Analysis of the meteorological factors affecting the short-term increase in O3 concentrations in nine global cities during COVID-19.

Surface ozone (O3) is a major air pollutant around the world. This study investigated O3 concentrations in nine cities during the Coronavirus disease 2019 (COVID-19) lockdown phases. A statistical model, named Generalized Additive Model (GAM), was also developed to assess different meteorological factors, estimate daily O3 release during COVID-19 lockdown and determine the relationship between the two. We found that: (1) Daily O3 significantly increased in all selected cities during the COVID-19 lockdown, presenting relative increases from -5.7% (in São Paulo) to 58.9% (in Guangzhou), with respect to the average value for the same period in the previous five years. (2) In the GAM model, the adjusted coefficient of determination (R2) ranged from 0.48 (Sao Paulo) to 0.84 (Rome), and it captured 51-85% of daily O3 variations. (3) Analyzing the expected O3 concentrations during the lockdown, using GAM fed by meteorological data, showed that O3 anomalies were dominantly controlled by meteorology. (4) The relevance of different meteorological variables depended on the cities. The positive O3 anomalies in Beijing, Wuhan, Guangzhou, and Delhi were mostly associated with low relative humidity and elevated maximum temperature. Low wind speed, elevated maximum temperature, and low relative humidity were the leading meteorological factors for O3 anomalies in London, Paris, and Rome. The two other cities had different leading factor combinations.

Introduction

The abrupt outbreak of the COVID-19 pandemic had significant societal and environmental impacts globally (Guzman, 2021). COVID-19 emerged and spread rapidly through the air to all parts of the world (Delikhoon et al., 2021), with the World Health Organization (WHO) labelling the outbreak a global pandemic on March 11, 2020 (https://www.who.int/). To curb the spread of virus among humans and to avoid the collapse of medical systems, most national governments implemented lockdown measures aimed at containment (Shi and Brasseur, 2020). As a consequence of the lockdown, economic activity associated with transport and mobility were nearly eliminated in many countries, and emissions from the transport and industrial sectors decreased markedly because of the significant reduction in human activities (Bauwens et al., 2020; Forster et al., 2020).

Recently, many studies have analyzed the impact of COVID-19 based on the changes in pollutants, such as particulate matter (PM) with an aerodynamic diameter <2.5 μm and 10 μm (PM2.5 and PM10), nitrogen dioxide (NO2), sulfate dioxide (SO2), carbon monoxide (CO), and ozone (O3). For example, Bauwens et al. (2020) and Forster et al. (2020) recorded the decline in PM concentrations over some major cities globally. Venter et al. (2020) found declines in the population-weighted concentration of ground-level NO2 (−60%), and PM2.5 (−31%) in 34 countries during the lockdown until May 15, 2020. Sharma et al. (2020) reported that air quality improved because of the decreased levels of PM2.5, PM10, CO, and NO2 emissions in 22 Indian cities during the lockdown phase. Y. Wang et al. (2020) linked NO2 reductions to the transportation sector in northern China, whereas the decrease in PM, CO, and SO2 emissions was linked to the industrial sector during the COVID-19 control period. Conversely, increases in near-surface O3 have been reported. Sicard et al. (2020) found an average increase of 17% in O3 in four southern European cities compared with that in the same spring period in the three previous years. Zhao et al. (2020) and Kumari and Toshniwal (2022) also reported that O3 levels increased significantly during the COVID-19 lockdown.

Here, we focus on near surface O3 changes in nine global cities during the COVID-19 lockdown. This pollutant is produced by the photochemical reaction of volatile organic compounds (VOCs) and nitrogen oxides (NOx = NO + NO2), and is enhanced by hydrogen oxide radicals (HOx = OH + peroxy radicals), which act as oxidants (Li et al., 2020). Increased near surface O3 concentrations pose a serious threat to human health (X. Wang et al., 2020). For instance, Chen et al. (2020) reported an increasing trend in O3-related mortality with an increase in O3 concentrations from 2014 to 2018. Zhang et al., 2020a, Zhang et al., 2020b presented a relationship between higher concentrations of air pollutants (increased O3 in particular) and increased risk of COVID-19 infection. Apart from human health, high O3 concentrations also induce plant cell death and yield reductions (Li et al., 2019; Xue et al., 2020). The studies mentioned above suggest that increased O3 pollution during the lockdown has become an increasing concern that underlies the reported air quality improvement (Fu et al., 2021; Ran et al., 2020).

