Surface ozone changes during the COVID-19 outbreak in China: An insight into the pollution characteristics and formation regimes of ozone in the cold season.

The countrywide lockdown in China during the COVID-19 pandemic provided a natural experiment to study the characteristics of surface ozone (O3). Based on statistical analysis of air quality across China before and during the lockdown, the tempo-spatial variations and site-specific formation regimes of wintertime O3 were analyzed. The results showed that the O3 pollution with concentrations higher than air quality standards could occur widely in winter, which had been aggravated by the emission reduction during the lockdown. On the national scale of China, with the significant decrease (54.03%) in NO2 level from pre-lockdown to COVID-19 lockdown, the maximum daily 8-h average concentration of O3 (MDA8h O3) increased by 39.43% from 49.05 to 64.22 μg/m3. This increase was comprehensively contributed by attenuated NOx suppression and favorable meteorological changes on O3 formation during the lockdown. As to the pollution states of different monitoring stations, surface O3 responded oppositely to the consistent decreased NO2 across China. The O3 levels were found to increase in the northern and central regions, but decrease in the southern region, where the changes in both meteorology (e.g. temperature drops) and precursors (reduced emissions) during the lockdown had diminished local O3 production. The spatial differences in NOx levels generally dictate the site-specific O3 formation regimes in winter, with NOx-titration/VOCs-sensitive regimes being dominant in northern and central China, while VOCs-sensitive/transition regimes being dominant in southern China. These findings highlight the influence of NOx saturation levels on winter O3 formation and the necessity of VOCs emission reductions on O3 pollution controls.

Introduction

Tropospheric ozone (O3) is one of the most threating air pollutants in terms of its adverse effects on human health (e.g. inducing respiratory and cardiovascular diseases) and vegetation growth (e.g. foliar injury and yield reduction) (Brunekreef and Holgate 2002; Feng et al. 2015). As a typical secondary air pollutant, O3 in the troposphere is produced mainly from two sources: (1) intrusion from stratospheric air with higher O3 concentrations; and (2) photochemical reactions of precursors including nitrogen oxides (NOx) and volatile organic compounds (VOCs) (Sillman 1999), which play dominant roles in ground-level O3 formation. Due to considerable emissions of O3 precursors from human activities and unfavorable climate conditions (e.g. stagnant hot weather), O3 pollution episodes with non-attainment days occurred frequently worldwide (Kleinman et al. 2005; Tong et al. 2017; Geddes et al. 2021). Besides, obvious increasing trends (1 ~ 5 ppb/year) of surface O3 have been observed among the developed urban areas of North America, Europe and Asia over past decades, which has posed great threats to the health of local people and regional economic development (Xie et al. 2019; Finch and Palmer 2020; Mitchell et al. 2021; Mousavinezhad et al. 2021).

