Introduction

Since COVID-19 was first reported in Wuhan, China, it has been spreading around the world for more than 2 years. Different countries have taken different measures to control the spread of the virus, and a complete lockdown measure is an effective way to comprehensively control the virus. At the same time, the COVID-19 lockdown has unexpectedly reduced pollutant emissions from anthropogenic sources and improved air quality. Under the lockdown measures such as restrictions on industrial activities, transport and travel (Lal et al. 2020), NO2, NO, and CO concentrations in Lyon were reduced by 67%, 78%, and 62%, respectively, compared to the normal situation (Sbai et al. 2021). In Delhi, the ban on almost all industrial activities and mass transportation resulted in reductions in PM10 and PM2.5 which were as high as approximately 60% and 39%, respectively, under the nationwide lockdown measures in India, compared to the pre-lockdown phase (Mahato et al. 2020). Furthermore, compared to pre-pandemic levels, NO2 concentration has been reduced by 25.5% due to non-essential business closures in the USA (Berman and Ebisu 2020). Wuhan, China, the first city in the world to be under strict lockdown, included actions such as quarantining, traffic restrictions, and factory closures (Cui et al. 2020), reducing NO2, PM10, PM2.5, and SO2 concentrations by 50.6%, 41.2%, 33.1%, and 16.6%, respectively, compared to the pre-lockdown period (Sulaymon et al., 2021). In Jiangsu Province, China, the mean change of PM2.5 decreased by 18%, and PM10 decreased by 19% from pre-COVID to active COVID (Bhatti et al. 2022). The concentrations of major pollutants SO2, NOx, PM2.5, and Volatile organic compounds (VOCs) in the Yangtze Delta region were reduced by up to 26%, 47%, 46%, and 57% respectively, during the city lockdown period (Li et al. 2020). NOx emissions were reduced by 36% over China, compared to the emissions before the outbreak (Feng et al. 2020), and the National Aeronautics and Space Administration (NASA) has published satellite images of the massive reduction in NO2 over China due to the economic slowdown and reduced human activities (NASA 2020).

Although the concentrations of NO2, NO, PM10, PM2.5, and SO2 decreased due to the implementation of epidemic control measures, O3 concentrations increased during the lockdown. Compared to the same period in 2017–2019, the daily O3 mean concentrations increased at urban stations by 24% in Nice, 14% in Rome, 27% in Turin, 2.4% in Valencia, and 36% in Wuhan during the lockdown in 2020 (Sicard et al. 2020). Similarly, O3 contents in the industrialized Gujarat state in western India increased by 16–48%, compared to the pre-lockdown (Selvam et al. 2020). This phenomenon is mainly due to stable HCHO concentrations in urban areas, which provides sufficient fuel for tropospheric O3 generation, especially when there is not enough NO to consume O3 through the titration effect (Pei et al. 2020).

Shijiazhuang, the capital city of Hebei province, is suffering from severe air pollution with a ranking of 167th in air quality among 168 cities in China in 2020 (MEP 2021a). It is also one of the cities in China to have imposed two lockdowns since the outbreak of COVID-19. The different control measures during the two lockdown periods provide a valuable opportunity to explore the effect of different measures on air quality. This paper analyzes the trends of the air quality index (AQI) and the characteristics of the changes in PM2.5 and O3, the most concerned atmospheric pollutants in China, during the two lockdowns. It then focuses on the causes of O3 growth in the lockdown, identifying the key species of O3 formation and changes in the sources of VOCs in the two lockdowns in Shijiazhuang in terms of NOx and VOCs, the main precursors of O3. This is important to explore the impact of the ban on air quality and to infer the extent to which regional emission reductions will improve air quality.

Methodology

Study area

Shijiazhuang (SJZ), the third-largest city in the Beijing-Tianjin-Hebei area, is located at 113°30′–115°20′E and 37°27′–38°47′N with an area of 14,464 km2, as shown in Fig. 1. SJZ consists of 22 districts or counties with a population of 11,023,586 as of November 1, 2020. Petroleum processing, chemical raw material and chemical product manufacturing, pharmaceutical manufacturing, electricity, and construction are the main industries in SJZ (Liu et al. 2018). At the same time, it is also the transportation center of Hebei Province.

