Risk tradeoffs between nitrogen dioxide and ozone pollution during the COVID-19 lockdowns in the Greater Bay area of China.

Photochemical regime for ozone (O3) formation is complicated in the sense that reducing emission of nitrogen oxides (NOx) may increase O3 concentration. The lockdown due to COVID-19 pandemic affords a unique opportunity to use real observations to explore the O3 formation regime and the effectiveness of NOx emission control strategies. In this study, observations from ground networks during the lockdowns were used to assess spatial disparity of the Ratio of Ozone Formation (ROF) for nitrogen dioxide (NO2) reduction in the Greater Bay Area (GBA) of China. The health risk model from Air Quality Health Index (AQHI) system in Hong Kong was adopted to evaluate the risk tradeoffs between NO2 and O3. Results show that the levels of O3 increase and NO2 reduction were comparable due to high ROF values in urban areas of central GBA. The ozone reactivity to NO2 reduction gradually declined outwards from central GBA. Despite the O3 increases, the NOx emission controls reduced the Integrated Health Risk (IHR) of NO2 and O3 in most regions of the GBA. When risk coefficients from the AQHI in Canada or the global review were adopted in the risk analyses, the results are extremely encouraging because the controls of NOx emission reduced the IHR of NO2 and O3 almost everywhere in the GBA. Our results underscore the importance of using a risk-based method to assess the effectiveness of emission control measures and the overall health benefit from NOx emission controls in the GBA.

Introduction

Ozone (O3) pollution in China has become prominent over time since the initiation of Clean Air Action Plan in 2013 (B. Zhao et al., 2020; Ziemke et al., 2019). Control of O3 pollution is challenging, partly due to the non-linear connections between O3 and its precursors, such as nitrogen oxides (NOx) and volatile organic compounds (VOCs) (de Foy et al., 2020; Zavala et al., 2020). Based on the reactivity of O3 to its precursors, the photochemical regime of O3 formation can be classified into categories such as NOx-limited and VOC-limited (Zhang et al., 2022). Due to high NOx emissions, dominant O3 formation regime in urban areas of China is believed to be VOC-limited (Ding et al., 2013; Shao et al., 2009). As a result, reducing the NOx emissions in urban areas often increase ambient O3 concentrations (Lin et al., 2021; Wang et al., 2019).

At the beginning of the novel coronavirus disease 2019 (COVID-19) pandemic, China enforced the nationwide city lockdowns (Azman and Luquero, 2020; He et al., 2020), which drastically reduced pollutant emissions (Su et al., 2020). Zheng et al. (2021a) compared the pollutant emissions in China during the lockdown period with the levels in the same month in previous year, and found that the emission of SO2, NOx, and fine particle decreased by 27%, 36%, and 24%, respectively. In particular, the significant reductions in NOx emission notably increased ambient O3 concentrations in many urban areas of China (Kang et al., 2021). For instance, Liu et al. (2021) compared ambient O3 concentrations under the lockdowns with those in three weeks before the lockdowns in China and concluded that the average O3 concentrations in northern and central China increased by 112% and 73%, respectively. These drastic pollution variations are well recorded by ground observations. The COVID-19 lockdown therefore affords a unique opportunity for us to use real observations to explore the ozone formation reactivity and the potential consequences of the emission control measures.

Numerous epidemiological studies have documented the links between the air pollution exposure and increased risks of respiratory and cardiovascular diseases (Guo et al., 2021; Holm and Balmes, 2022; Zheng et al., 2021a). Given the different variations in NO2 and O3 concentration, it is unclear whether the control of NOx emission can reduce the overall health risk. The risk tradeoff between the NO2 and O3 pollution is complicated in the sense that the O3 variation largely depends on the non-linear photochemical formation and the health risk of O3 pollution can differ from that of NO2 pollution. Understanding the risk tradeoff between NO2 and O3 is essential to guide the designs of pollution control strategies that aim to protect public health (Hossain et al., 2021).