Research on surface O3 pollution began in the early 1970s, and since then great efforts have been made on investigating the chemical mechanism of O3 production and loss (Guicherit and Vandop 1977; Bojkov 1986; Sillman 1999; Jacob 2000; Vingarzan 2004; Wang et al. 2017). In the troposphere, VOCs could be oxidized by the hydroxyl radical (OH) to produce peroxy radicals, which could further react with nitrogen monoxide (NO) to produce nitrogen dioxide (NO2). Photolysis of NO2 at wavelengths ≤ 424 nm could result in the production of atomic oxygen (O), which finally prompts the O3 formation via the combination reaction of atomic oxygen and molecular oxygen (O2). As a widely confirmed common feature, surface O3 formation is nonlinearly dependent on its precursors (i.e. VOCs and NOx). The relative contents of VOCs and NOx in the troposphere control the surface O3 formation regime, with O3 variation being sensitive to the less abundant air pollutants (Sillman and He 2002; Jin and Holloway 2015; Wang et al. 2017). In general, surface O3 is likely to be NOx-sensitive in remote and rural areas, where the air masses are aged with less freshly emitted air pollutants and low NOx/VOCs ratios (Spirig et al. 2002; Wang et al. 2010; Ding et al. 2013). On the contrary, it tends to be VOCs-sensitive (NOx-saturated) in urban and populous areas, where the NOx/VOCs ratios are usually high and surface O3 will decrease with reduced VOCs emissions but increase with reduced NOx emissions. In polluted urban areas with much higher levels of NO, O3 production could be strongly suppressed by NO (NO + O3 → NO2 + O2). This situation is usually referred to as the NOx-titration regime, which is also characterized by increased O3 with decreased NOx (Wang et al. 2017). The NOx-titration chemistry mainly occurs in regions characterized by net loss of surface O3, especially during winter and nighttime. This is different from VOCs-sensitive chemistry, which occurs in regions characterized by photochemical O3 production (Sillman 1999). In addition to the spatial variation, the O3 sensitivity to precursors was also found to change obviously with time owing to varied atmospheric chemical conditions. For example, Mazzuca et al. (2016) found that O3 production tended to be VOCs-sensitive in the morning and NOx-sensitive in the afternoon in Houston, America; Wang et al. (2019) found that O3 formation in eastern China had changed from VOCs-sensitive to the mixed regimes of VOCs-sensitive and NOx-sensitive from 2012 to 2016, due to significant NOx reductions (> 25%). Accurate determination of the dominant precursors for O3 formation is difficult given the tempo-spatial variability of O3 sensitivity, but it is of great significance to the development of effective O3 reduction strategies.

Besides the influence of precursors, meteorology can also play an important role in the variation of surface O3 by affecting precursor concentrations and photochemical conditions. Generally, surface O3 level is positively correlated with solar radiation and air temperature, and negatively correlated with relative humidity and wind speed (Abdul-Wahab et al. 2005; Pudasainee et al. 2006). Elevated O3 concentrations are often observed on days with strong solar radiation, high air temperature and low wind speeds, which favor the photochemical production and accumulation of O3 (Toh et al. 2013; Han et al. 2020). Meteorological effects on surface O3 could cover or even reverse the effects of precursors. For example, a recent investigation found that the long-term trend of surface O3 in eastern and central China could change from decreasing to increasing after removing the meteorological effects (Lin et al. 2021). Therefore, both meteorology and precursors should be taken into consideration before an accurate conclusion being reached on the O3 pollution mechanism and efficient pollution control measures.

As O3 levels are usually high under intense sunlight and high air temperature, most studies have focused on O3 pollution in relatively warm seasons with limited researches being carried out in winter. However, high O3 episodes could also occur during wintertime due to unfavorable climate conditions (e.g. temperature inversion) in the cold season (Edwards et al. 2014; Mansfield and Lyman 2021). With the intensification of global warming and the enforcement of pollution control policy (e.g. NOx emission reduction for winter haze mitigation), unanticipated upward trend of winter O3 concentration has been observed recently in Central Europe and East Asia (Feng et al. 2020; Gebhardt et al. 2021; Li et al. 2021a, b). During the wintertime, both human immunity and air quality usually get worse with the temperature drop (Dowell 2001; Wang et al. 2014). Elevated wintertime O3, although with relatively low ambient levels, might still pose risks to human health via both direct oxidative stress (Di et al. 2017) and indirect health effects from boosting the formation of secondary particulate matter in winter (Huang et al. 2021). Therefore, a better understanding of winter O3 formation is significant to air quality improvement and protection of human health.