In addition to the major influence of the anthropogenic emissions of O3 concentrations, meteorological factors play an important role in the formation, dispersion, transport, and dilution of O3. For example, Lu et al. (2019) estimated the influence of meteorology on O3 concentrations by using the GEOS-Chem chemical transport model in China. Gong et al. (2018) used the generalized additive model (GAM) to investigate the impact of changes in meteorological factors on O3 variations in 16 important Chinese cities. Specifically, COVID-19 restrictions directly led to reduced anthropogenic activities, thereby providing a unique opportunity to investigate the relationship between atmospheric pollutants, especially O3, and meteorological factors. However, to our knowledge, no controlled studies have compared the differences in the relationship between O3 concentrations and meteorological elements in various cities across the planet during the COVID-19 pandemic.

In this study, we attempt to quantify the effect of lockdowns in various cities on O3 levels in nine global cities and used GAM to examine the leading meteorological factors that affected O3 concentrations during lockdown. GAM is an effective regression model, which can flexibly handle the complex nonlinear relationships between air pollutants (e.g., PM2.5, O3, SO2, and CO) and meteorological factors. (Habeebullah, 2020; Hůnová et al., 2019; Ma et al., 2020). This study aims to describe the monthly and daily variations in O3 concentrations in nine global cities during pre- and post-lockdown periods. We also used the GAM approach to quantitatively identify the leading factors that controlled daily O3 variations in each city during the COVID-19 lockdown.

Materials and methods

Study area

In this study, we selected nine major metropolitan cities worldwide, namely, Beijing (China), Guangzhou (China), Wuhan (China), Seoul (South Korea), Delhi (India), Sao Paulo (Brazil), London (UK), Paris (France), and Rome (Italy), to investigate the impact of their lockdown on surface O3 and its relationship with meteorological factors. The primary reasons for selecting the nine cities are following. First, the selected cities have high population density, anthropogenic emissions, energy consumption, and air pollution levels. Second, the selected cities implemented strict lockdown measures during the COVID-19 pandemic. As strict lockdown significantly reduces personnel movement and outdoor activities, these cities provide an ideal experimental environment to study the impact of meteorological factors on O3. Finally, these cities also have entirely different geographical settings and climatic conditions. For example, Beijing experiences a temperate monsoon climate, whereas Delhi is semi-arid. To analyze the global impact of restriction measures on O3 and the relationship between O3 and meteorological factors, the locations were selected from different continents across the globe, as shown in Fig. 1 and Table S1. Based on the above facts, the selected cities provide suitably diverse areas for this study.

Fig. 1

Nine global metropolitan cities considered in this study.

O3 concentration data

The daily records of surface O3 concentration spread across the nine selected cities were downloaded from the World Air Quality Index (WAQI) portal (https://www.aqicn.org). The O3 concentration of each selected city is the average value of all monitoring values in that city. The data were collected from January 1 to June 30 (from 2015 to 2020) for each selected city. In addition, for each city, the daily average O3 concentrations (5-year average O3 concentrations) from January to June were also calculated based on data from 2015 to 2019, to present a baseline. We assume that the 5-year average O3 concentrations is normal, whereas O3 concentration varies in March, April, and May 2020 because of the country-specific measures undertaken during the COVID-19 lockdown. Therefore, the 5-year average O3 concentrations in March, April, and May were compared with those in 2020. Moreover, the information regarding lockdown start and end dates was collected from government reports of the selected cities, as shown in Table S1.

Meteorological data

The multi-scale interactions of meteorological conditions have a complex effect on air quality (He et al., 2017; Liu et al., 2017). The daily average meteorological data in the nine global cities from January to June (from 2015 to 2020) were obtained from the National Centers for Environmental Information reanalysis dataset (https://www.ncei.noaa.gov/data/) and used to analyze their relationship with O3 concentration. The daily average meteorological data included maximum temperature (Tmax, ℉), mean dew point (DEWP, ℉), mean visibility (VISIB, miles), mean wind speed (WDSP, knots), relative humidity (Rel.hum, %), and precipitation (PRCP, inches). A detailed description of each meteorological parameter can be found in Table S2.

