Ambient Ozone and Fine Particular Matter Pollution in a Megacity in South China: Trends, Concurrent Pollution, and Health Risk Assessment

: Over the past several years, Shenzhen’s air quality has signiﬁcantly improved despite increased ground-level ozone (O 3 ) and the challenges in reducing ﬁne particulate matter (PM 2.5 ). We investigated concentration trends, concurrent pollution features, and long-term exposure health risks to enhance our understanding of the characteristics of O 3 and PM 2.5 pollution. From 2016 to 2022, there was a decrease in PM 2.5 levels, but an increase in O 3 . Additionally, the premature mortality attributed to long-term air pollution exposure decreased by 20.1%. High-O 3 -and-PM 2.5 days were deﬁned as those when the MDA8 O 3 ≥ 160 µ g m –3 and PM 2.5 ≥ 35 µ g m –3 . Signiﬁcantly higher levels of O 3 , PM 2.5 , nitrogen dioxide (NO 2 ), O X (O X = O 3 + NO 2 ), and sulfur dioxide (SO 2 ) were observed on high-O 3 -and-PM 2.5 days. Vehicle emissions were identiﬁed as the primary anthropogenic sources of volatile organic compounds (VOCs), contributing the most to VOCs (58.4 ± 1.3%), O 3 formation (45.3 ± 0.6%), and PM 2.5 formation (46.6 ± 0.4%). Cities in Guangdong Province around Shenzhen were identiﬁed as major potential source regions of O 3 and PM 2.5 during high-O 3 -and-PM 2.5 days. These ﬁndings will be valuable in developing simultaneous pollution control strategies for PM 2.5 and O 3 in Shenzhen.


Introduction
Over the past few decades, there has been a remarkable increase in surface ozone (O 3 ) levels in East Asia, particularly in China [1][2][3].Despite rapid reductions in fine particulate matter (PM 2.5 ) levels [4,5], PM 2.5 -related hazards remain significant due to adverse impacts on visibility, climate change, and human health [6][7][8].Tropospheric aerosols play an important role in cooling the climate system by reflecting solar radiation and enhancing cloud reflection [9].Long-term or short-term exposure to O 3 and PM 2.5 has a detrimental effect on human health, e.g., mortality from cardiovascular and respiratory diseases [10][11][12][13].The Global Burden of Disease Study in 2019 reported that air pollution has emerged as the fourth leading cause of death worldwide [14].The premature deaths related to long-term exposure to O 3 and PM 2.5 were reported to be 1.39 million and 147.7 thousand, respectively, in 2020 across China [15].Given the serious O 3 and PM 2.5 pollution issues in China [16,17], it is essential to identify the causes of pollution for O 3 and PM 2.5 to implement coordinated prevention and control measures.
Tropospheric O 3 is recognized as one of the most important products formed via the photochemical oxidation of volatile organic compounds (VOCs), catalyzed via nitrogen oxides (NO X ) and hydrogen oxide radicals (HO X ) in the presence of sunlight [18,19].Ambient PM 2.5 has been shown to affect O 3 production by attenuating solar radiation and HO X radical uptake [20,21].Shao et al. [22] demonstrated a 37% increase in O 3 production attributed to reduced PM 2.5 in Beijing from 2006 to 2016.Nevertheless, other researchers reported that the increase in O 3 elevates the atmospheric oxidation capacity, which promotes the formation of secondary PM 2.5 [23,24].Despite the different formation mechanisms of ambient O 3 and PM 2.5 and their complex relationships, the same precursors, including VOCs and nitrogen dioxide (NO 2 ), make it possible to establish mitigation strategies for both O 3 and PM 2. 5 .
Shenzhen is a coastal city located in the southern region of China, with a population of more than 17 million.The region is influenced by a subtropical monsoon climate, characterized by relatively high temperatures throughout the year.As the central city of the Greater Bay Area, Shenzhen ranks as the third-largest city in China for gross domestic product (GDP).In recent years, the air quality in this area has significantly improved.The annual average concentration of PM 2.5 was 16 µg m −3 in 2022, close to the World Health Organization (WHO)'s recommended first-stage interim target of 15 µg m −3 .However, O 3 concentrations in Shenzhen have not shown declining trends, and the levels of O 3 and PM 2.5 are still significantly higher than the WHO's air quality guideline (AQG) levels [16,25].This situation highlights the coordinated control of O 3 and PM 2.5 pollution, which will help improve air quality and reduce environmental health risks.