In late December of 2019, a novel coronavirus disease (COVID-19) was identified, which has spread fast to become a global pandemic. In order to slow down its infection, most countries have taken social distancing measures with strict restriction on industrial production and transportation. As a consequence of the lockdown, anthropogenic emissions of primary air pollutants (e.g. NOx and VOCs) as well as their atmospheric concentrations decreased significantly (Li et al. 2020; Venter et al. 2020; Doumbia et al. 2021). This unexpected global event provided a natural experiment to investigate the response of surface O3 to emission reductions of its precursors, with a great number of researches being carried out on this subject since its outbreak. As to the related studies in China, more than 200 scientific papers on O3 pollution during the pandemic have been published till now, with most of them focusing on the O3 variation characteristics and regimes of specific locations and regions. For example, using comprehensive measurements and modeling techniques, Le et al. (2020) found that the reduced NOx during pandemic lockdown resulted in significant O3 enhancement in urban areas of Wuhan, Beijing, Shanghai and Guangzhou and the Beijing-Tianjin-Hebei (BTH) region of China, due to the nonlinear production chemistry and titration of O3 in winter; Li et al. (2020) and Huang et al. (2021) found an obvious increase in surface O3 of eastern China in response to the sharply NOx decrease during the COVID-19 lockdown, which in turn facilitated the formation of secondary particulate matter; Xu et al. (2020) investigated the impact of COVID-19 event on air quality of central China, and reported obvious increases by 12.7%, 14.3%, and 11.6% in January, February, and March, respectively, for the wintertime O3. Based on the satellite remote sensing technique, Ghahremanloo et al. (2021) found that the significant increases in surface O3 in East China from February 2019 to February 2020 are likely the result of significantly decreased NOx concentrations and less NOx saturation during the daytime. In comparison with the extensive studies on city or regional scales, very limited researches about the impact of COVID-19 pandemic on surface O3 have been done on the national scale of China. For example, based on model simulations, Zhao et al. (2020) examined the temporal − spatial variations of six air pollutants including O3 during the COVID-19 lockdown, with significant O3 increase being found caused by the nonmeteorological factors. In contrast, Liu et al. (2021) found that both meteorological changes and emission reductions contributed to the O3 increase during the COVID-19 lockdown, with O3 changes depending on the levels of NOx and VOC reductions in different regions of China. In this study, the changes of nationwide surface O3 in winter during the COVID-19 lockdown were analyzed, and the major influencing factors on surface O3 over different regions of China were evaluated. In comparison with the previous studies, which depended mainly on modeling techniques, this study was entirely based on observations and statistical analyses. A simple procedure was firstly developed in this research for removing the meteorological effects when determining the dominant precursor for O3 formation specific to each monitoring station (detailed in the Methods Section). Based on these analyses in the context of the specific event of COVID-19 pandemic, it is expected to provide useful insights into the formation mechanism and reduction measures on surface O3 of China.

Materials and methods

Study period and monitoring data

This study focused on the O3 variation from December 25, 2019 to February 22, 2020, which is divided into two time periods, i.e. one month before (December 25, 2019 ~ January 23, 2020) and one month after (January 24 ~ February 22, 2020) the initiation of the first-level public health emergency response to COVID-19 pandemic in China. Hourly data of surface O3 and NO2 at 1641 ground stations from 31 provinces were obtained from the China National Environmental Monitoring Center (http://106.37.208.233:20035/). Meteorological data including air temperature (TEMP), relative humidity (RH), pressure (P), wind direction (WD), wind speed (WS) and rainfall were acquired from National Oceanic and Atmospheric Administration (NOAA)'s National Climate Data Center archive (http://www.ncdc.noaa.gov/oa/ncdc.html).

In this study, the daily average concentrations of NO2 (DAV NO2) and the maximum daily 8-h average concentration of O3 (MDA8h O3) were calculated for statistical analyses. As NOx (= NO + NO2) was not the conventional pollution index and was unavailable to this study, NO2 was used as a surrogate marker of NOx given the close positive relationship between these two indices (Clapp and Jenkin 2001). In order to address any potential problem concerning data quality, a filtering procedure was performed as follows (Yin et al. 2019; Mousavinezhad et al. 2021): (1) all the non-positive values were eliminated; (2) the data with values greater than P75 + 1.5 × (P75–P25) or less than P25–1.5 × (P75–P25) (P25 and P75 represent the 25th and 75th percentiles, respectively) were removed as outliers; (3) only the daily statistical values were retained if more than 70% of its hourly values were valid; (4) only the data of a station were used if more than 70% of its daily statistical values were valid. After applying this criterion on data quality control, the 1641 stations were finally reduced to 1451 for further analysis in this study.