Fig. 1
figure 1

Location of SJZ

Full size image

For the last 2 years, COVID-19 is raging around the world. In China, Wuhan was the epicenter of the epidemic in 2020. Following a nationwide lockdown, SJZ was less affected by COVID-19, with 29 confirmed cases throughout 2020 (http://www.nhc.gov.cn/). From 25 January to 9 February 2020, SJZ implemented the first-level response to a major public health emergency, which is defined as lockdown I. The lockdown I period includes the Chinese Spring Festival and Lantern Festival. However, on 5 January 2021, SJZ suddenly became the epicenter of the epidemic, with 862 confirmed cases in less than a month (http://www.nhc.gov.cn/). SJZ entered a state of war on January 8 and lifted that state on 29 January. Therefore, the period from 9 to 28 January was defined as lockdown II. During lockdown II, dusty weather was observed from 12 to 16 January, air quality data for these 4 days were removed from this study to truly explore the contribution of the lockdown measures to air quality changes.

Data source

Hourly concentrations of PM2.5 (μg/m3) and NOx (μg/m3), 8-h moving average concentrations of O3 (μg/m3), and hourly values of the AQI were obtained from China Environmental Monitoring Station.

Satellite-derived data including tropospheric formaldehyde (HCHO) and tropospheric NO2 column density were obtained from the Tropospheric Monitoring Instrument (TROPOMI) level-2 retrievals. TROPOMI is a satellite instrument on the Copernicus Sentinel-5 Precursor (S5P) satellite launched by the European Space Agency on 13 October 2017 to monitor air pollution. TROPOMI data was provided on ESA’s Sentinels Scientific Data Hub website.

VOC sampling

The monitoring site is located in the Shijiazhuang Ninth High School (38.03°N, 114.28°E), which is part of a mixed educational, residential, and commercial area in the urban area. Online VOC measurement was conducted using AirmOzone Analysis System (ASS) consisting of the AirmoVOC C2-C6 analyzer and AirmoVOC C6-C12 analyzer developed by Chromatotec®. The AirmOzone system uses a flame ionization detector (FID) and monitors a total of 71 VOC species in this study, including 24 alkanes, 12 alkenes, 1 alkyne, 16 aromatics, and 18 halocarbons. The system consists of an internal automatic calibration (AirmoCal) and an external manual calibration. The internal automatic calibration frequency is once a day and the external calibration frequency using VOCs standard gas is once a week, thus ensuring a stable and efficient continuous monitoring of the instrument.

Potential source contribution function

The potential source contribution function (PSCF) is a conditional probability function (Hui et al. 2018), which was applied in this study to locate potential sources of pollution. Using global data provided by the National Center for Environmental Forecasting, the backward trajectories of PM2.5 in SJZ during the two lockdowns were calculated in this study with a 24-h backward trajectory height of 100 m.

PSCF can be used to locate the potential source area of the observation points, which is defined as PSCFij = mij/nij. Where nij represents all trajectories in the airflow through the cell grid in the study area, mij represents the number of pollution trajectories through the grid. In this work, the geographic area covered by the trajectories was divided into an array of 0.1° × 0.1° grid cells, and the PM2.5 reference value was set at 75 μg/m3. The higher PSCF values indicate that the area corresponding to the grid is a potential source area for high concentrations of pollution at the observation site and that the trajectories through this area are a transport pathway that has a significant impact on the observation site.

Since grid cells always have the same PSCF value, it is difficult to distinguish between slightly higher and far higher thresholds (Li et al. 2017). To reduce the effect of small values of nij, PSCF values are generally multiplied by an arbitrary weighting function Wij to better reflect the uncertainty of these small values (Polissar 1999). Wij is defined in Eq. (1).

$${W}_{ij}=\left\{\begin{array}{l}1.00\quad {n}_{ij}>80\\ 0.75\quad 80\geq {n}_{ij}>20\\0.42\quad 20\geq {n}_{ij}>10\\ 0.05\quad {n}_{ij}\leq 10\end{array} \right.$$
(1)

Positive matrix factorization

Positive matrix factorization (PMF) receptor model version 5.0 is developed by the US EPA and used for source analysis of VOCs in SJZ in this study. The PMF model is based on the fundamental principle of mass conservation to identify and apportion source contributions from a given data matrix using Eq. (2) (Assan et al. 2018; Guan et al. 2020; Paatero and Tapper 1994).