Monthly relative rate of change

The monthly relative rate of change was used to compare the differences in O3 concentrations in different periods (He et al., 2021). The monthly relative rate of change was given by:where, y is the monthly relative change rate of O3 concentration in month m, x is the 5-year average O3 concentrations in m month, and x is the O3 concentration in m month in 2020.

Statistical model

GAM is a flexible and free regression model that considers the additive effects of predictors on predicted values and their non-linear relationships. This model is usually used to quantify the influence of meteorology on O3 concentration time series. In this study, we applied this statistical model, which was provided in the “mgcv package” of R software, to each city separately to characterize the relationship between the O3 concentrations and meteorological factors. The equation is as follows:where Y indicates the dependent variable; x 1, x 2., … x n are explanatory variables of Y; β is the intercept; f , f ., … f are smooth functions of the explanatory variable; and ε is the residual.

In this study, to explore the main drivers influencing O3 changes, without considering the degree of pollution caused by O3, we considered the daily average O3 concentration of each city as dependent variables, and the daily mean or daily maximum of meteorological factors as independent variables. Meanwhile, we used the Gamma distribution for mathematical rationality, as the frequency distribution of O3 concentration in most cities is not a normal distribution (Fig. S1). In addition, we used F-statistics in GAM to identify important predictors. Previous studies have shown that F-statistic is an effective indicator to identify the independent variable that is the most important. It comprehensively considers the degree of freedom (e.d.f.) and the p value of each variable. In general, the larger the F statistic, the greater the importance of the variable. The modeling process and verification process of the GAM model can be found in the supplementary information (SI).

Results

Overview variations of O3

O3 concentrations increased in January and February of 2020 in all cities, except Delhi, Rome and Sao Paulo, compared to the 5-year average O3 concentrations from 2015 to 2019. Guangzhou (62.1%) had the largest increase in average O3 concentrations in January. Whereas in February, Pairs (16.6%) and London (18.6%) exhibited the largest increases in O3 concentrations. The largest decreases in O3 levels were observed in São Paulo (−15.2%). Compared to the 5-year average O3 concentrations from 2015 to 2019, March 2020 saw significant increases in O3 concentrations in Beijing (21.5%), Wuhan (31.4%), and São Paulo (28.1%). All cities except Beijing (−10.0%) and Seoul (−2.9%) in May 2020, exhibited significant increases in O3 concentrations in April and May 2020, compared with the average concentrations from the previous 5 years. In June, Beijing, Delhi, London, Paris, Seoul, and São Paulo exhibited significant increases in O3 concentrations in 2020 compared with those in the previous 5-year average. In contrast, Guangzhou, Wuhan, and Rome exhibited decreases in O3 levels in June 2020 compared with the previous 5-year average (Table 1 ).

Table 1

The percentage change in O3 concentration from January to June 2020 compared with the previous 5-year average.

Region	January	February	March	April	May	June
Beijing	17.4%	3.2%	21.5%	20.4%	−10.0%	9.5%
Delhi	−37.5%	−18.9%	−12.9%	13.5%	35.2%	3.0%
Guangzhou	62.1%	21.7%	39.1%	50.7%	16.1%	−22.7%
London	18.6%	60.5%	38.8%	45.0%	44.5%	34.4%
Paris	16.6%	76.6%	24.3%	37.1%	23.6%	12.4%
Rome	−27.3%	−6.3%	5.3%	7.5%	12.8%	−4.0%
Sao Paulo	−14.1%	−15.2%	31.4%	27.3%	29.8%	28.8%
Seoul	11.2%	8.4%	14.1%	31.5%	−2.9%	22.0%
Wuhan	31.0%	38.8%	49.7%	33.0%	4.7%	−20.6%

Fig. 2 shows the time series of daily O3 concentrations for the nine selected cities. Prior to the COVID-19 lockdown, daily O3 concentrations were close to the average for the same period in the previous five years in São Paulo (−5.7%) and Rome (−7.6%), while they were much lower in Delhi (−22.6%) and much higher in Guangzhou (58.9%), London (36.2%) and Paris (38.0%). All selected cities exhibited a significant increase in O3 concentrations during the lockdown period. The most significant increase compared with the previous 5-year average concentrations was observed in Wuhan (49.3%), followed by London (47.5%), Paris (31.8%), São Paulo (28.9%), Guangzhou (20.1%), Delhi (20.0%), Beijing (18.4%), Seoul (17.9%), and Rome (4.8%). It is worth noting that Wuhan, London, São Paulo, and Paris are highly polluted cities where anthropocentric activities are the main source of emissions.