Data analyses

Statistical analysis

Independent-samples t test and non-parametric tests (Kolmogorov–Smirnov Z) on two independent samples were used to evaluate whether there are statistically significant differences between the means of air pollutants for the periods before and after the initiation of public health emergency response to COVID-19 pandemic. Results were given as the average (avg) ± the standard deviation (sd), with the differences being considered as significant if p < 0.05. Spearman correlation analysis was used to check the relationships between surface O3 and other variables (NO2 and meteorological parameters). All data statistics were performed using the statistical package SPSS 19.0 (SPSS Inc., Chicago, IL, USA).

Removing the meteorological effects

The variations of O3 with its precursors, which is indicative of the O3 formation regimes (e.g. VOCs-sensitive or NOx-sensitive), could be masked by its responses to the variations of meteorological variables (Lin et al. 2021). In order to remove the meteorological effects and accurately determine the O3 formation regimes for each monitoring site, a simple statistical procedure was developed and applied to the dataset. Firstly, the major meteorological variables which are significantly correlated with surface O3 were selected based on the correlation analysis. Secondly, for each of the dominant meteorological variables and the two air quality variables (MDA8h O3 and DAV NO2), the differences between the daily values for each of the 30 days during the COVID-19 lockdown and those for each of the 30 days during the pre-lockdown were calculated. Therefore, a total of 900 differences were obtained for each variable of each monitoring site. Thirdly, only the differences of MDA8h O3 and DAV NO2 were adopted for further analysis on O3 formation regimes if the corresponding absolute differences for the dominant meteorological variables were lower than certain critical values. In terms of this study, the critical values for the major meteorological variables, i.e. temperature, relative humidity and wind speed, were selected to be 1 °C, 5% and 1 m/s, respectively, based on the following reasons: (1) The ranges of variation for daily average temperature, relative humidity and wind speed from pre-lockdown to COVID-19 lockdown are 0 ~ 30℃, 0 ~ 68% and 0 ~ 6 m/s, respectively, during the study period. Given these big changes in meteorological variables on the nationwide scale, the relatively small meteorological fluctuations, which are defined by the selected critical values, were assumed to be unable to alter the general trend of variation of surface O3 with its precursors, especially when the precursor emissions plummeted due to the enforcement of strict lockdown measures; (2) There are enough data values left for each station for further analyses on O3 formation regime after applying the selected critical values to removing the meteorological effects on O3 variations. Finally, the average differences for MDA8h O3 and DAV NO2, which are independent of the influence of major meteorological factors, were calculated for each monitoring site.

Results and discussion

Temporal variations of surface O3 and NO2 in response to the COVID-19 lockdown

The time series of geographically (1451 stations) averaged MDA8h O3 and DAV NO2 before and during the COVID-19 lockdown in China were shown in Fig. 1, which indicated obviously increased concentration of surface O3 but decreased concentration of NO2 during the lockdown. On the national scale of China, the average value of NO2 during the COVID-19 lockdown was significantly (p < 0.001) lower than that during the pre-lockdown period (Table 1 and Fig. 2), with the decrease rate of 54.03%. This decline in the concentration of O3 precursor was mainly caused by the reductions in primary pollutant emissions due to the sharp restriction in transportation and industrial production during the pandemic in China (Doumbia et al. 2021; Zhang et al. 2021). On the contrary, the average MDA8h O3 (64.22 μg/m3) during the COVID-19 lockdown was significantly (p < 0.001) higher than that (49.05 μg/m3) before the lockdown, with the increase rate of 39.43%. This increase in O3 level was also accompanied with an hour delay (from 15:00 to 16:00) for the peak time of hourly O3 concentration (Fig. 2). As reported by Strode et al. (2019), the changing emissions of precursor gases could significantly affect the diurnal cycle of surface O3. During the COVID-19 lockdown in China, the diurnal human activities and precursor emissions on the national scale were generally postponed, which can be partly indicated by the delay (from 14:00 to 15:00) of diurnal valley values of NO2 from pre-lockdown to COVID-19 lockdown (Fig. 2). This might account for the one hour delay of the diurnal O3 peak occurrence. During the wintertime, the general NOx levels are usually high in China due to increased emissions from domestic heating and adverse diffusion conditions with lowered atmospheric boundary layer (Lin 2012). Under this situation, O3 production is usually sensitive to VOCs and is suppressed by NOx. The significant O3 suppression by NOx in cold seasons has been confirmed over most areas of China (Liu et al. 2010; Lu et al. 2019; Yang et al. 2019), which means that the O3 production is inversely related to NOx emission during the wintertime. This relationship might largely account for the significant increase in MDA8h O3 during the lockdown (Table 1), when significant decreases of NOx was observed in China. Besides the precursor influence on the overall O3 level in China, meteorological variables might also play a role in surface O3 variations during the study period. During the COVID-19 lockdown, significant increase and decrease were observed in air temperature and relative humidity, respectively (Table 1 and Fig. 1), which had probably promoted the O3 rise during the lockdown. This positive influence of meteorology on surface O3 in China during the pandemic has also been confirmed by Gaubert et al. (2021) and Yin et al. (2021).