$${X}_{ij}= \sum_{k=1}^{p}{g}_{ik}{f}_{kj}+ {e}_{ij}$$
(2)

where Xij represents the VOC concentration matrix with i number of samples and j number of measured VOCs, which are resolved by the PMF to provide p number of possible source factors with the source profile f of each source and mass g contributed by each factor to each sample, leaving the residuals e for each sample (Sarkar et al. 2017). To determine the solution, the minimum value Q is calculated from Eq. (3) (Paatero and Tapper 1994):

$$Q=\sum_{i=1}^{m}\sum_{k=1}^{n}\left[\frac{{x}_{ij}-\sum_{k=1}^{p}{g}_{ik}{f}_{kj}}{{u}_{ij}}\right]$$
(3)

where Q is the object function and a critical parameter for PMF, m and n are sample and species numbers. The uncertainties (Unc) were calculated following the US EPA recommended method as follows: (1) if the concentration values were below or equal to the MDL, their uncertainty was calculated using the following equation: Unc = 5/6 × MDL; (2) if the concentration values were greater than the MDL, the calculation used was: Unc = [(Error fraction × concentration)2 + (MDL)2]1/2 (Polissar et al. 1998; Reff et al. 2007).

The PMF model was run ranging from 4 to 10 factor numbers to determine the best solution for this study, consistent with the chemical environment in the lockdown of SJZ. The six-factor solution was considered to be the best solution for this dataset based on the constraints imposed by the Q/Qexp theoretical ratio, the physical likelihood of each factor, and the rotational ambiguity of the solution.

Air quality index

The AQI is a comprehensive pollutant evaluation index that allows for a comprehensive assessment of air quality. According to the Technical Regulation on Ambient Air Quality Index of China, AQI levels are divided into six classes, as shown in Table 1. The AQI is calculated as the maximum value of the air quality sub-index for all air pollutants. Table 2 shows the corresponding air pollution sub-index levels and the corresponding air pollutant concentrations. Six major air pollutants (SO2, NO2, PM10, CO, O3, and PM2.5) are selected; their concentrations are grouped into six different categories based on concentration breakpoints. In Table 2, 24-h refers to 24-h average, and 8-h refers to 8-h average. IAQI values can be calculated from Eq. (4) by linear interpolation of the reference scale values given in Table 2:

$$IAQIp=\frac{{PI}_{high}-{PI}_{low}}{{BP}_{high}-{BP}_{low}}\times \left({C}_{p}-{BP}_{low}\right)+{I}_{low}$$
(4)

where IAQIp refers to the air pollution sub-index for pollutant p; Cp represents the integer concentration of pollutant p; BPhigh represents the breaking point greater than or equal to Cp; BPlow represents the breaking point less than or equal to Cp; PIhigh refers to the air pollution sub-index corresponding to BPhigh; and PIlow refers to the air pollution sub-index corresponding to BPlow. Finally, the AQI is calculated as the maximum value of the air pollution sub-indices for all air pollutants, and the air pollutant with the maximum value is identified as the primary pollutant.

Table 1 Air quality index and levels
Full size table
Table 2 Air quality sub-index levels and their corresponding air pollutant concentrations
Full size table

Results and discussion

Characteristics of the changes in air quality and AQI

Figure 2 shows the changes in AQI and the proportion of each level during the two lockdowns and from 10 to 25 January 2019 (defined as the same period in 2019). Compared to the same period in 2019, the air pollution levels under the COVID-19 lockdowns decreased significantly, with a 13.2% decrease in AQI for lockdown I and a 32.4% decrease for lockdown II. In terms of the changes in AQI levels, there were 4 days of serious pollution and 1 day of heavy pollution during the same period in 2019, no serious pollution in either lockdown, 6 days of heavy pollution in lockdown I, and 2 days of heavy pollution in lockdown II. In terms of the number of mild and moderate pollution days, the three periods were similar. There were no excellent days in the same period in 2019, and lockdown I and lockdown II had 1 and 2 additional days, respectively. Overall, the improvement in air quality under the lockdown was mainly reflected in the decrease in the number of serious pollution days and the increase in the number of excellent days. Comparing the two lockdowns, lockdown II was more stringent and had better air quality.