In this study, the pollution characteristics of O 3 and PM 2.5 in Shenzhen were investigated.This study quantified the trends, correlations, and premature mortality associated with long-term exposure to ambient O 3 and PM 2.5 from 2016 to 2022.The influence of the meteorological conditions on the high-O 3 -and-PM 2.5 days was analyzed.The positive matrix factorization (PMF) model and the TrajStat model were applied to determine the contributions of VOC sources and potential source regions of O 3 and PM 2.5 during high-O 3 -and-PM 2.5 days, respectively.The outcomes are expected to assist local governments in formulating effective strategies for controlling O 3 and PM 2.5 .These findings also have implications for other international coastal regions that seek to further improve air quality.

Data Collection
The locations of the sampling sites are depicted in Figure 1.The 14 sites were all ambient air quality monitoring stations where data on O 3 , PM 2.5 , NO 2 , sulfur dioxide (SO 2 ), and carbon monoxide (CO) were collected.Meteorological parameters, such as temperature, relative humidity, air pressure, wind speed, and wind direction, were collected from the meteorological monitoring stations situated near the sampling sites.The details of the 14 sites are presented in Table S1.We have defined 13 sites, excluding the Lianhua (LH) site, as basic air quality monitoring stations (BAQMS).Hourly O 3 , PM 2.5 , NO 2 , and SO 2 concentrations were monitored at 13 BAQMS from 2016 to 2022.In accordance with the National Ambient Air Quality Standards (NAAQS) of China [26], the reference state for O 3 , NO 2 , and SO 2 was standardized at 273 K and 101.325 kPa, while the concentrations of PM 2.5 were reported based on real-time temperature and pressure.At the LH site, 51 VOC species were continuously measured using an online gas chromatograph with a 1 h time resolution (Air-moVOC GC-866, CHROMATE-SUD, France) from 2019 to 2022.The LH site is a typical urban site located in the center of Shenzhen, which is surrounded by parkland, residential and commercial blocks.This site was also selected as a typical urban site in previous studies in Shenzhen [27][28][29].

Data Analysis
In this study, the levels of O3, PM2.5, NO2, SO2, and meteorological parameters in Shenzhen were determined using data from 13 BAQMS.Data from the LH site were used for source apportionment and to analyze the precursor VOCs' impact on O3 and PM2.5 formations.The months of March to May, June to August, September to November, and December to February were defined as spring, summer, autumn, and winter, respectively.We defined the high-O3-and-PM2.5 days as those with daily maximum 8 h average (MDA8) O3 concentrations ≥ 160 µg m −3 and PM2.5 daily average concentrations ≥ 35 µg m −3 to investigate the concurrent pollution of O3 and PM2.5.Similarly, high-O3 days were categorized as days MDA8 O3 levels ≥ 160 µg m −3 and daily averages of PM2.5 < 35 µg m −3 .The statistical analyses were conducted using the SPSS Statistics 26.0 software package.Since the concentrations of the air pollutants did not satisfy the normality distribution assumption, the correlation analysis was conducted using the Spearman correlation coefficient method.

Source Contribution Investigation
The US Environmental Protection Agency (EPA) positive matrix factorization model (PMF, version 5.0) was applied to investigate the emission sources of VOCs at the LH site, a representative urban site in central Shenzhen.The PMF model is a multivariate factor analysis tool that decomposes a matrix of speciated sample data into two matrices, i.e., factor contribution and factor profile decomposition [30,31].Next, the sources of VOCs in the specific sites can be identified.This model has been frequently used for the source apportionment of ambient VOCs, and detailed introductions and applications of the PMF model can be found in previous studies [32,33].In this study, data collected from 23 high-O3-and-PM2.5 days at the LH site were applied to the PMF model.A total of 19 non-methane hydrocarbons (NMHCs) and one trace gas (CO) were input into the model, and the uncertainties for each species were determined by summing 10% of the VOC concentrations [33].Values below the method detection limit (MDL) were replaced by half the MDL, and the uncertainties were set at 5/6 of the MDL [34][35][36].