Fig. 2

Daily O3 concentrations in nine selected cities from January 1 to June 30. The red and blue solid lines represent the daily O3 concentrations in 2020 and the average O3 concentrations during 2015–2019, respectively. The gray rectangle indicates the lockdown period.

Impact of meteorology on O3 during the COVID-19 lockdown

The GAM model was used to fit the O3 concentrations for each city using the meteorological data listed in Table S2. The fitting effect of GAM was measured based on the adjusted R2 value. Fig. 3 shows predicted and observed O3 concentrations in the nine cities based on GAM fitting data. The results showed that the GAM model explained most of the changes in O3, the adjusted R2 ranging from 0.51 (Sao Paulo) to 0.85 (Rome). The residuals of daily O3 concentration were normally distributed, and the average residuals of all cities were close to 0 (Fig. S4). The standard deviations of the residuals were between 2.72 (Delhi) and 6.89 (Wuhan) (Table S3). In all cities, except Sao Paulo (51%), the total deviance for meteorological factors reported by GAM were over 65% (Fig. S5). These results indicated that the GAM model can explain the variation in O3 concentration well.

Fig. 3

The GAM fitting result of predicted and observed O3 concentrations in selected cities. The color bars indicate the grades of O3 concentrations.

To identify the leading meteorological factors influencing O3 variations in each city during the COVID-19 lockdown, we statistically analyzed the F statistic value of each parameter in the GAM model. Table 2 displays the F value for the three most important variables of each city during the lockdown. For example, relative humidity was the most important meteorological factor in Guangzhou and Rome as their F statistics were as high as 107.62 and 39.12, respectively; these values were also much higher than those for the second factors in these cities. Specifically, mean dew point, daily average relative humidity, and maximum temperature were the top three leading factors affecting the changes in O3 concentrations in Beijing, Wuhan, and Delhi. Mean wind speed, maximum temperature, and relative humidity were the top three leading factors affecting O3 changes during the COVID-19 lockdown in London, Paris, and Rome. The other three cities had different leading factor combinations. Daily average relative humidity, mean visibility, and mean dew point are the top three leading factors of O3 changes in Guangzhou during the COVID-19 lockdown. Mean wind speed, mean visibility, and relative humidity were the top three leading factors of O3 changes in Seoul during the COVID-19 lockdown. Maximum temperature, mean wind speed, and precipitation are the top three leading factors of O3 changes in Sao Paulo during the COVID-19 lockdown.

Table 2

Leading meteorological factors influencing O3 for all cities during the COVID-19 lockdown.

City	Rel.hum	Tmax	DEWP	VISIB	WDSP	PRCP	DOY
Beijing	13.86***	15.51***	23.73***	1.8	5.77**	0.93	17.66***
Guangzhou	107.62***	2.1*	6.62***	7.25***	1.46	4.14*	14.93***
Wuhan	6.27***	3.27*	4.61***	2.63	2.91*	0.44	13.43***
Seoul	4.15**	2.7*	3.93***	13.22***	17.18***	0.98	36.88***
Delhi	3.66**	1.67	5***	2.97*	1.42	0.32	8.13***
London	4.25*	4.32***	1.82	1.71	22.27***	–	9.66***
Paris	46.98***	9.88***	6.14***	9.81***	87.93***	4.73***	21.94***
Rome	39.12***	8.42**	3.51**	0.79	17.41***	1.28	35.55***
Sao Paulo	2.53*	9.83***	1.96	0.12	7.08***	3.26*	8.46***

Note: * indicates < 0.05, ** indicates < 0.01, *** indicates < 0.001.