Fig. 1

Evolution of the daily geographically averaged MDA8h O3, DAV NO2 and meteorological factors from December 25, 2019 to February 22, 2020 in China

Table 1

The statistics (avg ± std) of O3, NO2 and meteorological variables before and during the COVID-19 lockdown in China

	Pre-lockdown	COVID-19 lockdown	DIFF*	Change rate (%)
MDA8h O3 (μg/m3)	49.05(± 12.90)	64.22 (± 8.35)	15.17(± 13.86)	39.43(± 38.80)
DAV NO2 (μg/m3)	38.73(± 14.50)	17.85(± 8.56)	-20.89(± 8.44)	-54.03(± 11.38)
TEMP (°C)	2.89(± 9.18)	4.15(± 7.56)	1.26(± 2.09)	96.84(± 446.83)
RH (%)	71.43(± 11.73)	66.11(± 12.95)	-5.32(± 5.53)	-7.73(± 9.49)
P (hPa)	1025.29(± 3.95)	1025.77(± 3.34)	0.47(± 1.22)	0.05(± 0.12)
WS (m/s)	2.26(± 0.45)	2.46(± 0.51)	0.20(± 0.23)	9.37(± 10.76)
Rainfall (mm)	66.78(± 74.38)	90.76(± 97.47)	23.98(± 88.45)	312.27(± 942.98)

*DIFF: Significant differences (p < 0.001) for different variables between pre-lockdown and COVID-19 lockdown

Fig. 2

Geographically averaged (± std) hourly variations of O3 and NO2 concentrations during the periods of pre-lockdown and COVID-19 lockdown

Changes in the spatial distributions of O3 and NO2 due to the COVID-19 lockdown

For the 1451 ground stations analyzed in this study, a total of 1450 (99.93%) stations exhibited decreased NO2 in response to the COVID-19 lockdown (Fig. 3). The changes in NO2 for most of the stations were within -40 ~ -5 μg/m3 (Fig. 4). Bigger drops in NO2 (< -30 μg/m3) were found in Central East China, where higher levels of NO2 were observed during the pre-lockdown period (Fig. 3). As to the changes of surface O3, a total of 1256 stations (86.56%), which mainly located in the northern and central areas of China, showed elevated levels in MDA8h O3, with the changes for most stations in the range of 0 ~ 40 μg/m3 (Figs. 3 and 4). The stations with bigger increases (> 30 μg/m3) in surface O3 mostly located in Central East China, which is generally in accordance with the significant decreases in its precursors.

