 outbreak). Compared to those before CNY, PM2.5, BC, SNA (sulfate, nitrate and ammonium), other ions, elements, and NO2 and CO mass concentrations decreased by 9.9%-64.0% during CNY. The lockdown policy had strong (weak) effects on the diurnal variations in aerosol chemical compositions (gas pollutants). Compared to those before CNY, source concentrations and contributions of vehicle exhaust during CNY decreased by 72.9% and 21.7%, respectively. In contrast, increased contributions from coal combustion and industry were observed during CNY, which were recorded to be 2.9 and 1.7 times higher than those before CNY, respectively. This study highlights that the lockdown policy that was imposed in Suzhou during CNY not only reduced the mass concentrations of air pollutants but also modified their diurnal variations and the source contributions of PM2.5, which revealed the complex responses of PM2.5 sources to the rare, low emissions of anthropogenic pollutants that occurred during the COVID-19 lockdown.

Introduction

With rapid economic development and acceleration of urbanization, the Yangtze River Delta (YRD), which exhibits complex emission patterns of air pollutants, has suffered serious air pollution problems in recent years (Huang et al., 2019; Li et al., 2018, 2020a; Wang et al., 2014a, 2016). China has prioritized a series of legislative actions for air quality improvement; for instance, the Air Pollution Prevention and Control Action Plan (the “Action Plan”) was issued in September 2013. Currently, the air quality in the YRD has improved significantly with PM2.5 (particles with aerodynamic diameters of <2.5 μm) concentrations decreasing by 13.3 μg m−3 (22.1%) from 2014 to 2019 (Dai et al., 2021). However, considering the impact of pollutant emissions, unfavorable meteorological elements and transport processes, haze events also occur frequently in winter in the YRD (Huang et al., 2020b; Kang et al., 2019; Li et al., 2018, 2020a; Wang et al., 2014a). For example, airflow from the northwest transports air pollutants to the southeast in China and is triggered by Siberian high pressure in winter; air pollutants can be transported from the North China Plain (NCP) to the YRD by cold fronts (Kang et al., 2019; Liu et al., 2020). Moreover, the chemical compositions of aerosols vary according to the prevailing meteorological conditions (Li et al., 2019; Tang et al., 2016; Wu et al., 2018). For instance, Wu et al. (2018) reported that the peak sizes of sulfate, nitrate and OA all shifted toward larger sizes with increasing relative humidity (RH) in Nanjing during wintertime, which reflects the effects of aqueous-phase processing. Hence, it is crucial to reveal the formation mechanism of regional haze pollution by clarifying the distributions and potential sources of air pollutants.

Positive matrix factorization (PMF) is the most popular and widely used model for source apportionment studies internationally and is used in the study of haze processes and urban PM2.5 pollution (Jaeckels et al., 2007; Kim et al., 2018; Li et al., 2020b; Okuda et al., 2010). For example, by using the PMF model, Zheng et al. (2020) suggested that primary emissions have decreased while secondary formation has increased since the lockdown in Wuhan. Moreover, the accuracy of PMF results is highly dependent on the time resolution of the observation data. Yu et al. (2019) reported that 1-h measurement data were preferred over 23-h averages for source apportionments during PMF model simulations due to the limits of inter-sample variability. Numerous studies have been carried out that involve PMF studies on air pollutants in the YRD (Mo et al., 2017; Wang et al., 2020; Tang et al., 2016; Yu et al., 2019), which are crucial to air quality control and subsequent policy formulation. Overall, a previous study demonstrated the great impacts of air pollution control policies, such as the activities of APEC, the 2nd World Internet Conference and the 2016 G20 Summit, on pollutant emissions; these policies also significantly affected the chemical compositions, spatiotemporal variations and formation mechanisms of aerosols (Cai et al., 2017; Shen et al., 2017; Sun et al., 2016; Yu et al., 2018). However, the true contribution of emission controls to particle matter (PM) reduction remains poorly constrained, largely due to variations in meteorological factors and regional atmospheric transport (Sun et al., 2016). Moreover, emission controls have usually targeted certain areas of China for a guarantee of previous events, with a focus on Hangzhou and its surroundings for the G20 Summit, for instance (Yu et al., 2018), which cannot eliminate the impacts of regional transport of air pollutants as a result.