China has three major cities clusters, namely the Beijing-Tianjin-Hebei economic zone, the Yangtze River Delta region, and the Guangdong–Hong Kong–Macau Greater Bay Area (GBA). The GBA is situated in the southern coast of China. It includes nine major cities of Guangdong Province and two special administrative regions (SARs), Hong Kong and Macau. This cluster of Chinese cities was the first to urbanize due to the reform and opening-up policies in the 1980s (Zhao et al., 2021). With 59 million people reside in a small area of about 8000 km2, the GBA of China is one of the most densely populated city clusters in the world. The rapid urbanization, however, makes it one of the most polluted city clusters in China (Lin et al., 2018). Air pollution assessments in the GBA can provide clues to foresee the impacts of pollution controls in urban areas of China in the future.

A number of studies have been conducted to understand the O3 variations during the COVID-19 lockdowns in the GBA. By analyzing the O3 variations in southern China (mainly covers the GBA), Liu et al. (2021) found that O3 concentrations remained steady during the lockdowns, compared to those in three weeks before the lockdowns. Another study performed in Guangzhou, the capital of Guangdong Province, found a decline in O3 concentration from one month before the lockdown to the lockdown period (Li et al., 2022). The meteorological variations were believed to play an important role in the O3 variation during the lockdowns in Guangzhou (Shek et al., 2022).

Exploration of the spatial disparity of O3 variation during the lockdowns in the GBA is still limited. Such a regional study is essential to improving our knowledge of the O3 formation regime and developing appropriate pollution control strategies to protect public health in the GBA. In this study, observations from ground networks during the lockdowns will be used to ascertain the spatial disparity of the O3 formation regime in the GBA of China. The health risk model, which is currently used as the basis of the official Air Quality Health Index (AQHI) in Hong Kong, will be adopted to evaluate the short-term health risks of ambient NO2 and O3 (Wong et al., 2013). Then, the risk tradeoffs between NO2 and O3 will be evaluated to understand whether the NOx emission reductions effectively reduced the overall health risk during the lockdowns. Finally, characteristics of the risk-based method for evaluating the emission control measures will be discussed.

Data and methodology

Study period

In response to the COVID-19 outbreak, a series of containment measures were implemented in China. On January 23, 2020, a strict lockdown was imposed in Wuhan, the epicenter of the outbreak. As the virus spread rapidly, the Chinese New Year vacation was extended. China began resuming economic activities from the end of February. On March 11, 2020, Hubei province announced that work could resume in phases. Accordingly, the primary study period of our analysis is from January 23 to March 10, 2020, when the lockdowns were imposed over China in response to the COVID-19 outbreak (Song et al., 2021). To assess the impacts of emission reductions, air pollution levels during the same period (January 23–March 10) in 2019 and 2020 were comparatively evaluated.

Study area

The Guangdong–Hong Kong–Macau GBA of China, shown in Fig. 1 , is a large plane surrounded by elevated terrains on three sides. It includes nine major cities of Guangdong Province, including Guangzhou (GZ), Shenzhen (SZ), Foshan (FS), Zhaoqing (ZQ), Zhuhai (ZH), Jiangmen (JM), Zhongshan (ZS), Huizhou (HZ), Dongguan (DG), and two SARs, Hong Kong (HK) and Macau (MC). This city cluster was the first region in China to urbanize and thus experienced a prompt economic growth during the past few decades. However, the air pollution issues are also prominent and impose a significant threat to human health (Wu et al., 2019).

Fig. 1

Cities in the Guangdong–Hong Kong–Macau Greater Bay Area (GBA) of China. Blue points mark the locations of air quality stations. Red squares mark the locations of meteorological stations.

Air quality and meteorological data

Hourly NO2 and O3 data for the study period (i.e., January 23–March 10) in 2019 and 2020 were obtained from ground monitoring networks in Guangdong (http://www.cnemc.cn/), Hong Kong (https://www.epd.gov.hk/), and Macau (https://www.smg.gov.mo/). To take into account the effects of meteorological variations, hourly meteorological values, including temperature, relative humidity, and wind, in the GBA were acquired from the global telecommunications system of the World Meteorological Organization (WMO). The weather station that is within 50 km from an air quality station was used to monitor the meteorological values for that air quality station. As shown in Fig. 1, NO2 and O3 data at 75 air quality stations in the GBA were used in this study.