Fig. 3

Spatial distributions of averaged MDA8h O3 and DAV NO2 during the periods of pre-lockdown (a) and COVID-19 lockdown (b), and the differences (c) between the two periods

Fig. 4

Variations of station numbers with the differences of MDA8h O3 and DAV NO2 between pre-lockdown and COVID-19 lockdown

Based on the O3 ambient air quality standards (AAQS) for human health and ecological environment protection in China (MDA8h O3 of Grade I: 100 μg/m3, and Grade II: 160 μg/m3), the number of stations with at least one non-attainment day (i.e. pollution episode) over AAQS Grade I has increased from 260 (17.91%) during the pre-lockdown to 807 (55.58%) during the COVID-19 lockdown (Table 2). While for the AAQS Grade II, the number of exceeding stations has decreased from 39 (2.69%) during the pre-lockdown to 22 (1.52%) during the COVID-19 lockdown. During the period of pre-lockdown, most of the stations with non-attainment days in surface O3 mainly located in the coastal regions of southeast China (Fig. 5a). In contrast, the stations with O3 ever the AAQS spread all over China during the COVID-19 lockdown (Fig. 5b). These results indicated that: (1) O3 pollution episodes with adverse health effects could happen in a large area of China during the COVID-19 lockdown of wintertime, which has also been reported in previous study by Li et al. (2021a); (2) The spatial scope of the winter O3 pollution in China had expanded greatly from the period of pre-lockdown to that of the COVID-19 lockdown.

Table 2

Statistics on the stations exceeding the ambient air quality standards

		Number of exceeding stations	Percent of exceeding stations %	Average exceeding days	Range of exceeding days
Grade I (100 μg/m3)	Pre-lockdown	260	17.91	3.67	1 ~ 18
	COVID-19 lockdown	807	55.58	2.99	1 ~ 19
Grade II (160 μg/m3)	Pre-lockdown	39	2.69	1.23	1 ~ 2
	COVID-19 lockdown	22	1.52	1.14	1 ~ 2

Fig. 5

Spatial distribution of stations with MDA8h O3 exceeding AAQS for at least one day during the study periods of pre-lockdown (a) and COVID-19 lockdown (b)

It is worth noting that although most (86.56% in percent) stations had exhibited elevated O3 levels during the lockdown, there is still a considerable amount (195) of stations showing obvious decreased O3 concentrations (Fig. 3c), with most of them locating in the south region of China. In the meantime, the severe O3 pollution episodes (> 160 μg/m3 of Grade II AAQS), which occurred mainly in the southern coastal region of China before the lockdown, has also been reduced during the COVID-19 lockdown (Table 2 and Fig. 5). Two possible reasons might account for this tempo-spatial difference in surface O3 during the study period: (1) The NOx concentrations in winter are relatively low in the southern part of China (Fig. 3a), where no winter heating with large NOx emissions was applied due to the relatively high air temperature in winter (Fig. 6). Therefore, surface O3 was unlikely to be in the NOx-titration regime in the southern China. As reported by previous researches, surface O3 in southern coastal area of China (e.g. Pearl River Delta region) is in the VOCs-sensitive (Liu et al. 2010; Zou et al. 2015) or transitional regimes (Jin and Holloway 2015; Liu et al. 2021) in winter. This could lead to the O3 decreases with the simultaneous emission reductions of both precursors during the COVID-19 lockdown (Liu et al. 2021); (2) As suggested by the recent studies of Hu et al. (2021a) and Ma et al. (2021), meteorological factors could greatly affect the tempo-spatial variations of surface O3 in China. During the study period, most stations with decreased MDA8h O3 during the COVID-19 lockdown located in the regions characterized by decreased temperature, increased relative humidity (RH) and rainfall (Fig. 6c). These meteorological changes, which were unfavorable to atmospheric O3 formation, might comprehensively lead to the decrease in surface O3 during the COVID-19 lockdown. Similar limiting effects on regional O3 production from meteorological variables in China has also been reported by Lin et al. (2021).