To assess the influence of anthropogenic sources on the formation of O3 and secondary organic aerosol (SOA), the maximum incremental reactivity (MIR) method [37,38] and the fractional aerosol coefficient (FAC) method [39,40] were utilized to compute the ozone formation potentials (OFPs) and secondary organic aerosol formation potentials (SOAFPs) of all VOC species, respectively.OFPs and SOAFPs can be calculated using the following equations:

Data Analysis
In this study, the levels of O 3 , PM 2.5 , NO 2 , SO 2 , and meteorological parameters in Shenzhen were determined using data from 13 BAQMS.Data from the LH site were used for source apportionment and to analyze the precursor VOCs' impact on O 3 and PM 2.5 formations.The months of March to May, June to August, September to November, and December to February were defined as spring, summer, autumn, and winter, respectively.We defined the high-O 3 -and-PM 2.5 days as those with daily maximum 8 h average (MDA8) O 3 concentrations ≥ 160 µg m −3 and PM 2.5 daily average concentrations ≥ 35 µg m −3 to investigate the concurrent pollution of O 3 and PM 2.5 .Similarly, high-O 3 days were categorized as days MDA8 O 3 levels ≥ 160 µg m −3 and daily averages of PM 2.5 < 35 µg m −3 .The statistical analyses were conducted using the SPSS Statistics 26.0 software package.Since the concentrations of the air pollutants did not satisfy the normality distribution assumption, the correlation analysis was conducted using the Spearman correlation coefficient method.

Source Contribution Investigation
The US Environmental Protection Agency (EPA) positive matrix factorization model (PMF, version 5.0) was applied to investigate the emission sources of VOCs at the LH site, a representative urban site in central Shenzhen.The PMF model is a multivariate factor analysis tool that decomposes a matrix of speciated sample data into two matrices, i.e., factor contribution and factor profile decomposition [30,31].Next, the sources of VOCs in the specific sites can be identified.This model has been frequently used for the source apportionment of ambient VOCs, and detailed introductions and applications of the PMF model can be found in previous studies [32,33].In this study, data collected from 23 high-O 3 -and-PM 2.5 days at the LH site were applied to the PMF model.A total of 19 non-methane hydrocarbons (NMHCs) and one trace gas (CO) were input into the model, and the uncertainties for each species were determined by summing 10% of the VOC concentrations [33].Values below the method detection limit (MDL) were replaced by half the MDL, and the uncertainties were set at 5/6 of the MDL [34][35][36].
To assess the influence of anthropogenic sources on the formation of O 3 and secondary organic aerosol (SOA), the maximum incremental reactivity (MIR) method [37,38] and the fractional aerosol coefficient (FAC) method [39,40] were utilized to compute the ozone formation potentials (OFPs) and secondary organic aerosol formation potentials (SOAFPs) of all VOC species, respectively.OFPs and SOAFPs can be calculated using the following equations: where VOC i represents the concentration of VOC species i (µg m −3 ).MIR i denotes the MIR coefficient of species i [37].FAC i represents the FAC of species i [39,40].In this study, the OFPs and SOAFPs of 19 VOC species used as input in the PMF model were calculated.

Identification of Potential Source Regions
We conducted simulations of the backward trajectories of air masses and identified the potential source regions of O 3 and PM 2.5 using a GIS-based model TrajStat to examine the impact of regional transport [41].This model has been extensively used in previous studies [42][43][44].The meteorological data used for the model were obtained from the global data assimilation system (GDAS) dataset (available at https://www.ready.noaa.gov/archives.php,accessed on 5 December 2023).The LH site, situated in the central area of Shenzhen, was selected as the target location for the backward trajectory study.A total of 58 high-O 3 -and-PM 2.5 days were identified during the study period in Shenzhen.The model was run in a 24 h backward mode at a height of 200 m with a 1 h interval on high-O 3 -and-PM 2.5 days during 2016 and 2022.Considering that the back trajectories at different heights below 1000 m did not differ significantly [45], the height of 200 m was chosen to reduce the effects of surface friction and to represent the concentrations of well-mixed air pollutants [46].In previous studies, the height of 200 m has been widely used to investigate the back trajectories of air masses and potential source regions of air pollutants [46][47][48].