To further explore the impacts of meteorological factors on O3 in each selected city during the COVID-19 period, we developed a GAM model to assess multiple meteorological variables and O3 concentration response variables. This was done to identify the smooth regression function of the meteorological variables and the influence of meteorological factors on O3 concentrations, to analyze the specific influences of meteorological factors on O3 concentrations. Fig. 4 illustrates how relative humidity, maximum temperature, mean dew point, mean visibility, mean wind speed, and precipitation impacted the O3 concentration in Wuhan during COVID-19. The effect responses for other cities are not shown for brevity. The GAM model successfully visualized the nonlinear relationship between O3 and different explanatory variables. For example, in Fig. 4, all selected meteorological factors (except mean visibility) have a nonlinear relationship with O3 concentration. Specifically, mean dew point fluctuated and decreased as O3 concentration increased. O3 concentration had a nonlinear positive correlation with the maximum temperature. When the temperature was below 55 °F (approximately 13 °C), O3 concentration decreased with increases in maximum temperature. However, when the air temperature is more than 70 °F (approximately 21 °C), O3 concentration significantly increased with increase in maximum temperature. O3 concentration was negatively correlated with relative humidity and precipitation, and decreased as relative humidity increased. In contrast, O3 concentration was positively correlated with average wind speed. When the wind speed was less than six knots (approximately 3 m/s), O3 concentration significantly increased with increase in wind speed, whereas when the wind speed was greater than six knots, the magnitude of increases in O3 concentration decreased. These results indicate that high temperature, mean dew point, low humidity and wind speed increased O3 concentration in Wuhan during the COVID-19 lockdown. The effects of meteorological factors in other cities on O3 concentrations can be found in the SI.

Fig. 4

Response curves of O3 concentrations to meteorological parameters in Wuhan. (a) Relative humidity, (b) Maximum temperature, (c) Mean dew point, (d) Mean visibility, (e) Mean wind speed, and (f) Precipitation. Spline smoothing function f(x) is presented on the vertical axis, with labels including the degrees of freedom for nonlinear smoothing. Gray shading refers to the 95% confidence interval for the response, and the lines on the X axes represent the distribution of data points.

In addition, we used the GAM results to evaluate whether the observed O3 concentration anomalies during COVID-19 lockdown were within the expected range or if the concentrations were close to those for the typical O3 changes during the same period. Fig. 5 shows the time series of O3 concentration predicted by the GAM model with all meteorological factors. By comparing the adjusted R2, we found that the GAM model displayed the variations in O3 concentration well, with relatively low standard deviation for the daily observations (Fig. 5). The dotted black lines indicate the daily concentrations of O3 in 2019 as the reference, whereas the gray rectangle indicates the COVID-19 lockdown period. We also plotted the observed concentrations (red dotted lines) and the concentrations predicted by GAMs (blue line) from January 1st, 2020 to June 30th, 2020. Through the analysis of Fig. 5 and Section 3.1, the selected cities exhibited a significant short-term increase in O3 levels during the COVID-19 lockdown period. When we predicted O3 concentration with all the meteorological data (blue) and leading meteorological data (green) in 2020, the O3 concentrations predicted by the GAM model in Delhi and Sao Paulo were lower than the values observed in 2020. The O3 concentrations predicted by the GAM model in the other cities were close to the observed value in 2020. In contrast, when the lockdown was lifted in these cities, the O3 meteorological forecast values of all cities (except Guangzhou and Wuhan) quickly became close to the reference values and remained so. The observation O3 value and meteorological forecast value in 2020 were lower than the reference values due to the continuous precipitation from June to July in Guangzhou and Wuhan.

Fig. 5

Time series of daily O3 concentrations in nine global cities. The time series included correspond to observations made during the reference period in 2019 (black dot), observations from January 1, 2020 to June 30, 2020 (red dot), GAM predictions using the daily meteorology of that period (blue), and GAM predictions for the same period after replacing leading meteorological factors (green).

Fig. 2, Fig. 5 also show that the observed O3 concentrations in Wuhan, London, and Paris were considerably higher than those in the six other cities during the COVID-19 period. Specifically, during the intermediate stage of the lockdown, the O3 concentrations in these cities were much higher than those in the summer of the reference period. By predicting the O3 concentrations in Wuhan, London, and Paris using all the meteorological data and leading meteorological data and comparing them to the O3 concentrations in the same period in 2019, we found that the predicted O3 concentrations under the two conditions remained higher than the reference values. This indicated that the meteorological conditions during the COVID-19 lockdown favored the O3 production. During the COVID-19 lockdown, Delhi and Sao Paulo also observed significant increases in O3 concentrations. When we predicted the O3 concentrations of Delhi and Sao Paulo for the COVID-19 lockdown period using meteorological data from 2020, we found that the predicted values for the two cities were lower than the observed values. However, when the lockdown was lifted in both cities, the predicted and observed values of O3 concentrations gradually approached the reference values, indicating that meteorological factors had a certain impact on the O3 concentrations in Delhi and Sao Paulo during the COVID-19 lockdown. Human emissions and other factors also played important roles.