To minimize the impacts of meteorological variations, we extract the NO2 and O3 data at hours when temperature and relative humidity is within ±10 °C and ±20% from the two-year average during January 23–March 10, respectively. In addition, we only use the NO2 and O3 data under stagnant air conditions with wind speed ≤3 m/s to minimize the interaction of pollutants from different cities. The averages of valid sample size for NO2 in 2019, NO2 in 2020, O3 in 2019, and O3 in 2020 were 55.6%, 59.4%, 55.3%, and 52.3%, respectively. Then, the average NO2 and O3 concentration in 2019 and 2020 during January 23–March 10 are compared and analyzed.

Then, the changes in the average NO2 (ΔNO2) and O3 (ΔO3) concentration between 2019 and 2020 during the lockdowns are evaluated. A number of studies compared the ΔNO2 and ΔO3 during the lockdowns around the world (Fenech et al., 2021; S. Zhao et al., 2020). To explore the ozone formation regimes, the Ratio of Ozone Formation (ROF) for NO2 reduction is defined as:

If both NO2 and O3 concentrations decline in specific region, the corresponding ROF value is negative. For a region with the VOC-limited regime, the ROF value is positive because O3 concentration increases when NO2 concentration decreases.

Both NO and NO2 are important in the NOx-O3 cycle. However, NO is not a criteria air pollutant from current national air quality monitoring network. In this study, the ROF value represents the ratio of the variations in O3 and NO2 concentrations.

Health risk model

The health risk model, which is currently used as the basis of the AQHI in Hong Kong (Wong et al., 2013), is used to quantify the short-term health risks of NO2 and O3. The model estimates the percentage of added health risk (%AR) of hospital admissions for respiratory and cardiovascular diseases for different air pollutants. The concept of %AR is the same to excess risk, which estimates the excess rate of occurrence of specific health effect that is associated with air pollution exposure. Based on the %AR for different air pollutants, the AQHI system in Hong Kong determines the risk categories (including “Low”, “Moderate”, “High”, “Very High”, and “Serious”) and reports them to the public. The application of this health risk model has a long record. Some recent studies applied it to evaluate the long-term variations of air quality and health burden (Hossain et al., 2021; Tan et al., 2023). Another health risk study concluded that the AQHI system in Hong Kong helped to reduce the cardiovascular hospitalization for elderly population (Mason et al., 2020). This health risk model was also applied to evaluate the representativeness of air quality monitoring network (Hohenberger et al., 2021) and estimate the high-resolution exposure distribution (Che et al., 2020).

The %AR for specific pollutant i can be expressed as (Wong et al., 2013; Zhou et al., 2021):where denotes the concentration of specific pollutant (e.g., NO2 and O3); and denotes the risk coefficient for specific pollutant, which is derived from epidemiological studies (Orellano et al., 2020; Zheng et al., 2021b). Based on the Hong Kong AQHI system, the value for NO2 and O3 is 0.000446 (95% Confidence Interval (CI): 0.000439, 0.000459) and 0.000512 (95% CI: 0.000499, 0.000519) per μg/m3, respectively (Wong et al., 2013).

Then, the Integrated Health Risk (IHR) of NO2 and O3 pollution is calculated as:

In the present study, the IHR of NO2 and O3 pollution will be used to assess the risk tradeoffs between the NO2 reduction and O3 increase.

Results

Variations of NO2 and O3

NO2 and O3 concentrations under similar meteorological conditions in 2019 and 2020 over the GBA are assessed. Panels (a) and (b) in Fig. 2 display the spatial distribution of the average NO2 concentration in the GBA for the two years. In general, the highest NO2 concentrations were found in urban areas of central GBA (e.g., Guangzhou and Foshan) and Hong Kong, where the average NO2 concentrations exceeded 50 μg/m3 before the lockdowns in 2019. Panels (c) and (d) display the spatial distribution of the average O3 concentration in the GBA for the two years. In general, the O3 concentrations in urban areas of central GBA (20–40 μg/m3) were much lower than the other GBA regions (50–70 μg/m3). The difference in the spatial patterns of NO2 and O3 is mainly related to the NOx titration effect, with a reaction of O3 and NO to form NO2, in regions with high NOx emissions (Tang et al., 2021).