Fig. 2
figure 2

AQI and air quality degrees in SJZ for the same period in 2019 and the two lockdowns

Full size image

Through the calculation of IAQIp, the daily primary pollutants in the same period of 2019 were PM2.5 with a 100% contribution; lockdown I was PM2.5 with a 100% contribution; and lockdown II was PM2.5 and PM10 with 47% and 53% contributions, respectively. PM2.5 was the primary pollutant in SJZ, and the effect of the lockdown on the change of PM2.5 concentration will be analyzed in detail below. Meanwhile, O3, the main pollutant in SJZ in summer, gradually increased during three periods, as the PM2.5 concentration decreased. Therefore, the changes of O3 concentration during the lockdown period will also be discussed below.

Characteristics of the changes in PM2.5 concentrations

In terms of individual pollutants, PM2.5 has been the primary pollutant in the Beijing-Tianjin-Hebei region. In 2020, the proportion of non-attainment days with PM2.5 as the primary pollutant reached 48% (MEP 2021b). Figure 3 (a) shows the trends of PM2.5 hourly concentrations at different AQI levels for three periods. On heavy and serious pollution days, the average PM2.5 concentration for the same period in 2019 was 286.6 μg/m3, which decreased by 37.3% to 179.6 μg/m3 during lockdown I and by 42.7% to 164.3 μg/m3 during lockdown II. On mild and medium pollution days, the average PM2.5 concentrations were relatively similar for the three periods, at 97.3, 109.1, and 106.1 μg/m3. Compared to the same period in 2019, on good and excellent days, the average PM2.5 concentrations in the three periods were 43.8, 53.5 and 34.7 μg/m3, respectively; a decrease of 20.8% for lockdown II. As shown in Fig. 3 (b), the average PM2.5 concentrations decreased by 12.9% and 42.4% for the whole lockdown I and lockdown II periods, respectively, compared with the same period in 2019. In addition, PM2.5 concentrations decreased by 33.8% during lockdown II, compared to lockdown I. It can be concluded that the decrease in PM2.5 concentrations in the lockdown occurred mainly on days of heavy and severe pollution, that is, the level of heavy and serious pollution decreased.

Fig. 3
figure 3

(a) Temporal variation (different colors correspond to the AQI level colors in Fig. 2) and (b) average concentrations of PM2.5 for three periods in SJZ

Full size image

The significant reduction in PM2.5 concentrations was largely due to the implementation of the lockdown measures. During lockdown I, the policy required people to avoid going outside unnecessarily and allowed one person per household per day to go out to work or shop with a transport permit. However, during lockdown II, due to the SJZ outbreak, the policy required all people to implement home quarantine, except for medical personnel and those ensuring basic city operations. The difference in policy between lockdown I and lockdown II directly resulted in a significant change in urban traffic mobility. In this study, the intensity of intra-city travel was introduced to characterize urban traffic mobility. The intra-city travel intensities during lockdown I and lockdown II were 1.9 and 1.4, respectively, both significantly lower than 4.0 during the same period in 2019 (http://qianxi.baidu.com). The value during lockdown II was 26.3% lower compared to lockdown I, which proved that lockdown II had fewer traffic sources under the stricter lockdown policy. At the same time, the lockdown policy led to a significant decrease in the level of industrial activity. Compared to the same period in 2019, industrial electricity consumption decreased by 18.8% in lockdown I and 40.0% in lockdown II (SBS 2021). With a 26.2% decrease in industrial electricity use in lockdown II compared to lockdown I, industrial emissions during lockdown II can be estimated to be 26.2% lower than during lockdown I (Li et al. 2021). Moreover, the decrease in PM2.5 concentrations during lockdown II was associated with a reduction in firework displays. Lockdown I spanned the Chinese Spring Festival and Lantern Festival, which are two fireworks festivals in China. Fireworks shows have a significant impact on PM2.5 concentrations, reaching 50% of PM2.5 during the Chinese Spring Festival (Kong et al. 2015; Zhang et al. 2017).