The 2020 Chinese New Year (CNY) holiday was originally scheduled to take place from January 24 to 30, 2020. To control the spread of the 2019 novel coronavirus (COVID-19), the 2020 CNY holiday was delayed until February 10 (a delay of 18 days in total). During the CNY holiday, quarantines and roadblocks were imposed in urban and rural areas, which dramatically influenced the country’s economy. Energy demand and industrial output were reported to decrease sharply during lockdown periods (Le Quéré et al., 2020, Myllyvirta, 2020). These restrictions are believed to have caused large reductions in air pollutant emissions in China (Huang et al., 2020a; Chang et al., 2020; Shi and Brasseur, 2020). During the COVID-19 lockdown, satellite-derived NO2 column density data showed substantial decreases of 40% on average over Chinese cities relative to the same period in 2019 (Bauwens et al., 2020). Bao and Zhang (2020) found that travel restrictions reduced human mobility by 69.85%; reductions in AQI, PM2.5 and CO levels were partially mediated by human mobility; SO2, PM10 (particles with aerodynamic diameters of <10 μm) and NO2 levels were completely mediated. Primary pollutants were reported to have decreased dramatically due to the lockdown policy, while air oxidation and secondary pollutant levels (O3, secondary aerosols) were reported to have increased (Chang et al., 2020; Huang et al., 2020a; Xu et al., 2020; Zheng et al., 2020). However, the variations in chemical compositions and local and regional transport sources of PM2.5 during the lockdown period remain unclear.

As an influential city in the YRD, Suzhou had a permanent residential population of 10.75 million people, and the number of residents who owned vehicles was 4.1774 million in 2019 (http://www.tjcn.org/tjgb/10js/36361_4.html). Numerous investigations of atmospheric pollution in Suzhou have been conducted recently (Costabile et al., 2006; Tian et al., 2016; Wang et al., 2015; Yang et al., 2012). Yang et al. (2012) reported that the health impacts of O3 are more evident in the cooler seasons in Suzhou. Tian et al. (2016) reported that OM, (NH4)2SO4 and NH4NO3 are major contributors to visibility impairment, but this share differs from that of haze events. Wang et al. (2017) reported a summer aerosol concentration in Suzhou of 12,797 ± 5931 cm−3 with a unimodal distribution of the spectrum peaking at 100–110 nm. Due to the several impacts of the CNY and COVID-19 lockdown, emission reduction levels were substantial. As such, investigating the changes in the chemical components of aerosols that occurred during this special period can deepen our understanding of the impact of anthropogenic sources on air pollutants in urban areas. To date, numerous studies have been carried out on the impact of the lockdown on air quality and on source apportionment of aerosols in the NCP and in Hubei Province (Cui et al., 2020; Dai et al., 2020; Sun et al., 2020; Zheng et al., 2020); however, few studies have been conducted on the YRD. Given that emission sources, economic structures and meteorological patterns differ in different regions of China, the impact mechanisms of the lockdown policy on the distributions and sources of air pollutants were distinct in each region. This study examines variations in air pollutants, including PM2.5, gas pollutants, water-soluble ions (WSIs), black carbon (BC) and elements, that were measured in the major city of Suzhou in the YRD before, during and after the 2020 CNY holiday. Moreover, the distinct PM2.5 sources that were involved in different stages were identified using a 1-h dataset that was obtained via positive matrix factorization (PMF) model analysis to reveal the characteristics sources to which the residents of Suzhou were exposed and to develop better pollution management strategies.