Fig. 2

Spatial distribution of average NO2 concentration in (a) 2019 and (b) 2020 over the GBA of China. Spatial distribution of average O3 concentration in (c) 2019 and (d) 2020 over the GBA of China.

Panels (a) and (b) in Fig. 3 display spatial distribution of the change in (a) NO2 and (b) O3 concentration from 2019 to 2020 over the GBA of China. Due to the controls of NOx emission, NO2 concentrations declined at most stations (i.e., 73 out of 75 stations) in the GBA. The regional average NO2 concentration decreased from 36.0 ± 14.2 μg/m3 in 2019 to 26.5 ± 12.9 μg/m3 in 2020. In particular, substantial reductions in NO2 concentrations were found in central GBA. By contrast, increases in O3 concentrations were found at most stations (i.e., 58 out of 75 stations) in the GBA. The regional average O3 concentration increased from 43.7 ± 13.2 μg/m3 in 2019 to 48.1 ± 10.4 μg/m3 in 2020, with the most substantial increases (>10 μg/m3) occurred in urban areas of central GBA. The O3 response to NO2 reduction is related to the amount of NOx reduction and the O3 formation regime. The O3 photochemical regime tends to be the VOC-limited in urban areas with high NOx emissions. Therefore, the active O3 responses to the NO2 decreases occurred near the sources of high NOx emissions in central GBA during the lockdowns.

Fig. 3

Spatial distribution of the change in (a) NO2 and (b) O3 concentration from 2019 to 2020 over the GBA of China.

We now focus on the ROF for NO2 reduction over the GBA. Fig. 4 shows spatial distribution of the ROF for NO2 reduction between 2019 and 2020 over the GBA of China. Positive ROF values were found at most stations (58 out of 75 stations) in the GBA due to the VOC-limited regime. In particular, the ROF values reached a level of 1 in urban areas of central GBA, where the NOx emissions were high. These results indicate that the NO2 reduction can produce a comparable amount of O3 in urban areas of central GBA.

Fig. 4

Spatial distribution of the Ratio of Ozone Formation (ROF) for NO2 reduction between 2019 and 2020 over the GBA of China.

The ozone reactivity to NO2 reduction for each city in the GBA is now assessed. Fig. 5 shows the change in city average of NO2 (blue bars) and O3 (green bars) concentration from 2019 to 2020 and the corresponding ROF for NO2 reduction (red bars) over the GBA of China. NO2 concentration shows a great reduction in all cities. On the city average, the NO2 reduction ranged from −6.9 ± 4.5 μg/m3 in Huizhou to −13.8 ± 5.7 μg/m3 in Zhaoqing. In contrast, the city-wide O3 concentration increased in most cities. The highest O3 increases were found for cities in central GBA (e.g., by 12.6 ± 2.1 μg/m3, 10.6 ± 3.0 μg/m3, and 8.5 ± 4.7 μg/m3 in Zhaoqing, Foshan, and Guangzhou, respectively). As a result, the ROF values were highest in central GBA (e.g., 0.91 ± 0.41, 0.88 ± 0.37, and 0.82 ± 0.53 for Zhaoqing, Foshan, and Guangzhou, respectively). These results indicate that the NO2 reduction produced a comparable amount of O3 on the city scale in central GBA and confirm that the ozone reactivity to NO2 reduction gradually declined outwards from central GBA. For instance, the city average ROF values were only 0.11 ± 0.59 and −0.02 ± 0.48 for Shenzhen and Macau, respectively.

Fig. 5

The change in city average of NO2 (blue bars) and O3 (green bars) concentration from 2019 to 2020 and the corresponding ROF for NO2 reduction (red bars) over the GBA of China. The y axis on the left gauges the NO2 and O3 concentration changes, while the y axis on the right gauges the ROF values.