The lockdown policy can only reduce the emissions of local sources but not the contribution of regional transport sources. To clarify the contribution of regional transmission sources to PM2.5, we applied the PSCF model, and the results are shown in Fig. 4. With the city of SJZ as the center of the circle, three regions are divided into a radius of 50 km according to the administrative area of SJZ. The area within a radius of 50 km represents the contribution of local sources; the area between 50 and 100 km represents short- and medium-range transport sources; while the other areas correspond to long-range regional transport sources (> 100 km).

Fig. 4
figure 4

The potential source contribution function of PM2.5 during (a) the same period in 2019, (b) lockdown I, and (c) lockdown II in SJZ

Full size image

From Fig. 4 (a), during the same period in 2019, the highest PSCF values of 0.6–0.7 were found in areas within 50 km, which are industrial areas located in the south and southwest of SJZ. The short and medium distance transmissions between 50 and 100 km were mainly in the western and northeastern areas of SJZ, originating from the eastern part of Yangquan, the southern part of Baoding and the northern part of Xingtai, which were PM2.5 potential source areas, with PSCF values mostly between 0.2 and 0.4. When the distance exceeds 100 km, PM2.5 was more likely to come from the northern part of Baoding and the northwestern part of Yangquan, with PSCF values between 0.1 and 0.3. As shown in Fig. 4 (b), the high PSCF values were widely distributed within 50 km, indicating that the potential sources of PM2.5 during lockdown I were mainly local, especially in the city center, with PSCF values > 0.8. Also, the likelihood of PM2.5 from non-industrial areas increased significantly compared to the same period in 2019, indicating a significant increase in pollution from residential emissions during the lockdown. Short- and medium-range transmissions within 50–100 km decreased significantly, with only small transmissions from Yangquan, Xingtai and Baoding, with PSCF values mostly between 0.4 and 0.7. Long-range transmissions over 100 km no longer exist. During lockdown II, SJZ was closed, but the activity level in the surrounding areas was normal. As seen in Fig. 4 (c), the distribution of local sources within 50 km of SJZ was significantly reduced, indicating a low contribution of local sources to PM2.5. The higher PSCF values were only distributed near the urban center, with values between 0.5 and 0.8. The transmissions over 50 km were mainly from the east, including the transmissions from the eastern part of SJZ and Hengshui, with PSCF values reaching 0.7.

As can be seen from the discussion above, traffic mobility decreased by 52.5% and industrial activity decreased by 18.8% during lockdown I, compared to the same period in 2019, while both traffic emissions and industrial emissions decreased even more during lockdown II. Compared to lockdown I, urban traffic mobility decreased by 26.3%, industrial activity decreased by 26.2%, and fireworks emissions were completely reduced in lockdown II, resulting in a 33.8% decrease in PM2.5. This suggests that the stricter the lockdown and the lower the level of anthropogenic source activity, the greater the improvement in PM2.5.

Characteristics of the changes in O3 concentrations

From Fig. 5 (a), the average O3 concentrations on heavy and serious pollution days during these three periods were 8.6, 48.9, and 26.3 μg/m3, which increased nearly fivefold and twofold during the two lockdowns, respectively. On mild and medium pollution days, the average O3 concentration for the same period in 2019 was 16.5 μg/m3, increasing 195.2% to 48.7 μg/m3 during lockdown I and 116.4% to 35.7 μg/m3 during lockdown II. On good and excellent days, the average O3 concentrations for the three periods were 31.8, 49.4, and 45.8 μg/m3, respectively, with two lockdown periods increased by 55.3% and 44.0%, respectively. Compared with the same period in 2019, the average O3 concentration increased by 160.6% for the whole lockdown I and 108.4% for lockdown II (Fig. 5(b)), with the increases in O3 mainly concentrated on heavy and serious polluted days. There are two main reasons for the dramatic increase in O3 in lockdowns. First, PM2.5 concentrations decreased by 27.2% on average, which led to a decrease in the absorption efficiency of hydrogen peroxide radical (•HO2) for PM2.5 and an increase in peroxyl radical-mediated O3 production (Wang et al. 2020). Second, in Fig. 5 (b), the average NO concentration for both lockdowns (3.6 μg/m3) was more than 10 times lower compared to the same period in 2019 (43.7 μg/m3), and insufficient NO titration during the lockdowns leads to the accumulation of O3 (Pei et al. 2020).