Methods and materials

Observation datasets

The observation site is located at the South Gate Station in Suzhou (31.29°N, 120.63°E), which is a national control point of the China Environmental Monitoring Station; this site is surrounded by residential areas and is located 100 m from the southwest road of Suzhou to the south, which makes it a mixed region of commercial and residential areas. Air pollutants and meteorological data were recorded with a time resolution of 1 h, and the observation period ran from January 1 to February 29, 2020. WSIs in PM2.5 were detected using an Ambient Ion Monitor-Ion Chromatograph (AIM-IC, URG-9000D, Thermo Scientific™, USA), which can continuously measure ion mass concentrations, including Na+, NH4 +, K+, Mg2+, Ca2+, Cl−, NO3 − and SO4 2− (Malaguti et al., 2015). Quality control experiments, including collection efficiency, flow calibration and air tightness tests, were carried out each month. An internal check standard of lithium bromide injected into each sample was also used to confirm IC analysis accuracy. Additionally, ICs were calibrated by testing concentration gradients of the standard agent (Merck, Germany) using the external calibration method (Tang et al., 2020). Twenty-three elements (e.g., K, Fe, Zn, Ca, Si, Mn, Pb, Cu, Ti, As, V, Ba, Cr, Se, Ag, Cd, Ni, Au, Co, Sn, Sb, Tl, and Hg plus Pd for quality assurance) were monitored by an Xact 625 ambient metals monitor (Cooper Environmental Services (CES), Beaverton, OR, USA) via X-ray fluorescence (XRF) (Furger et al., 2017). For each sample analysis, the detector energy gain was automatically adjusted using pure Pd as an internal standard. The XRF response of the element of interest was calibrated using the standard thin film provided by the manufacturer. The measured mass was in good agreement with the standard mass for each element, and the deviation was less than 5%. BC mass concentrations were observed with a Model AE-31 aethalometer (Magee Scientific, USA), which simultaneously measures the attenuation of light as it passes through a particle laden filter spot and through a particle free filter spot (reference spot) of filter tape at seven wavelengths (e.g., 370, 470, 520, 590, 660, 880 and 950 nm) (Hansen and Schnell, 2005). Although measurements were conducted every 5 min, the data were averaged over a period of 1 h to decrease uncertainties originating from instrumental noise, flow rates, filter spot areas and detector responses (Corrigan et al., 2006). Silica gel in the drier was used upstream of the AE-31 instrument to change the monitored relative humidity (RH) to ensure that the RH of the sample air was below 40%. Air in the container was released to the area surrounding the drier to reduce the temperature difference between the interior of the container and drier. Temperatures within the container were maintained at approximately 25 °C. PM2.5 concentrations were measured by a Synchronized Hybrid Ambient Real-time Particulate Monitor (SHARP Model 5030i, Thermo Scientific™, USA), and gas pollutants, NO2, SO2, CO and O3, were monitored by a series of monitors (e.g., 42i, 43i, 48i, and 49i, Thermo Scientific™, USA) (Wei et al., 2020). PM2.5 concentrations were measured using the β absorption method. NO2, SO2 and O3 concentrations were measured using chemiluminescence, ultraviolet fluorescence, and UV spectrophotometry methods, respectively. CO was measured using the nondispersive infrared absorption method and gas filter correlation infrared absorption method. Wang et al. (2014b) provide a more detailed introduction to these instruments. Meteorological data (e.g., temperature, relative humidity (RH), wind speed, wind direction, visibility and precipitation) were observed with the CSI-CR1000 (Logan, UT, USA) automatic weather station.

Measurement periods were classified as being before (January 1 to 23, 2020), during (January 24 to February 10, 2020), and after (February 11 to 29, 2020) the CNY holiday of 2020.

Source apportionment

The U.S. Environmental Protection Agency (EPA) PMF version 5.0 has been widely applied for PM2.5 source apportionment (Bari and Kindzierski, 2018; Callén et al., 2014; Ji et al., 2018; Paatero and Tapper, 1994; Zheng et al., 2020; Zíková et al., 2016). A detailed description of the PMF model can be found in the supplementary material. WSIs, BC, NO2, SO2 and elements were included for PMF analysis. Species were categorized as strong, weak, and bad according to their signal-noise ratios and percentages below detection limits (Callén et al., 2014). Species were categorized as “bad” when the signal-to-noise (S/N) ratio was <0.5, “weak” if the S/N ratio was greater than 0.5 but less than 1 and “strong” if the S/N ratio was greater than 1 (US EPA, 2014). Finally, datasets of (sample number × species number) 552 × 26, 432 × 26 and 456 × 26 were introduced into the model before, during and after CNY periods, respectively. Choosing an optimal factor number is challenging. Considering too many factors results in meaningless outputs, while considering too few factor numbers may result in mixed sources (Bressi et al., 2014). The PMF model was run with a random seed, and the lowest Q was considered as the base run solution. Q-values, resulting source profiles and scaled residuals were examined. Resolution conditions can be found in the supplementary material. Finally, five factors that were in effect before and after CNY and six factors that were in effect during CNY were considered to be the most reasonable solutions. The uncertainties of the PMF model are usually estimated by bootstrapping (BS), displacement (DISP), and bootstrapping with displacement (BS-DISP). All factors are mapped over 96%–99% in BS. There was no factor swap and no decrease in the Q value in DISP. There were no swaps with DISP, and 97%–98% of the BS-DISP runs were successful.