Risk tradeoffs between NO2 and O3

The %AR of hospital admission for respiratory and cardiovascular diseases is calculated to understand the short-term health risks of NO2 and O3 pollution. Fig. 6 shows spatial distribution of the %AR of hospital admission for respiratory and cardiovascular diseases for (a) NO2 in 2019, (b) NO2 in 2020, (c) O3 in 2019, and (d) O3 in 2020 over the GBA of China. The spatial pattern of %AR for each pollutant is similar to the pattern of concentration. In general, the highest %AR values for NO2 were found in central GBA and Hong Kong, while the %AR values for O3 in central GBA were lower than the other GBA regions. Panels (e) and (f) show the IHR (i.e., integrated %AR) of NO2 and O3 in 2019 and 2020, respectively. Overall, the IHR for NO2 and O3 shows a north-to-south increasing gradient in the GBA.

Fig. 6

Spatial distribution of the %AR of hospital admission for respiratory and cardiovascular diseases for (a) NO2 in 2019, (b) NO2 in 2020, (c) O3 in 2019, and (d) O3 in 2020 over the GBA of China. Panels (e) and (f) show the IHR (i.e., integrated %AR) of NO2 and O3 in 2019 and 2020, respectively.

Then, the changes in the %AR of hospital admission for respiratory and cardiovascular diseases from 2019 to 2020 are quantified. Panels (a) and (b) in Fig. 7 respectively show spatial distribution of the change in %AR for NO2 and O3 from 2019 to 2020 over the GBA of China. Due to the NOx emission controls, the %AR for NO2 extensively declined, whereas the %AR for O3 extensively increased in the GBA. The most substantial changes were found in central GBA, where the %AR for NO2 decreased by 0.4–0.8 but the %AR for O3 increased by 0.4–0.8.

Fig. 7

Spatial distribution of the change in %AR of hospital admission for respiratory and cardiovascular diseases from 2019 to 2020 for (a) NO2 and (b) O3 over the GBA of China. Panel (c) shows spatial distribution of the change in the IHR (i.e., integrated %AR) of NO2 and O3 from 2019 to 2020 over the GBA of China.

Spatial distribution of the change in the IHR of NO2 and O3 from 2019 to 2020 over the GBA is shown in panel (c) of Fig. 7. The IHR of NO2 and O3 decreased at most stations (55 out of 75 stations) in the GBA. These results indicate that the control of NOx emission during the lockdowns reduced the overall health risk in most regions of the GBA, even though it increased O3 concentration to a certain extent. In particular, the IHR increased in urban areas of central GBA, where a large amount of O3 was produced due to the non-linear ozone chemistry.

The integrated health effect of NO2 and O3 for each city in the GBA was now assessed. Blue and green bars in Fig. 8 represent city average of the change in %AR of hospital admission for respiratory and cardiovascular diseases from 2019 to 2020 for NO2 and O3 over the GBA of China, respectively. The %AR for NO2 greatly reduced during the lockdowns in all cities. On the city average, the reductions in the %AR for NO2 ranged from −0.31 ± 0.20 in Huizhou to −0.62 ± 0.26 in Zhaoqing. In contrast, the %AR for O3 increased in most cities. The highest increases in the %AR for O3 were found for cities in central GBA (e.g., by 0.66 ± 0.11, 0.55 ± 0.16, and 0.44 ± 0.24 in Zhaoqing, Foshan, and Guangzhou, respectively).

Fig. 8

City average of the change in %AR of hospital admission for respiratory and cardiovascular diseases from 2019 to 2020 for NO2 (blue bars) and O3 (green bars) over the GBA of China. Red bars represent the city average of the change in the IHR (i.e., integrated %AR) of NO2 and O3 from 2019 to 2020 over the GBA of China.

Red bars in Fig. 8 represent city average of the change in the IHR (i.e., integrated %AR) of NO2 and O3 from 2019 to 2020 over the GBA of China. For cities in central GBA, the adverse effects of O3 increases on health risk were almost balanced out by the beneficial effects of NO2 reduction. The IHR varied by 0.03 ± 0.28, 0.01 ± 0.23, and −0.03 ± 0.29 for Zhaoqing, Foshan, and Guangzhou, respectively. In other cities, the NOx emission controls greatly benefited the overall public health although they increased ozone concentrations to a certain extent. For instance, the IHR declined by −0.37 ± 0.49, −0.57 ± 0.45, −0.17 ± 0.24 for Zhongshan, Macau, and Hong Kong, respectively.