Fig. 6

Spatial distributions of averaged meteorological variables during the periods of pre-lockdown (a) and COVID-19 lockdown (b), and the differences (c) between the two periods

Determination of the O3 formation regimes for different monitoring sites

As has been discussed in the previous section, both increase and decrease in O3 concentrations have been observed among the monitoring sites with the consistent decreased NO2 during the COVID-19 lockdown (Fig. 3). This indicated that the O3 sensitivities to its precursors might be different for the monitoring sites. Based on the correlation analysis, the dominant meteorological factors affecting winter O3 were identified to be air temperature, relative humidity and wind speed, which showed significant correlations (p < 0.01) with surface O3 (Table 3). Using the statistical procedure introduced in Sect. 2.2.2, the influences of these factors on O3 variations were eliminated. As shown in Fig. 7a, after removing the major meteorological effects, a total of 1166 sites (Type I) showed increased O3 concentrations with the decreases of NO2 during the lockdown period. This indicated that the majority of the ground monitoring stations had the NOx-titration or VOCs-sensitive regime of O3 formation, which is characterized by increased O3 with decreased NOx (Sillman 1999). The levels of NOx saturation for these sites were probably high, which had significantly suppressed local O3 formation during the pre-lockdown period. As reported in the recent studies on air quality changes during the COVID-19 pandemic (Doumbia et al. 2021; Gaubert et al. 2021; Hu et al. 2021b; Li et al. 2021a; Liu et al. 2021; Wang et al. 2021), besides the reduced NOx emissions, the VOC emissions as well as its ambient concentrations across China had also been reduced due to the countrywide restriction on both transportation and industrial activities. For the 1166 monitoring sites of Type I, the VOCs emission reductions during the lockdown, which could have led to the decrease in O3 levels, were probably not large enough to offset the promotion effects on O3 increase from NOx reduction. This predominant effect of NOx reduction on O3 increase has been widely confirmed by recent studies on air quality changes during the COVID-19 lockdown (Shi and Brasseur 2020; Liu et al. 2021; Yin et al. 2021), which suggests that the conventional emission reduction strategy (NOx reduction only) might be less sufficient to reduce surface O3 concentrations in China.

Table 3

Spearman correlation coefficients between surface O3 and other environmental variables

	DAV NO2	TEMP	RH	P	WS	WD	Rainfall
MDA8h O3	-0.751**	0.224*	-0.591**	0.137	0.270*	0.114	-0.209

* indicates p < 0.01; ** indicates p < 0.001

Fig. 7

Spatial distribution of stations with different trends of variation of MDA8h O3 (red circles: O3 increase ↑; blue circles: O3 decrease ↓) with decreased NO2 during the COVID-19 lockdown (a), and the box plots of DAV NO2 (b) and MDA8h O3 (c) before and during the lockdown for the stations with different O3 variations with decreased NO2

In contrast with the widely observed O3 rise in China, there is still a considerable amount of sites (223 in total, Type II) showing simultaneous decreases in both O3 and NO2 concentrations (Fig. 7a). The pre-lockdown NO2 concentrations for the sites of Type II were generally lower than those for the sites of Type I (Fig. 7b). This indicated a relatively low level of NOx saturation for the minor group of sites, which could facilitate the formation of the non-NOx-titration regimes (i.e. NOx-sensitive or VOCs-sensitive) for surface O3 production. As the simultaneous decreases in both NOx and VOCs emissions, which is the case during the COVID-19 lockdown, could comprehensively lead to the net O3 declines no matter surface O3 is sensitive to NOx, or to VOCs or in the transition state between both regimes, it is not easy to determine the precise O3 formation regimes for the 223 sites. Based on the previous studies concerning O3-NOx-VOCs chemistry (Sillman 1999; Sillman and He 2002; Peralta et al. 2021; Yang et al. 2021), it can be inferred that the NOx-sensitive regime was usually characterized by a higher ratio (> 4) of concentration changes between O3 and NOx (i.e. △O3/△NOx), and insensitivity of O3 to VOC changes alone. As the data of NOx (sum of NO and NO2) for the monitoring sites were not available to this study, two extreme scenarios with much low (10%) and high (90%) NO2/NOx ratios, which are typical of the ambient NO2 fractions for the polluted urban sites and remote clean sites, respectively, were adopted for the calculations of NOx concentrations and the △O3/△NOx ratios. As shown in Fig. 8, the △O3/△NOx ratios for the polluted scenario were lower than 0.8 in all the sites, while those for the clean scenario were lower than 4 in 96.9% of all the sites. This indicated that the O3 formation for most of the 223 sites (Type II) was probably in the VOCs-sensitive or the transition regime.