Results and discussion

Changes in air pollutant loading

PM2.5 concentrations during CNY were noticeably lower than those before CNY (Fig. 1 c), whereas their hourly concentrations still exceeded 75 μg m−3 several times under periods with rigorous restrictions, such as from 00:00–16:00 on January 30, from 03:00–15:00 on January 31, from 13:00–19:00 on February 3 and from 07:00–11:00 on February 9, when the maximum PM2.5 concentrations varied from 89 to 118 μg m−3. According to Chinese air quality standards, a 24 h-mean PM2.5 concentration of more than 75.0 μg m−3 indicates the presence of air pollution. These pollution episodes were weaker than those that occurred before CNY but were much stronger than those occurring after CNY. During CNY, PM2.5 had a mean concentration of 38.8 μg m−3, which was 36.9% lower than that measured before CNY (61.4 μg m−3) and slightly higher than that measured after CNY (31.8 μg m−3) (Fig. 2 b).

Fig. 1

Time series of air pollutants and meteorological factors.

Fig. 2

Pollutant concentrations before, during and after CNY.

As shown in Fig. 1c, the variations in chemical compositions of PM2.5 were similar to those of PM2.5, while large discrepancies were still found among the different chemical species. Given the impact of fireworks, K+ concentrations were mostly greater than 1.0 μg m−3 from 22:00 on January 24 to 03:00 on January 25 and peaked at 00:00 on January 25 with a value of 3.1 μg m−3, which exceeded the levels measured in non-fireworks periods by more than 10-fold. The mean concentrations of BC, SNA (i.e., sulfate, nitrate and ammonium), other ions and elements during CNY were 1.5, 10.3, 1.2 and 1.6 μg m−3, respectively, (Fig. 2a), and were 51.1%, 48.2%, 9.9% and 23.3% lower, respectively, than the concentrations measured before CNY and were 0.5%, 29.7%, 23.6% and 41.7% higher, respectively, than those measured after CNY. Although people started returning to work in China after the Spring Festival (February 10), Jiangsu Province modified its level-I response to a level-II response on February 25, 2020 due to the public health emergency. Hence, the lockdown policy was not fully lifted after CNY, which resulted in relatively low pollutant concentrations. Li et al. (2020a) reported that industrial activity is the dominant PM2.5 contributor with a relative contribution of 32.2%–61.1% in the YRD, while the relative contribution from residential sources was higher during the lockdown period. Moreover, high PM2.5 concentrations were observed on New Year’s Eve (January 24) during CNY, which is attributed to fireworks. A short PM2.5 peak from January 27 to February 1 was also recorded and was due to unfavorable meteorological conditions, which are discussed in detail elsewhere (Shen et al., 2020). Hence, air pollutants reached higher concentrations during CNY than after CNY.

During CNY, NO2 concentrations were significantly lower than they were before and after CNY, and they varied inconsistently with the PM2.5 concentrations, which were different from the changes observed before CNY (Fig. 1b). For instance, NO2 concentrations were relatively high at 70–73 μg m−3 from 21:00 on January 21 to 00:00 on February 1, during which the PM2.5 concentrations were only 42–49 μg m−3. In contrast, the NO2 variations generally conformed to the PM2.5 concentrations during the haze period before CNY (i.e., January 11 to 16). In addition, meteorological conditions also influenced PM2.5 concentrations, such as high RHs and low wind speeds, which also led to similar variations in air pollutants (Fig. 1a and c). As such, the correlation coefficient of NO2 and PM2.5 was 0.57 before CNY and was thus much higher than that during (0.19) and after (0.32) CNY, which is in agreement with other results (Bauwens et al., 2020; Chang et al., 2020; Xu et al., ，2020). The correlation coefficient of NO2 and NO3 − was 0.47 before CNY and was thus higher than that during (0.16) and after (0.26) CNY. Such high correlations before CNY reveal that vehicle exhaust contributes greatly to NO2 and NO3 − and further increases PM2.5 concentrations. Hence, the NO2 variations were mostly similar to the PM2.5 variations. A dramatic reduction in traffic sources was the main driver of decreased NO2 concentrations during CNY and reached a mean of 16.2 μg m−3 which was 64.0% and 23.1% lower than before (45.0 μg m−3) and after (21.0 μg m−3) CNY, respectively (Fig. 2d). The decreased contributions of traffic sources to PM2.5 also led to lower correlation coefficients between NO2 and PM2.5 during CNY.