The potential source contribution function (PSCF) model combined with backward trajectories was applied to identify potential O 3 and PM 2.5 source regions.The study area was divided into i × j grids, and the PSCF value of grid ij can be calculated using Equation (3).
where n ij represents the number of endpoints within the ij grid, and m ij is the number of endpoints exceeding the pollutant concentration threshold.In total, there were 1392 trajectories and 34,800 endpoints.The area covered by the trajectories was divided into 3599 grids with a resolution of 0.15 × 0.15 • .The O 3 and PM 2.5 thresholds were set at 115.7 µg m -3 and 44.0 µg m −3 , respectively.These values represent the average O 3 and PM 2.5 concentrations during high-O 3 -and-PM 2.5 days.Since the PSCF model only indicates the distribution of pollution trajectories within a grid and is unable to demonstrate the actual pollution levels of specific trajectories, the concentration-weighted trajectory (CWT) model is utilized to attach weights to the trajectories based on their associated concentrations.The CWT is calculated using Equation (4).
where C ij denotes the average weight concentration of trajectory a in the ij grid.C a is the concentration of trajectory a, and b represents the total number of trajectories.t ija is the time that trajectory a remains in the ij grid.
If the value of n ij is lower than three times the average number of endpoints for each grid n ave , the weighting function W ij should be applied to reduce uncertainty by multiplying the PSCF and CWT.The W ij is calculated using Equation (5).

Health Risk Assessment
This study utilized the Poisson regression relative risk model based on epidemiological research to estimate the premature mortality associated with long-term exposure to O 3 and PM 2.5 .All-cause mortality, as well as cardiovascular and respiratory mortality, was calculated using Equation ( 6) [2,49].
where ∆M i,y represents the premature mortality of health endpoint i attributable to longterm exposure to O 3 or PM 2.5 in year y.P y is the population of Shenzhen in year y; I i is the baseline mortality rate for health endpoint i in year y.βi denotes the long-term exposure-response coefficient (ERCs) of O 3 or PM 2.5 for health endpoint i (as shown in Table S2).It is defined as the percentage change in mortality for every 10 µg m −3 increase in O 3 or PM 2.5 concentration.C y is the exposure concentration of O 3 or PM 2.5 in year y.C 0 refers to the baseline concentration of O 3 or PM 2.5 , which is considered a safe level with relatively low health risks.In line with the WHO Global Air Quality Guidelines [50], the long-term exposure scenario for O 3 is defined as the peak season concentration.This is specifically determined by calculating the average of the MDA8 O 3 concentration over six consecutive months with the highest six-month running average O 3 concentration.The exposure scenario for PM 2.5 is defined by the annual average concentration.The baseline concentrations of peak season MDA8 O 3 and annual PM 2.5 are assumed to be 60 µg m −3 and 5 µg m −3 , respectively.The economic loss caused by premature mortality was assessed using the value of a statistical life (VSL) method, which is commonly applied to evaluate the health economic loss resulting from air pollution [51][52][53].Due to the unavailability of VSL in Shenzhen, the VSL from Chongqing [54] was used to assess the economic loss.This study utilizes per capita disposable income (PCDI) to calculate the VSL in Shenzhen, considering the influence of residents' income [55].The equation is as follows [56]: where VSL y represents the VSL in Shenzhen in year y, and VSL base indicates the VSL of Chongqing in 2018, which is USD 3.88 million [54].I base denotes the PCDI of the benchmark city in the base year, i.e., the PCDI of Chongqing in 2018 (available at http: //tjj.cq.gov.cn/zwgk_233/tjnj/2019/indexch.htm,accessed on 5 December 2023).I y is the PCDI of Shenzhen in year y.β E is the income elasticity coefficient, which the Organization for Economic Cooperation and Development recommends to equal 0.8 [57].The population and PCDI data in Shenzhen for each year were obtained from the Shenzhen Statistical Yearbook (available at http://tjj.sz.gov.cn/zwgk/zfxxgkml/tjsj/tjnj/,accessed on 5 December 2023).Mortality rates related to various diseases were provided by the Health Commission of Shenzhen Municipality (available at http://wjw.sz.gov.cn/gkmlpt/index#2504,accessed on 5 December 2023).