Discussion

In this study, the temporal variation of O3 concentration and influence of meteorological factors in nine cities around the world were analyzed for the period encompassing the COVID-19 lockdown. We determined that the concentration of O3 significantly increased during the COVID-19 lockdown period in 2020 compared with the O3 levels measured during the same period in 2019, as confirmed in previous studies (Kumari and Toshniwal, 2022; Shi and Brasseur, 2020; Venter et al., 2020a). The increase in O3 concentration can be attributed to the following reasons.

Reduction of O3 precursors

Previous studies revealed that O3 is a secondary pollutant, and its concentration depends on the local availability of its precursors (Xue et al., 2020). In urban areas, O3 formation depends on the VOC:NOx ratio (Pusede and Cohen, 2012). In general, urban areas are characterized by low ratios because of high NOx concentrations. For example, Zeng et al. (2018) and Anav et al. (2019) reported that high concentrations of VOCs were observed on days when high O3 levels were reported in Wuhan and in Southern Europe. Thus, a reduction in VOC emission would reduce O3 formation, but a reduction in NOx emission would increase O3 formation (Eqs. (1), (2), (3), (4), (5), (6), (7)).

During the COVID-19 lockdown, global NO2 concentrations dropped significantly (Venter et al., 2020a). Sicard et al. (2020) found that the daily average NO2 concentrations in Nice, Rome, Turin, Valencia, and Wuhan decreased by 62.8%, 45.6%, 30.4%, 69.0%, and 57.2%, respectively, compared with the baseline conditions (2017–2019). Similarly, Ordóñez et al. (2020) revealed that the daily maximum NO2 decreased consistently across Europe, with reductions ranging from 5% to 55% for the same period in 2015–2019 for 80% of the sites considered. In addition, Delhi (−60.0%) (Kumari and Toshniwal, 2022), Sao Paulo (−54.3%) (Nakada and Urban, 2020), Beijing (−33.3%) (He et al., 2020), Seoul (43.0%), and Guangzhou (30.0%) (Bauwens et al., 2020) also exhibited significant decreases in NO2 concentrations. These phenomena suggest that, in the investigated cities, strict lockdown measures led to a reduction in NOx emissions, leading to higher O3 concentrations (Zhang et al., 2020a, Zhang et al., 2020b). In addition, the increased O3 concentrations in European cities can be explained by the reduced road traffic emissions owing to strict lockdown measures as the overall contribution of road traffic emissions to O3 levels is approximately 12–35% in some European cities (Mertens et al., 2019).

Effects of leading meteorological factors

In this study, we used GAM to analyze the relationships between O3 concentrations and meteorological factors in nine global cities. The results of the GAM model revealed that the maximum temperature, relative humidity, dew point, and mean wind speed played important roles in the rise in O3 levels during the COVID-19 lockdown period in all considered cities. Moreover, the O3 concentrations in nine global cities had significantly positive (negative) correlations with meteorological factors during COVID-19. Many studies have reported that meteorological factors are the main driving factors for daily changes in air quality (He et al., 2017; Wang et al., 2017; Zhao et al., 2016). Our study shows that the maximum temperature had a significantly nonlinear relationship with O3 concentration and a significantly positive effect on O3 concentration. During the COVID-19 lockdown, O3 concentration increased as the air temperature increased in selected cities. This finding is consistent with the results from previous studies (Li et al., 2020; Lia et al., 2020; Y. Wang et al., 2020). An increase in the air temperature is a major reason for an increase in the O3 concentration. For example, rising air temperatures lead to increases in O3 concentrations in Europe (Hůnová et al., 2019) and in China (Sun et al., 2019).