SO2 concentrations showed flat variations during the whole observation period (Fig. 1b), and they had mean concentrations of 5.0, 4.0 and 5.2 μg m−3 before, during and after CNY, respectively (Fig. 2c). CO concentrations showed variations that were consistent with those of PM2.5, with a mean of 0.7 mg m−3 that was measured during CNY, which was 5.2% lower and 16.5% higher than the concentrations measured before and after CNY, respectively (Fig. 2f). The mean O3 concentrations reached 66.8 and 62.7 μg m−3 during and after CNY, respectively, which were 2.0 and 1.9 times higher than that before CNY, respectively (Fig. 2e). Such opposing changes are mainly explained by the unprecedented reduction in NO emissions that led to less O3 titration by NO (Sicard et al., 2020). Greater decreases in NO2 concentrations during and after CNY resulted in weaker titration effects and even higher O3 concentrations (Fig. 2). In addition, the sharp reduction in PM2.5 concentrations that occurred during and after CNY could also have resulted in weakened solar radiation extinction; such conditions favor ozone formation. A detailed investigation of O3 enhancement in the YRD during the lockdown has been conducted elsewhere (Huang et al., 2020a).

Diurnal variations of air pollutants

Fig. 3 b shows that the diurnal variations in PM2.5 exhibited a bimodal distribution during CNY and peaked at 10:00 and 21:00. Li et al. (2020a) found that human activities decreased significantly during COVID-19 in the whole YRD: industrial operations, vehicle kilometers traveled, and construction work were significantly reduced, which led to decreased SO2, NOx, PM2.5 and VOC emissions of approximately 16%–26%, 29%–47%, 27%–46% and 37%–57%, respectively. Given the sharp decrease in PM2.5 concentrations that resulted from the response to traffic sources during CNY, the two PM2.5 peaks that occurred during rush hour were caused by diurnal variations in the boundary layer. Notably, the extent of PM2.5 reduction at night was larger than that during the daytime during CNY and ranged from 25.2% (14:00) to 35.1% (08:00) from 07:00–15:00 and exceeded 38% in other time segments, with a maximum of 42.7% occurring at 03:00. Meanwhile, slight changes in PM2.5 concentrations were observed when precipitation occurred during CNY (Fig. 3a). Given that light rain dominated during the observation period with hourly precipitation rates mostly remaining below 1 mm h−1, scavenging effects on PM2.5 were weak as a result. Meanwhile, these stable weather conditions with high RHs were conducive to pollutant accumulation and to the formation of secondary aerosols, which could be the dominant driver of PM2.5 increases. Fig. 4 b reveals that wind speeds averaged approximately 3.1 m s−1 with few changes occurring throughout the study period, which suggest that there were few distinctions in diffusion conditions; as such, the discrepancies in air pollutant concentrations that occurred in different stages were caused mainly by changes in emission sources.

Fig. 3

Diurnal variations of PM2.5 and precipitation before, during and after CNY (note: precipitation refers to the cumulative precipitation during each period).

Fig. 4

Diurnal variations of meteorological factors before, during and after CNY.

Although diurnal PM2.5 variations underwent considerable changes in different stages (Fig. 3b), visibility values presented unimodal distributions (Fig. 4d). Moreover, the correlations between PM2.5 and visibility weakened with decreases in PM2.5 concentrations, with coefficients of −0.51, −0.46 and −0.33 measured before, during and after CNY, respectively. In particular, the differences between PM2.5 variations that were measured during and after CNY were much smaller than those of visibility according to comparisons between Fig. 3, Fig. 4d, which show that PM2.5 was not the principal factor that affected visibility when its concentrations were low. For instance, PM2.5 concentrations during CNY were 1.5–1.6 times higher than those after CNY from 09:00–11:00, during which the PM2.5 difference was the greatest, yet the difference in visibility was less than that measured from 13:00–17:00 (Fig. 4d). As the RH after CNY was lower than that during CNY from 13:00–18:00 (Fig. 4c), aerosol water contents probably had greater impacts on visibility when PM2.5 concentrations were low.