Fig. 8

Spatial distribution of stations with different △O3/△NOx ratios and the corresponding statistics in box plots (a: scenario with NO2/NOx ratio of 10%; b: scenario with NO2/NOx ratio of 90%)

Across the whole country of China, the sites with VOCs-sensitive/transition regimes of O3 formation (Type II: low level of NOx saturation) were scattered near the sites with NOx-titration/VOCs-sensitive regimes (Type I: high level of NOx saturation) (Fig. 7a), which means that the O3 sensitivities to precursors could change even within small spatial regions. This might be accounted for by the spatial difference of precursor emissions related to the urban–rural development gradient, with different levels of O3 precursors among different city areas (i.e. central, suburban and rural) (Geng et al. 2008; Wang et al. 2017). Previous studies based on ground monitoring data (e.g. Liu et al. 2021; Yang et al. 2021) and satellite retrievals (e.g. Jin and Holloway 2015) had also showed this mixed state of O3 formation regimes around China. Although varied regimes of O3 formation could be observed across China, the NOx-titration/VOCs-sensitive regimes (Type I) were predominant in northern and central China during the wintertime, while the VOCs-sensitive/transition regimes (Type II) were predominant in the southern China. The NOx concentrations were different across China with generally higher and lower levels in northern and southern regions, respectively, during the cold season (Fig. 3a). This could significantly affect the region-specific O3 responses to precursor reductions during the COVID-19 lockdown, which might account for the spatial distribution differences for varied O3 formation regimes.

Conclusions

Our results showed that the O3 pollution with concentrations higher than the Chinese AAQS could occur widely in China in the recent winter, and the pollution state had been worsened by the emission reduction and meteorological changes during the COVID-19 lockdown. On the national scale of China, surface O3 increased significantly with the drastic decrease of NO2 concentration, which could be accounted for by the reduced suppression effects of NOx as well as the favorable meteorological conditions (e.g. increased air temperature) on O3 formation during the lockdown period. As to the pollution states of different monitoring stations, with the consistent decrease in NOx, surface O3 responded oppositely for the sites in different regions of China. In contrast with the O3 increases in the northern and central China, obvious decreases in O3 levels had been observed in south China. The low NO2 concentrations in the southern part of China indicated a less or non-saturated state of NOx, which had contributed to the O3 decrease with the substantial precursor reductions during the COVID-19 lockdown. Besides, the changes in meteorology (e.g. decreased temperature and increased relative humidity) in south China during the lockdown had further promoted surface O3 declines.

After removing the major meteorological effects on O3 variations, the site-specific formation regimes of wintertime O3 in China were analyzed in this study. Most of the monitoring sites were characterized by the NOx-titration/VOCs-sensitive regimes of O3 formation, while a minority of the sites had the regimes toward the transition zone. Although the sites with different O3 formation regimes were distributed in a mixed state across the country, the sites with the NOx-titration/VOCs-sensitive regimes (high NOx-saturation level) were dominant in the northern and central China, while those with the VOCs-sensitive/transition regimes (low NOx-saturation level) were dominant in the southern China. This highlights the importance of implementing (1) VOCs emission controls throughout the country, and (2) the region-specific emission controls according to local NOx levels for effective O3 mitigation. In northern and central China, VOCs emission reductions should be strengthened to achieve the net declines of surface O3. Stringent regional joint-control within these regions might be an effective measure. While in southern China, where the NOx levels were generally low, cooperative controls on both NOx and VOCs emissions will be applicable. Given the increased winter O3 in the recent year, which might causing adverse health effects both directly and indirectly by enhancing the atmospheric oxidizing capacity (Feng et al. 2020; Huang et al. 2021; Li et al. 2021a), more attention should be paid to the pollution regimes of winter O3 in China, and the findings of our study could provide some useful insights into this issue.