Relative humidity was found to have a significantly negative effect on O3 concentration in the selected cities during the COVID-19 lockdown. We found that O3 concentration decreased linearly (Wuhan, Beijing, Seoul, Delhi, and Sao Paulo) or nonlinearly (Guangzhou, London, Paris and Rome) with increase in relative humidity. These results are consistent with studies in Europe (Mertens et al., 2019; Ordóñez et al., 2020), Brazil (Siciliano et al., 2020), and the South China (Fu et al., 2021). Previous studies have shown that water vapor can not only absorb and release energy through the change of water phase but also react with O3 (Ma et al., 2020; Wang et al., 2019), resulting in a decrease in O3 concentration. In addition, air humidity is usually positively correlated with the amount of cloud cover. Increase in humidity indicates an increase in cloud cover area. Simultaneously, the accompanying water vapor can decrease solar ultraviolet radiation, thus affecting photochemical reactions and the O3 concentration (Huang et al., 2019; Yin et al., 2019). Interestingly, however, the average relative humidity was less than 50% in cities where O3 concentrations changed significantly during the COVID-19 lockdown, such as Wuhan, London, and Paris. Li et al. (2018) and Wang et al. (2019) pointed out that when relative humidity was lower than 50%, O3 concentration increased with relative humidity; O3 concentration could also reach a peak in the range of 50%–60% relative humidity. This may be one of the reasons why O3 concentrations in these cities rose rapidly during the COVID-19 lockdown period.

Generally, wind speed tends to have a significantly negative effect on O3 concentration (He et al., 2017; Liu et al., 2017). When the wind speed is high, O3 and its precursors can be easily removed, thereby reducing O3 accumulation (Li et al., 2020). However, O3 concentrations increased as wind speed increased in all the selected cities during the COVID-19 lockdown. Specifically, in Asian cities, such as Wuhan, Beijing, Guangzhou, Delhi, and Seoul, O3 concentration had a significant nonlinear relationship with average wind speed, and the influence of wind speed on O3 concentration was less than that of other meteorological factors (e.g., temperature and relative humidity). This may be caused by the typical seasonal weather conditions in these cities involved a mean wind speed that was mostly lower than five knots (about 2.5 m/s). Liu et al. (2020) pointed out that the possibility of removing environmental pollutants is extremely small at low wind speed. Meanwhile, low wind speeds are not conducive to the horizontal dispersion and vertical transport of O3 and reduce the downward transport of O3 pollutants at high altitudes (Zhang et al., 2018). In addition, previous studies have pointed out that lower wind speeds reduce the amount of ozone mixed with urban air (Ma et al., 2020; Yi et al., 2015). As a result, the increase in wind speed was beneficial for the increase in O3 concentration. Similar results have also been reported in other global cities, such as Shanghai (Gu et al., 2020), central eastern China (Sun et al., 2019), and Barcelona (Tobías et al., 2020). In contrast, in London, Paris, and Rome, O3 concentration increased linearly as mean wind speed increased. Furthermore, the mean wind speed was an important meteorological factor for the increase of O3 concentration in London, Paris, and Rome during the period of COVID-19. This may be because of the reduced zonal airflow at low levels in many European cities (Ordóñez et al., 2020).

Conclusion

This study investigates the impact of the COVID-19 lockdown on O3 concentration in nine cities across the globe. The GAM approach was used to identify the leading factors that influenced daily O3 variations in the investigated cities during the COVID-19 lockdown. We found that the O3 concentration exhibited a remarkable increase during the post-lockdown period compared with that during the pre-lockdown period for all selected locations. Furthermore, meteorological effects contributed to increasing O3 concentrations in all selected cities, accounting for a large proportion of all observed changes. However, the main meteorological variables driving O3 anomalies varied with geographical location. They were dominated by relative humidity and maximum temperature in Beijing, Wuhan, Guangzhou, and Delhi; maximum temperature and mean wind speed in Sao Paulo; mean wind speed, maximum temperature, and relative humidity in London, Paris, and Rome.

Thus, our study only discusses the changes in O3 concentrations and the meteorological factors driving these changes before and after the COVID-19 lockdown in nine cities around the world. Furthermore, this analysis is equally applicable to other countries and regions around the globe. Our findings can be used to predict O3 concentrations and variation patterns based on meteorological factors and human activity levels, leading to policies to reduce O3 pollution and improve public health.

Credit author statement

Yunzhang Li, Zhongsong Bi, and Chao He conceptualized the idea, designed the research, and wrote the paper; Zhongsong Bi and Chao He analyzed the data and interpreted the results; all authors discussed the results and revised the manuscript.

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.