Trends of Air Pollutants
Figure 2 illustrates the trends of the daily average PM 2.5 , MDA8 O 3 , and NO 2 concentrations in Shenzhen from 2016 to 2022.It was found that the PM 2.5 levels decreased at a rate of −1.7 µg m −3 year −1 .The concentrations of PM 2.5 precursors (Figures S1 and S2), namely NO 2 , SO 2 , and VOCs, decreased at a rate of 1.5 µg m −3 year −1 , 0.3 µg m −3 year −1 , and 0.3 µg m −3 year −1 , respectively.A comparable decline was also observed in Beijing, Shanghai, Hong Kong, and other cities in China [15,[58][59][60], indicating the efficacy of the air pollution control measures.It is worth noting that the global spread of COVID-19 since 2019 has had a significant impact on anthropogenic activities, especially in China.Decreased industrial and transportation activities result in a reduction in air pollutant emissions, including NO 2 , PM 2.5 , SO 2 , and VOCs [61-64].Therefore, the impact of the COVID-19 pandemic on ambient air pollutant levels from 2019 to 2022 cannot be ignored.In contrast to PM 2.5 , the concentration of MDA8 O 3 increased by 1.5 µg m −3 year −1 .This finding is consistent with previous studies that have reported a significant increase in O 3 levels in urban areas in China [2,53,65], highlighting the importance of controlling ozone pollution.O X (O X = O 3 + NO 2 ) can be used to describe the atmospheric oxidation capacity of urban areas [66].O X levels decreased at a rate of −1.5 µg m −3 year −1 (Figure S1), indicating a decline in atmospheric oxidation capacity.Given the increasing trend in O 3 concentrations, the decrease in O X concentrations is primally attributed to reduced NO 2 concentrations.
namely NO2, SO2, and VOCs, decreased at a rate of 1.5 µg m year , 0.3 µg m year , a 0.3 µg m −3 year −1 , respectively.A comparable decline was also observed in Beijing, Sha hai, Hong Kong, and other cities in China [15,[58][59][60], indicating the efficacy of the air p lution control measures.It is worth noting that the global spread of COVID-19 since 2 has had a significant impact on anthropogenic activities, especially in China.Decrea industrial and transportation activities result in a reduction in air pollutant emissions cluding NO2, PM2.5, SO2, and VOCs [61-64].Therefore, the impact of the COVID-19 p demic on ambient air pollutant levels from 2019 to 2022 cannot be ignored.In contras PM2.5, the concentration of MDA8 O3 increased by 1.5 µg m −3 year −1 .This finding is c sistent with previous studies that have reported a significant increase in O3 levels in ur areas in China [2,53,65], highlighting the importance of controlling ozone pollution.(OX = O3 + NO2) can be used to describe the atmospheric oxidation capacity of urban ar [66].OX levels decreased at a rate of −1.5 µg m −3 year −1 (Figure S1), indicating a declin atmospheric oxidation capacity.Given the increasing trend in O3 concentrations, the crease in OX concentrations is primally attributed to reduced NO2 concentrations.The MDA8 O3 levels exhibited relatively low values in summer and high values in tumn.This differs from the seasonal variations of O3 in inland cities such as Beijing, Sha hai, Xi'an, and Wuhan, where the highest O3 concentrations consistently occur in the su mer [53,67,68].The levels of O3 in Shenzhen during the summer are influenced by un vorable conditions, such as rainy and humid weather, as well as the dispersion of cl air from the southern sea.Despite the lower temperatures compared to summer (Ta S3), Shenzhen maintains a relatively warm climate during the autumn season, particula in September (28.7 ± 0.02 °C) and October (25.5 ± 0.02 °C).The decrease in relative hum ity and precipitation during autumn, along with unfavorable synoptic systems, such typhoon periphery and subtropical high-pressure systems, contribute to the format and accumulation of O3.The periphery of typhoons and subtropical high-pressure s tems often create favorable conditions, such as intense solar radiation and low w speeds, for the photochemical formation of O3, as well as the accumulation of O3 and precursors [69][70][71].