NO2 exhibited bimodal distributions during different stages and peaked at 08:00 and 20:00 during and after CNY, whereas NO2 peaked at 10:00 and 19:00–21:00 before CNY (Fig. 5 a). The rush hours may have delayed the morning peak and advanced the evening peak. Diurnal variations in NO2 were similar in different stages, with the lowest values recorded at noon. High wind speeds at noon (Fig. 4b) were responsible for NO2 diffusion. Meanwhile, high temperatures (Fig. 4a) and low RHs (Fig. 4c) also encouraged photochemical reactions of NO2. During the lockdown period, traffic sources decreased significantly, which resulted in a lower contributions to NO2, which were also impacted by industrial and natural sources. As such, the diurnal variations in NO2 were little affected by the lockdown. Unimodal distributions were recorded for diurnal variations in SO2 that peaked from 10:00–13:00 in different stages with small discrepancies in concentrations (Fig. 5b), which can be attributed to slight variations in SO2 emission sources that were mainly affected by industrial coal burning. Diurnal variations in CO exhibited a bimodal distribution before CNY, with peaks occurring from 08:00–09:00 and from 19:00–21:00. However, trimodal distributions were observed for CO concentrations during and after CNY and peaked at 09:00, 13:00 and 19:00. CO is generally produced by incomplete combustion and is reported to be mainly generated from industry and vehicle exhaust in urban areas (Parrish et al., 2009; Sahu et al., 2015); as such, the CO diurnal variations showed a similar pattern as those of NO2. Unimodal distributions were consistently found for O3 and peaked from 13:00–16:00 in different stages (Fig. 5d), which are consistent with the variations in air temperature (Fig. 4a).

Fig. 5

Diurnal variations of trace gases before, during and after CNY.

The diurnal variations in BC exhibited a bimodal distribution before CNY and peaked at 07:00 and 19:00 (Fig. 6 a), and then changed to a trimodal distribution during CNY with peaks occurring at 07:00, 13:00 and from 17:00–19:00. A flat variation was found for BC after CNY, which was similar to those of SNA for different stages with higher concentrations than those measured before CNY.

Fig. 6

Diurnal variations of chemical components before, during and after CNY.

Diurnal variations in SO4 2−, NO3 − and NH4 + showed minor differences throughout the observation period (Fig. S1), consistent with those of SNA (Fig. 6b). Trajectory lengths represent the transport speeds of air masses. Fig. S2 shows that the 24-h backward trajectories were similar for different periods and mostly exceeded 300 km, which suggested great impacts from the foreign pollutant transport process on air pollutants in Suzhou. Meanwhile, northerly air masses were dominant during the observation period and came mainly from the northwest inland and northeast marine areas. Some southerly air masses were noticed before and after CNY, which were also from marine areas with source characteristics that were similar to the northeast marine air masses. Overall, such long-range transport of air masses was conducive to the aging process of aerosols, which further promoted secondary ion formation, such as SO4 2−, NO3 − and NH4 +, which showed weak diurnal variations as a result. SO4 2− exhibited a unimodal distribution that peaked at 14:00 during CNY. However, there were few diurnal differences in SO4 2− before and after CNY. SOR exhibited high and low concentrations in the daytime and nighttime, respectively, during CNY, which were consistent with those of SO4 2−. In contrast, lower and higher SOR values were recorded in the daytime and nighttime, respectively, before and after CNY (Fig. S3a). Mean SO4 2− concentrations reached 5.3, 3.6 and 1.6 μg m−3 before, during and after CNY, respectively, and showed larger discrepancies than those of SO2 (Fig. 2c). The correlation coefficients between SO2 and SO4 2− were 0.5, 0.5 and 0.2 before, during and after CNY, respectively. The SOR values after CNY were much lower than those before and during CNY, which revealed a minor contribution of SO2 conversion to SO4 2− after CNY. Overall, sulfate can be sourced from heterogeneous processes that were distinct from the photochemical reactions of SO2 during the observation period. Shen et al. (2021) also reported that nitrate and sulfate formation can be more strongly influenced by aerosol water contents, pH values and air oxidation compared with their precursors.

NO3 − exhibited mean concentrations of 9.7, 3.9 and 3.3 μg m−3 before, during and after CNY, respectively, with changes occurring in different stages that were similar to those of NO2 (Fig. 2d). However, these diurnal variations in NO3 − were much weaker than those of NO2 (Fig. S1b). NOR showed significant diurnal variations which peaked at 04:00 and 15:00 (Fig. S3b) and lagged behind those of NO2, which is ascribed to the conversion of NO2 to NO3 −. It is worth noting that NO3 − concentrations were much higher before CNY than during and after CNY, while NOR concentrations showed small differences before, during and after CNY, which suggest a more complex source of NO3 − before CNY. The correlation coefficients between NO2 and NO3 − were 0.5, 0.2 and 0.3 before, during and after CNY, respectively, which suggested that NO3 − levels were strongly impacted by vehicle exhaust before CNY when compared with the levels during and after CNY when sharp decreases in traffic emissions were observed. Nitrate, which normally originates from photochemical and heterogeneous reactions (Shen et al., 2021), may be generated mostly by the aerosol aging process during its transportation, and after reaching Suzhou, such effects of pollutant transportation on nitrate cannot be neglected as a result. NH4 + concentrations underwent diurnal variations similar to those of NO3 − (Fig. S1) with mean concentrations of 4.8, 2.7 and 2.3 μg m−3 measured before, during and after CNY, respectively.