The integrated health effect of NO2 and O3 is largely affected by the ROF for NO2 reduction. Increased ROF is associated with enhanced ozone formation and can thus increase the IHR of NO2 and O3. Fig. 9 shows the relationship between the change in the IHR of NO2 and O3 from 2019 to 2020 and the ROF for NO2 reduction at all stations in the GBA of China. Significant correlation coefficient of 0.89 was found. Regression relationship is identified as y = 0.51x - 0.45, indicating that the cut-off point of the ROF for reducing the overall health risk is 0.89. Most stations (55 out of 75 stations) in the GBA were in compliance with this standard and thus experienced a decreasing health risk during the lockdown period. These results confirm that control of NOx emission can benefit the public health in most regions of the GBA, even though it may increase O3 concentration to a certain extent. In some urban areas of the central GBA with large ROF values, control of NOx emission could be insufficient to protect public health. Development of synergistic control strategies for different ozone precursors (e.g., VOCs and NOx) is therefore required. These results are based on the analyses using the β coefficients from the Hong Kong's AQHI. The impacts of β coefficients will be assessed in next Section.

Fig. 9

Relationship between the change in the IHR of NO2 and O3 from 2019 to 2020 and the ROF for NO2 reduction at all stations in the GBA of China.

Impacts of the β coefficients

The β coefficients for air pollutants can be very different from one study to another. To better understand the impacts of the β coefficients, we evaluate the change in %AR using coefficients obtained from other famous studies. The AQHI in Canada reports the coefficient for daily mortality for NO2 and O3 to be 0.000871 and 0.000537 per μg/m3, respectively (Stieb et al., 2008). A global review of the relationship between the short-term exposure to air pollution and daily mortality showed that the coefficient for NO2 and O3 was 0.001292 and 0.000479 per μg/m3, respectively (Orellano et al., 2020). We then applied these coefficients to quantify the change in mortality risk in the GBA during the lockdown period. Fig. 10 shows the spatial distribution of the change in the integrated %AR of mortality for NO2 and O3 from 2019 to 2020 over the GBA of China using the β coefficients from (a) the AQHI in Canada and (b) the global review. Results show that the beneficial effects of NO2 reduction on mortality risk overwhelmed the adverse effects of O3 increase almost everywhere in the GBA. As a result, the control of NOx emission reduced the integrated mortality risk almost everywhere in the GBA. These results confirm the great necessity of NOx emission control to protect public health in the GBA.

Fig. 10

Spatial distribution of the change in the integrated %AR of mortality for NO2 and O3 from 2019 to 2020 over the GBA of China using the β coefficients from (a) the AQHI in Canada and (b) the global review.

The change in the overall health risk largely depends on the ratio between the coefficient for NO2 and O3 (i.e., ). This ratio is around 0.87 from the Hong Kong's AQHI system. It is much lower than the ratio from the AQHI in Canada (the ratio is 1.62) and the global review (the ratio is 1.67). Given the low ratio between the coefficient of NO2 and O3 from Hong Kong's AQHI, the beneficial effects of NOx emission controls on public health in most regions of the GBA are almost guaranteed.

Discussion

In this study, the observations from ground monitoring networks during the COVID-19 lockdown period were used to explore the O3 variations in the GBA of China. During the lockdowns, O3 concentrations extensively increased under the NOx emission controls over the GBA. In urban areas of central GBA, the levels of O3 increase and NO2 reduction were comparable. The ozone reactivity to NO2 reduction gradually declined outwards from central GBA. Despite the O3 increases, the NOx emission controls reduced the overall health risk of NO2 and O3 in most regions of the GBA. When the risk coefficients from the AQHI in Canada or the review in China were adopted in the risk analyses, the results are extremely encouraging because the controls of NOx emission during the lockdowns reduced the overall health risk of NO2 and O3 almost everywhere in the GBA.