From 2016 to 2022, MDA8 O3 concentrations significantly increased in spring, tumn, and winter (p < 0.05), while showing no significant trend in summer (p > 0.05).rate of increase in autumn is as high as 3.4 µg m −3 year −1 .PM2.5 concentrations were r tively high in winter and low in summer.The long-term trends in all seasons show significant decreases (p < 0.01), with the highest rate of −3.4 µg m −3 year −1 in winter.Si larly, all precursors (i.e., NO2, SO2, and VOCs) exhibited low levels in summer and h in winter (Figures S3 and S4).NO2 and SO2 levels significantly decreased in all seas between 2016 and 2022 (p < 0.05), whereas VOCs only decreased in spring (p < 0.05) contrast to other pollutants, the long-term trend of OX increased in autumn (0.7 µg   [53,67,68].The levels of O 3 in Shenzhen during the summer are influenced by unfavorable conditions, such as rainy and humid weather, as well as the dispersion of clean air from the southern sea.Despite the lower temperatures compared to summer (Table S3), Shenzhen maintains a relatively warm climate during the autumn season, particularly in September (28.7 ± 0.02 • C) and October (25.5 ± 0.02 • C).The decrease in relative humidity and precipitation during autumn, along with unfavorable synoptic systems, such as typhoon periphery and subtropical high-pressure systems, contribute to the formation and accumulation of O 3 .The periphery of typhoons and subtropical high-pressure systems often create favorable conditions, such as intense solar radiation and low wind speeds, for the photochemical formation of O 3 , as well as the accumulation of O 3 and its precursors [69][70][71].
From 2016 to 2022, MDA8 O 3 concentrations significantly increased in spring, autumn, and winter (p < 0.05), while showing no significant trend in summer (p > 0.05).The rate of increase in autumn is as high as 3.4 µg m −3 year −1 .PM 2.5 concentrations were relatively high in winter and low in summer.The long-term trends in all seasons showed significant decreases (p < 0.01), with the highest rate of −3.4 µg m −3 year −1 in winter.Similarly, all precursors (i.e., NO 2 , SO 2 , and VOCs) exhibited low levels in summer and high in winter (Figures S3 and S4).The number of MDA8 O3 exceedance days autumn was significantly higher than in other seasons, highlighting the importance controlling O3 pollution during this time of year.In contrast to the trend for MDA8 O3, th frequency of PM2.5 levels exceeding the NAAQS showed a significant downward tren declining from 38 days to 0 days.The frequency of PM2.5 exceedance days is much high in winter than in other seasons.Although PM2.5 levels in Shenzhen are below China NAAQS limit, they are significantly higher than the guideline level recommended by th WHO (5 µg m −3 ).Furthermore, there is still a significant difference compared to the leve observed in internationally advanced cities.In 2022, the annual average PM2.5 concentr tion in Shenzhen was 16 µg m −3 , which is notably higher than Tokyo (9.0 µg m −3 ), Londo (9.6 µg m −3 ), New York (9.9 µg m −3 ), and other advanced international cities [72].Therefor it is essential to simultaneously mitigate O3 and PM2.5 to improve air quality in Shenzhe  In contrast to the trend for MDA8 O 3 , the frequency of PM 2.5 levels exceeding the NAAQS showed a significant downward trend, declining from 38 days to 0 days.The frequency of PM 2.5 exceedance days is much higher in winter than in other seasons.Although PM 2.5 levels in Shenzhen are below China's NAAQS limit, they are significantly higher than the guideline level recommended by the WHO (5 µg m −3 ).Furthermore, there is still a significant difference compared to the levels observed in internationally advanced cities.In 2022, the annual average PM 2.5 concentration in Shenzhen was 16 µg m −3 , which is notably higher than Tokyo (9.0 µg m −3 ), London (9.6 µg m −3 ), New York (9.9 µg m −3 ), and other advanced international cities [72].Therefore, it is essential to simultaneously mitigate O 3 and PM 2.5 to improve air quality in Shenzhen.The number of MDA8 O3 exceedance days in autumn was significantly higher than in other seasons, highlighting the importance of controlling O3 pollution during this time of year.In contrast to the trend for MDA8 O3, the frequency of PM2.5 levels exceeding the NAAQS showed a significant downward trend, declining from 38 days to 0 days.The frequency of PM2.5 exceedance days is much higher in winter than in other seasons.Although PM2.5 levels in Shenzhen are below China's NAAQS limit, they are significantly higher than the guideline level recommended by the WHO (5 µg m −3 ).Furthermore, there is still a significant difference compared to the levels observed in internationally advanced cities.In 2022, the annual average PM2.5 concentration in Shenzhen was 16 µg m −3 , which is notably higher than Tokyo (9.0 µg m −3 ), London (9.6 µg m −3 ), New York (9.9 µg m −3 ), and other advanced international cities [72].Therefore, it is essential to simultaneously mitigate O3 and PM2.5 to improve air quality in Shenzhen.