Fig. 5
figure 5

(a) Temporal variation (different colors correspond to the AQI level colors in Fig. 2) and (b) average concentration of O3, NO, NO2, and NOx for three periods in SJZ

Full size image

To explain the different growths in O3 in the two lockdowns, we studied the changes in ozone-producing precursors of NOx (NOx = NO + NO2) and VOCs, respectively. Due to the occurrence of secondary pollution, from Fig. 5 (b), the mean NOx concentration (37.0 μg/m3) was slightly higher in lockdown II than in lockdown I (26.1 μg/m3), and the mean NO and NO2 concentrations were 2.6 μg/m3and 24.6 μg/m3, respectively, in lockdown I and 24.6 μg/m3 and 30.7 μg/m3, respectively, in lockdown II. Combining tropospheric NO2 column density data from TROPOMI satellite (Fig. 6 (a) and (b)), it can be found that local NO2 column density still decreased during lockdown II compared with lockdown I, especially in the southern part of SJZ where there are industrial areas, and the reduction of NOx by lockdown was positive.

Fig. 6
figure 6

The satellite image of HCHO and NO2 in SJZ urban area: (a) NO2 and (c) HCHO during lockdown I, (b) NO2 and (d) HCHO during lockdown II and (e) concentrations and OFP of overlapping species in the top ten species ranked by OFP in both lockdowns

Full size image

On the other hand, the mean concentrations of TVOCs were 42.1 and 37.3 ppbv for lockdown I and lockdown II, respectively. We calculated the ozone formation potentials (OFP) during the two lockdowns using the latest indigenous MIR values separately (Zhang et al. 2021), and eight of the top ten species with the largest OFP during the two lockdowns overlapped, as shown in Fig. 6 (e). With almost constant total concentrations for these eight species, the OFP of lockdown II decreased by 11.8%, mainly due to a 46.5% decrease in ethylene concentration. Among the above species, ethylene was the most concentrated species in lockdown I, reaching 6.45 ppbv, while in lockdown II the concentration decreased to 3.45 ppbv. Ethylene is a typical gasoline vehicle emission (Pei et al. 2022; Zhang et al. 2013), and the decrease in its concentration reaffirms the reduction in traffic sources in lockdown II. Propane increased by 45.6% in lockdown II, making it the most abundant species with an increase in OFP of 0.43 ppbv. Trans-2-butene increased more than threefold, corresponding to an increase in OFP of 2.96 ppbv and was one of the main contributors to the increase in OFP in lockdown II. Both propane and trans-2-butene are important species of volatile fuels (Lau et al. 2010), and propane is the main emission associated with liquefied petroleum gas (LPG) (Vega et al. 2000; Xuan et al. 2021). Also, tropospheric formaldehyde (HCHO) column density data were obtained, as shown in Fig. 6 (c) and (d). As the most abundant aldehyde in the atmosphere, HCHO is the most dominant volatile organic compound and pollutant in the troposphere (Peng et al. 2016), and its column concentration variation is positively correlated with energy production and vehicle emissions (Fan et al. 2021), which can be used to assess changes in human activity level. As can be seen in Fig. 6 (c) and (d), the decrease in HCHO column densities was more pronounced in the northern, eastern, and western regions of SJZ during lockdown II compared to lockdown I.

The lower O3 concentration in lockdown II may be due to a combination of a stronger NO titration reaction and a lower TVOC concentration. But which one dominates? The O3/NOx ratio can be used to qualitatively determine the sensitivity of O3 generation, and a ratio of 15 can be used as a threshold for O3 generation sensitivity conversion (Tonnesen and Dennis 2000). The O3/NOx ratios for both lockdowns and the same period in 2019 were much less than 15, and therefore the SJZ was judged to be a control area for VOCs. This is the same as the result of the previous study (Zhang 2020). Therefore, the difference in O3 concentrations between the two lockdowns in SJZ was mainly due to the different involvement of VOCs.