This study underscores the great necessity of changing the assessment paradigm of pollution control from using concentration-based methods to using risk-based methods. Traditional evaluations of emission control measures mainly focused on the variation of pollutant concentration. Our analyses underscore the importance of using a risk-based method to comprehensively assess the effectiveness of emission control strategies that aim to protect public health. This is particularly important for the assessment of the effectiveness of the NOx emission controls, which may temporarily increase O3 concentration due to the complexity and non-linearity of O3 chemistry. The health risk model used in this study comparatively evaluated the short-term health risks of NO2 and O3 pollution. The risk tradeoffs between the NO2 and O3 pollution can be done only when the concentrations are converted to health risk. Then, the integrated health impacts of different air pollutants can be evaluated.

This study also underscores the great health benefits from stringent NOx emission reductions, although O3 concentrations may increase. In a VOC-limited region, conventional wisdom to rely on the VOCs emission control as a tool for O3 management is experiencing great challenges due to the diversity and widespread nature of the VOCs species, and especially after the low-hanging fruit has been picked (Gao et al., 2021). In the long-term strategies, the deep reduction in NOx emission and a subsequent transition to a NOx-limited regime are promising for attaining the O3 target (Ou et al., 2016).

Our results are particularly important under the context of decarbonization. NOx emission controls align closely with the carbon neutrality policies. Many human activities (e.g., combustion of fossil fuels) that produce NOx also emit CO2. Under the contexts of carbon neutrality policies and fast energy transitions, the combustion sources are expected to go through very stringent controls, which will bring substantial reductions in both carbon and NOx emissions. In the initial phase of the decarbonization, O3 concentration may increase as NOx emission reduces. The “O3 increase” is less of a problem if we take an overall health perspective.

In addition to the ROF values, the change in the overall health risk of NO2 and O3 largely depends on the ratio between the β coefficient for NO2 and O3. In this study, three sets of β coefficients were obtained from epidemiological studies in Hong Kong, Canada, and review paper. Based on the β coefficients obtained from Hong Kong, we calculated the %AR of hospital admission for respiratory and cardiovascular diseases for NO2 and O3. Based on the β coefficients obtained from Canada and review paper, we calculated the %AR of mortality for NO2 and O3. Compared to the β coefficients obtained in Canada and review paper, the ratio between the coefficient for NO2 and O3 from Hong Kong's AQHI is much lower. Overall, the beneficial effects of NOx emission controls on public health in most regions of the GBA are guaranteed.

The accuracies of β coefficients for air pollutants affect the uncertainties of the risk calculations. For instance, the β coefficients from the AQHI system in Hong Kong were calculated based on the health data in Hong Kong from 2001 to 2005. They were estimated for potentially different mixes of air pollution, compared to current pollution conditions. To minimize the uncertainties caused by the risk coefficients, three sets of β coefficients from the global review and two official AQHI systems were used and compared in this study.

The major focus of this study is the O3 responses to NOx emission reductions. Accordingly, the combined health risks of the O3 increases and NO2 reductions in ambient environment were evaluated. These analyses provide clues to anticipate the impacts of carbon neutrality policy in the future. Human exposure level is determined by both indoor and outdoor air pollution. The way human interacted with air pollution can be different between 2019 and 2020 due to the lockdowns. To understand the impacts of human behavior on the health risk, detailed information on the human activities are required. Future studies can explore the impacts of human behavior on the health risk if more data are available.

Various control measures were implemented by the Chinese government to contain the spread of the COVID-19 virus in 2020. These control efforts drastically reduced human activities, leading to unprecedented declines in pollutant emissions. The pollution variations were well recorded by ground observations. Therefore, the lockdown due to the COVID-19 pandemic affords a unique opportunity for us to use real observations to explore ozone photochemical regime and the effectiveness of the emission control measures. This is very different from modelling studies that rely on air quality models. The modelling studies require complicated model simulations and detailed emission inventories (Doumbia et al., 2021). Future study can compare the results from observation- and simulation-based analyses to generate a comprehensive picture of the O3 formation regimes and the effects of emission controls in the GBA of China.

The ground air quality observations have their limitations. First, spatial representativeness of air quality stations is often limited (Yatkin et al., 2020). Second, the majority of air quality stations are located in residential areas to monitor the public exposure levels. Due to these characteristics of monitoring network, conclusions made in this study only represent the situations in residential areas. The socioeconomic factors, such as the working locations, can greatly affect the pollution exposure level. These socioeconomic factors can change greatly over time. Future studies can analyze the detailed socioeconomic impacts if detailed and dynamic population and economic data are available.