The NO3 −/SO4 2− ratios showed large discrepancies among different stages (Fig. S3c). Similar to NO2, the NO3 −/SO4 2− ratios were high during rush hour and were usually above 1.5 with a mean value of 1.7 before CNY, which revealed the noticeable contributions from vehicle exhaust. NO3 −/SO4 2− ratios decreased to a mean of 1.2 during CNY, with high and low values occurring in nighttime and daytime, mainly due to the reduced traffic emissions that resulted from the lockdown policy. The NO3 −/SO4 2− ratios increased again after CNY and averaged 2.5, which is ascribed to work resumption.

Large discrepancies were found for the other ions and elements in different stages (Fig. 6c and d). Before CNY, other ions and elements showed similar diurnal variations to those of PM2.5, with low concentrations measured from 12:00–15:00. During CNY, higher concentrations of other ions and elements were noted before dawn, and in daytime, the former peak was caused by fireworks that released K+, Mg2+, K, Cu and Ba ions and elements. The latter is ascribed to higher wind speeds and low RHs, which were also responsible for their unimodal distributions after CNY, with higher values recorded in the daytime.

Variations in source contributions

Based on the PMF-modeled results, five factors were derived before and after CNY (Fig. S4 and S5), and six factors were identified as part of the optimal solution for the period during CNY (Fig. 7 ). Strong correlations were found between the modeled and observed concentrations before, during and after CNY (Fig. S6). The slope of the fitting function reached over 0.98 with R values ranging from 0.98 to 0.99, which indicate that the observations were reasonably represented by the PMF simulation.

Fig. 7

Profile and contribution of each source during CNY.

The first factor is coal combustion in view of the higher observed loadings of BC, SO2, AS, Se and Pb. Se and Pb have been widely reported to be tracers of coal combustion (Liu et al., 2018; Zheng et al., 2020), while BC is mainly generated from the incomplete combustion of fossil and biomass fuels (Qin and Xie, 2012; Streets et al., 2001), and SO2 is normally produced by industrial coal burning processes (Kharol et al., 2020). The second identified factor refers to industrial processes with larger contributions of the heavy metal elements, Cr, Mn, Fe, Ni and Zn (Querol et al., 2007; Ji et al., 2018). The third factor, which is characterized by the higher loadings of Ca2+, Ca and Mg2+, pertains to dust sources (Zheng et al., 2020; Sudheer and Rengarajan, 2012). Vehicle exhaust was identified as the fourth factor, with higher NO2 fractions of 62.3%–79.3% observed in the three stages (Chan et al., 2011; Liu et al., 2017). Secondary processes were identified as the fifth factor due to the higher loadings of NO3 −, SO4 2− and NH4 +. Additionally, the sixth factor, fireworks, was also observed during CNY, as large contributions of K+, Mg2+, K, Cr, Cu and Ba were observed (Kong et al., 2015; Rai et al., 2020; Vecchi et al., 2008; Zheng et al., 2020).

Although source concentrations (11.2 μg m−3) and contributions (32.0%) of vehicle exhaust were higher than those of other sources during CNY, they decreased by 72.9% (39.7%) and 21.7% (19.5%) relative to those measured before (after) CNY (Fig. 8 ). These results exhibited bimodal distributions for the diurnal variations in vehicle exhaust after CNY (Figs. S7 and S8) with the peak hours corresponding to rush hours, which were consistent with results for NO2 and CO (Fig. 5). However, variations in vehicle exhaust showed subtle peaks before and during CNY. The fraction remained at high values from 12:00 before CNY and peaked slightly at 08:00 and 11:00 during CNY. Figs. S9–S11 show that vehicle exhaust levels showed high values under low wind speeds in different stages, which revealed that its origin was mainly from local emissions in Suzhou, which are ascribed to traffic arteries around the observation site (Fig. S12). As such, high concentrations of vehicle exhaust at night during the lockdown were also observed (Fig. S7d).

Fig. 8

Concentration and fraction of each source before, during and after CNY.