VOC profiles during the two lockdowns

From the above discussion, it can be found that VOCs are the key precursor of O3 generation in SJZ, and its change in the lockdown period can provide ideas for the regional management of O3. As shown in Fig. 7, alkanes were the most contributing species in both lockdowns, with their share further increasing by 37.8% in lockdown II, mainly concentrated in light alkanes (C < 6) associated with evaporation emissions of LPG and gasoline (Pei et al. 2022), consistent with the increase in propane described above. In addition, the diurnal variation in alkane concentrations was more severe in lockdown II (14.4 ~ 32.4 ppbv) than in lockdown I (16.2 ~ 19.3 ppbv), suggesting that alkane sources were characterized by heterogeneous temporal variation. Acetylene and aromatics were associated with coal combustion emissions (Barletta et al. 2005; Shi et al. 2015), whose concentrations increased by 13.6% and doubled, respectively, mainly associated with heating during lockdowns. The decrease in VOCs in lockdown II compared to lockdown I was mainly due to an 83.1% reduction in the contribution of halocarbons, which were mainly from industrial processes (Hui et al. 2018).

Fig. 7
figure 7

Changes of VOC species in (a) percentage for both lockdowns and daily concentrations in (b) lockdown I and (c) lockdown II

Full size image

The VOCs species also produced different daily variation characteristics during the two lockdowns (Fig. 7 (b) and (c)). In lockdown I, all substances showed a clear single-peak distribution with a peak at 9:00 and a trough at 17:00. In contrast, all substances in lockdown II, except alkanes, satisfied a bimodal distribution with two peak times at 9:00 and 21:00 and a trough at 16:00. According to Cheng et al., the types of VOCs detected in Chinese household cooking were alkanes (63.8% ± 4.5%), alkenes (21.9% ± 5.3%) and aromatics (14.3% ± 4.8%) (Cheng et al. 2016), and Chinese home cooking is the most common cooking cuisine in SJZ. Combined with the actual quarantine of people at home during lockdowns, the growth process of alkenes, alkynes, and aromatics overlapped with cooking time, and the contribution of cooking to VOCs may become apparent.

During the lockdown, the sources of VOCs may also change as human activity was greatly restricted. We have analyzed 25 species to identify the contribution of six emission sources to VOCs using PMF (Fig. 8).

Fig. 8
figure 8

PMF results for two lockdowns

Full size image

In lockdown I, Factor 1 explains the highest contribution of 70.5% for benzene and 53.5% for propane. Benzene is the most abundant VOC species for biomass combustion (Wang et al. 2009), and propane is also one of the major components of non-methane hydrocarbon emissions from biomass stoves in China (Liu et al. 2008; Tsai et al. 2003). Therefore, Factor 1 is identified as biomass burning.

Factor 2 is identified as a traffic source with a contribution of 27.8%, characterized by high contributions from ethane (52.0%) and ethylene (41.7%), which are typical of petrol vehicle emissions. It is worth noting that 30.6% of isoprene is also explained by this factor, and its emissions related to urban transport have been reported previously (Hellén et al. 2012; Sarkar et al. 2017).

Factor 3 is dominated by n-undecane (61.0%), acetylene (58.8%), and some aromatics. Acetylene and aromatics are important indicator species for fossil fuel (Barletta et al. 2005; Shi et al. 2015), therefore, Factor 3 is set as the coal-fired source.

Factor 4 was identified as a biogenic source because it explains more isoprene total mass than any other factor (38.8%).

Factor 5 was determined to be a fuel volatilization source, characterized by a high proportion of C4-C11 n-alkane. Evaporative emissions from China 5 and China 6 gasoline vehicles include C1-C12 alkanes, and C4-C5 alkanes account for the largest part from China 5 gasoline vehicles’ evaporative emissions (Liu et al. 2022), isopentane (42.3%) and n-pentane (41.8%), which are tracers of oil and gas volatilization.

Factor 6 was distinguished by n-heptane (43.8%), n-undecane (38.1%), and toluene (36.9%). Heavy alkanes and aromatics are the dominant VOCs in emissions of the printing industry, and toluene is the most abundant species of the total VOCs emitted from paint applications (Vega et al. 2000; Yuan et al. 2010). Factor 6 was therefore judged to be a solvent source.

In lockdown II, Factor 1 is mainly explained by alkanes and aromatics, including ethane (74.9%), propane (24.6%), toluene (6.4%), and styrene (21.4%), which fit the coal combustion VOC profile and was considered to be the source of coal combustion.

Factor 2 is mainly composed of ethylene (38.9%), acetylene (45.8%), n-butane (47.7%), isobutane (45.7%), benzene (5.3%), and p-diethyl benzene (41.3%). Ethylene and ethyne are important species emitted from biomass open burning (Fang et al. 2017) and have a similar composition to Factor 1 in lockdown I, so Factor 2 is identified as biomass burning.