In this study, we obtained the available meteorological data from ground monitoring network. Compared to ground measurements, satellite measurements often have a much higher spatial resolution. However, temporal resolutions of satellite measurements are often much lower than ground measurements. For instance, polar-orbiting satellites usually provide one or two measurements every day (Y. Zhao et al., 2020). In addition, the satellite data are often affected by cloud contamination and systematic biases (Bosilovich, 2006). Future studies can take high-resolution meteorological measurements into account if more measurements are available.

To minimize the impacts of meteorological variations, this study compared pollutant concentrations during the lockdowns in 2020 with the data under similar meteorological conditions in the same month in 2019. We found that O3 concentrations extensively increased in the GBA due to the controls of NOx emission. These results suggest that the dominant ozone formation regime in the GBA is VOC-limited, which is consistent to previous studies (Ding et al., 2013). It is noted that some of previous studies compared the O3 concentrations during the lockdowns with those in a few weeks before the lockdowns and found that O3 concentrations declined during the lockdowns (Shek et al., 2022). The difference in O3 variation suggests the potential impacts of meteorological variations.

Air pollution level in China experienced a drastic decline during the past decade. For instance, at the Guangzhou city monitoring station (113.26°E, 23.13°N), annual average NO2 concentration declined from 34.6 ppb in 2017 to 24.8 ppb in 2019. Given the large variation in the historical air pollution level, this study compared the pollutant concentrations in 2020 with the levels in 2019, rather than the years before 2019. To minimize the impacts of meteorological variations, we compared the pollutant concentrations under the same meteorological conditions during the same period (i.e., January 23–March 10) in 2019 and 2020. Future studies are warranted to assess the detailed impacts of meteorological factors on air pollution levels.

This study has several limitations. First, the analyses were performed only in specific season, when the lockdowns were implemented across China. However, the ozone chemistry is complicated and may vary over time. A comprehensive assessment covering an extended period is of necessity to understand the temporal variation in the ozone photochemical regime and the impacts of control measures. In addition to the evaluations of ozone concentration, our analyses also focused on the ratio of ozone increase and NO2 reduction, which is an important variable for all seasons. We found that the NOx emission controls often benefit the public health, unless the O3 increase is much larger than the NO2 decrease. The active O3 responses to the NO2 decreases tend to occur near the sources of high NOx emissions, such as the roadside areas. Second, this study is designed to focus on the risk tradeoffs between the variations of two air pollutants, namely NO2 and O3. The lockdowns affected the concentrations of other air pollutants, including particulate matter (PM), sulfur dioxide (SO2), and carbon monoxide (CO). A comprehensive evaluation of the overall health risk is warranted to include the health impacts of all air pollutants. Third, this study is designed to evaluate the short-term health risks of air pollution during a short study period. Exposure to ambient air pollution is not only associated with short-term health risks but also associated with a range of long-term health risks. Therefore, future assessments of emission controls are suggested to cover a long period to investigate the long-term health impacts of ambient air pollution.

Conclusion

This study used observations from ground networks during the COVID-19 lockdown period to explore the spatial disparity of the ozone reactivity to NO2 reduction in the GBA of China. The health risk model from the AQHI system in Hong Kong was adopted to evaluate the health risks of NO2 and O3. The risk tradeoffs between NO2 and O3 were then evaluated. Results show that the control of NOx emission extensively increased O3 concentrations in the GBA. In urban areas of central GBA, the levels of O3 increase and NO2 reduction were comparable. The ozone reactivity to NO2 reduction gradually declined outwards from central GBA. Despite the O3 increases, the NOx emission controls reduced the overall health risk of NO2 and O3 in most regions of the GBA. Our results underscore the importance of using a risk-based method to assess the effectiveness of emission control measures and the overall health benefit from NOx emission controls in the GBA.

Author contribution

Supervision: AKH Lau. Roles/Writing - original draft: CQ Lin, PKK Louie. Validation: YS Song, MH Tao. Investigation: ZB Yuan, CC Li, Y Li. Writing - review & editing: JCH Fung, Z Ning, XQ Lao.

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.