Slight changes in the contributions of secondary processes and dust were observed, with values of 20.3%, 20.9% and 16.0% found for the former and values of 12.3%, 10.0% and 6.1% found for the latter before, during and after CNY, respectively. The source concentrations of secondary processes and dust during CNY decreased by 53.2% and 62.9%, respectively, relative to those measured before CNY. The diurnal variations in secondary processes were consistent in different stages (Figs. S7 and S8), which showed that this source type was little affected by the CNY holiday and lockdown policy. High concentrations due to secondary processes were found under light southerly winds before CNY (Fig. S9), westerly winds during CNY (Fig. S10) and northerlies after CNY (Fig. S11). High wind speeds were accompanied by low RHs (Fig. 4) and were responsible for the high concentrations and proportions of dust sources during daytime (Figs. S7–S11). Due to human activities, dust concentrations peaked at 10:00, 15:00 and 21:00 before CNY (Fig. S7) and decreased noticeably during and after CNY due to the restrictions.

In contrast, increased contributions of coal combustion (20.7%) and industry (11.1%) were found during CNY, which were 2.9 and 1.7 times higher than those before CNY, respectively. The source concentrations of coal combustion and industry reached 5.5, 7.3, and 5.2 μg m−3 and 5.1, 3.9, and 4.4 μg m−3 before, during, and after CNY, respectively. Such flat variations in different stages reveal stable emissions from industrial processes, including power generation and iron and steel coking. The observed distributions of coal combustion were largely consistent with those of power plants (Figs. S9–S12). High concentrations of industry sources under northwesterly winds before and after CNY were observed, which highlight the strong impacts of industrial parks. During CNY, high industrial values were found in the northwestern and southwestern regions (Fig. S10), where power plants are distributed (Fig. S12), which indicate the strong influence of power plants on industrial concentrations. The diurnal concentrations of coal combustion showed subtle variations in different stages, which again confirmed a stable emissions source. It is worth noting that coal combustion and sea salt sources exhibited higher concentrations during CNY than before and after CNY from 04:00–15:00, which suggest the strong impacts of sea salt in view of the high wind speeds that occurred during this period (Fig. 4b). The proportions from coal combustion showed flat diurnal variations before and after CNY and reached peak values at 04:00, 09:00 and 13:00 during CNY (Fig. S8), probably due to the decreasing contributions of other sources, e.g., vehicle exhaust.

Finally, fireworks sources were also recognized, with concentrations and fractions of 1.8 μg m−3 and 5.3%, respectively, measured during CNY and values being higher at night (Figs. S7 and S8). These high fireworks emission concentrations largely corresponded to high wind speeds from the northwest, which suggest that their origins were from the surrounding suburbs (Fig. S10).

Conclusions

In this work, the characteristics of air pollutants (e.g., PM2.5, gas pollutants, WSIs, BC and elements) and PM2.5 sources in Suzhou of the YRD were investigated before, during, and after CNY. Compared to before CNY, the mass concentrations of PM2.5, BC, SNA, other ions, elements, NO2 and CO reached 38.8, 1.5, 10.3, 1.2, 1.6, 16.2 μg m−3 and 0.7 mg m−3, respectively, during CNY (COVID-19 lockdown), which represented decreases of 36.9%, 51.1%, 48.2%, 9.9%, 23.3%, 64.0% and 25.2%, respectively. The extent of PM2.5 reduction was greater at night, with a maximum of 42.7% occurring at 03:00. The lockdown policy had strong (weak) effects on diurnal variations in aerosol chemical compositions (gas pollutants). Compared to before (after) CNY, source concentrations and contributions of vehicle exhaust during CNY decreased by 72.9% (39.7%) and 21.7% (19.5%), respectively. However, slight changes in the contributions of secondary processes and dust were observed during CNY despite large reductions in their source concentrations (by 53.2% and 62.9%, respectively). In contrast, increased contributions of coal combustion and industrial activity were noticeable during CNY, being 2.9 and 1.7 times higher than those before CNY, respectively, whereas their source concentrations showed flat variations in different stages. This study highlights that the lockdown policy applied in Suzhou during CNY not only reduced air pollutant mass concentrations but also modified their diurnal variations and the source contributions of PM2.5, which revealed complex responses of PM2.5 sources to the rare low emissions of anthropogenic pollutants that occurred during the COVID-19 lockdown.

CRediT author statement

Honglei Wang, Conceptualization, Funding acquisition, Supervision, Writing – review & editing, Writing – original draft, Validation, Investigation. Qing Miao, Data curation, Methodology, Writing – original draft. Lijuan Shen, Conceptualization, Methodology, Software, Writing – original draft. Qian Yang, Data curation, Resources. Yezheng Wu, Data curation, Resources. Heng Wei, Data curation, Resources.

Declaration of competing interest

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