For Factor 3, high contributing species include n-hexene, n-butane, isobutane, n-pentane and isopentane, which have a similar structural composition to Factor 5 in lockdown I and are identified as fuel volatile sources.

Factor 4 was identified as a traffic source, with high loads of ethylene (36.9%), acetylene (25.7%), benzene (31.4%), and toluene (38.1%), typical of petrol vehicle emissions (Pei et al. 2022; Zhang et al. 2013).

Factor 5 was interpreted as biogenic due to the contribution of 54.4% isoprene.

Factor 6 explains mainly the heavy alkanes, including 82.3% n-heptane, 27.9% n-nonane, and 26.1% n-dodecane, which are major species in printing emissions (Yuan et al. 2010). Therefore, Factor 6 was identified as a solvent source.

The proportion of the sources of VOCs in the two lockdowns is shown in Fig. 7. The ranking of source contributions in lockdown I is traffic source (32.6%) > coal combustion source (23.4%) > biogenic source (14.8%) > fuel evaporation source (12.0%) = biomass burning source (12.0%) > solvent source (5.2%); in lockdown II it is coal combustion source (35.5%) > biogenic source (22.8%) > fuel evaporation source (15.3%) > traffic source (9.9%) > solvent source (8.7%) > biomass burning (7.8%). The reduction of traffic contribution is the main reason for the decrease of VOC concentration in lockdown II, which is consistent with the above conclusion of ethylene reduction. Compared with the two lockdowns, when the restrictions of other sources of VOCs reach a certain limit, reducing the emission from traffic sources is an effective way to reduce the concentration of VOCs.

Conclusion

Since the COVID-19 lockdown significantly limits human activities and was to some extent highly similar to regional abatement activities, the two lockdowns in SJZ provide important insights for improving regional air quality.

The restrictions on human activities caused by the two lockdowns were mainly mirrored in the reduction of urban traffic mobility by 52.5% and 65.0%, respectively, and the reduction of industrial activities by 18.8% and 40.0%, respectively. These restrictions indirectly improved the regional air quality, as reflected by reductions of 13.2% and 32.4% in AQI, and 12.9% and 42.4% in PM2.5 concentrations, respectively. Comparing these two lockdowns, it can be concluded that the stricter the lockdown, the greater the air quality improvement, with the decrease in AQI and PM2.5 occurring mainly on heavy and serious polluted days and good and excellent days. Meanwhile, the PSCF results show that the PM2.5 contribution from local sources dominated, which implies that the focus of improving regional air quality is on the control of local sources.

However, due to unreasonable emission reductions during the lockdown, O3 concentrations spiked during the two lockdowns, increasing by 160.6% and 108.4%, respectively. First, the decrease in PM2.5 concentrations increased the production of O3 mediated by peroxyl radicals. Second, the average concentrations of NO decreased more than tenfold, which led to a significant reduction in the titration consumption of O3. The difference in O3 increase between the two lockdowns was mainly due to an 11.8% decrease in OFP for the major VOC species in lockdown II, especially a 3 ppbv decrease in ethylene concentrations associated with traffic sources. The results of the PMF model again confirmed that the concentration of TVOCs decreased in lockdown II due to a decrease in the contribution of traffic sources. The characteristics and sources of VOCs showed that in lockdown II alkanes increased by 37.8%, especially light alkanes (C < 6) associated with LPG, and gasoline evaporative emissions, acetylene, and aromatic hydrocarbons related to coal combustion emissions increased by 13.6% and doubled, respectively, while the contribution of halocarbons from industrial processes decreased by 83.1%. In addition, the diurnal variation characteristics of VOCs in lockdown II, except for alkanes, overlap with the time of household cooking in China. All characteristics were consistent with those of lockdown II, where stringent household policies increased emissions from cooking and coal-fired heating.

The COVID-19 lockdown results suggest that regional air quality can be improved by limiting human activities, but a blind reduction of PM2.5 and NOx may significantly increase O3 concentrations. Since SJZ is in the VOC control area, future policy should strengthen the control of VOCs, especially for traffic sources. In addition, cooking emissions also need attention.