Table 1 presents the correlation coefficients (R) between MDA8 O3, OX, an
Shenzhen from 2016 to 2022.Remarkable positive correlations (R > 0) were throughout each season and the entire year, consistent with findings reported i River Delta region [24,73].The significantly positive correlation between MD PM2.5 (p < 0.01) indicated simultaneous trends in O3 and PM2.5 concentrations.T be attributed to the same precursors of O3 and secondary PM2.5, such as VOC [19,75].Furthermore, elevated O3 concentrations will increase the atmospheric capacity and promote the generation of secondary PM2.5, which contributes to th correlation between O3 and PM2.5 [23,24,73,76].This also explains why the stro relation between MDA8 O3 and PM2.5 is observed in summer (R = 0.62).It has be reported that O3 and PM2.5 are positively correlated in warm seasons, especially [23,77].This might be due to the relatively high solar radiation and temperatu the summer, which greatly enhances the ambient photochemistry, and promot mation of O3 and secondary PM2.5 [69,78,79].The correlation coefficients of PM higher than those of MDA8 O3-PM2.5, suggesting that PM2.5 concentrations tend synchronous with atmospheric oxidation.This may be related to the substantia tions in OX (about 30% in Shenzhen).Particulate nitrate is an important compon ondary PM2.5, which can be generated via the gas-phase oxidation reaction of hydroxyl radicals (OH) [19].As both O3 and particulate nitrate are generated pheric oxidation processes, the control of O3 and PM2.5 pollution is expected from the mitigation of atmospheric oxidation capacity [21].Table 1 presents the correlation coefficients (R) between MDA8 O 3 , O X , and PM 2.5 in Shenzhen from 2016 to 2022.Remarkable positive correlations (R > 0) were observed throughout each season and the entire year, consistent with findings reported in the Pearl River Delta region [24,73].The significantly positive correlation between MDA8 O 3 and PM 2.5 (p < 0.01) indicated simultaneous trends in O 3 and PM 2.5 concentrations.This might be attributed to the same precursors of O 3 and secondary PM 2.5 , such as VOCs and NO 2 [19,75].Furthermore, elevated O 3 concentrations will increase the atmospheric oxidation capacity and promote the generation of secondary PM 2.5 , which contributes to the positive correlation between O 3 and PM 2.5 [23,24,73,76].This also explains why the strongest correlation between MDA8 O 3 and PM 2.5 is observed in summer (R = 0.62).It has been widely reported that O 3 and PM 2.5 are positively correlated in warm seasons, especially in summer [23,77].This might be due to the relatively high solar radiation and temperature during the summer, which greatly enhances the ambient photochemistry, and promotes the formation of O 3 and secondary PM 2.5 [69,78,79].The correlation coefficients of PM 2.5 -O X were higher than those of MDA8 O 3 -PM 2.5 , suggesting that PM 2.5 concentrations tend to be more synchronous with atmospheric oxidation.This may be related to the substantial NO 2 fractions in O X (about 30% in Shenzhen).Particulate nitrate is an important component of secondary PM 2.5 , which can be generated via the gas-phase oxidation reaction of NO 2 with hydroxyl radicals (OH) [19].As both O 3 and particulate nitrate are generated in atmospheric oxidation processes, the control of O 3 and PM 2.5 pollution is expected to benefit from the mitigation of atmospheric oxidation capacity [21].

Influence of Precursors and Meteorological Parameters
When O 3 concentrations exceed China's Grade II NAAQS, they often coincide with increased PM 2.5 levels.This underscores the importance of examining the pollution characteristics of high-O 3 -and-PM 2.5 days, which typically occur in autumn, particularly in September (Figure S5).Table S4 presents the descriptive statistics of MDA8 O 3 , PM 2.5 , O X , NO 2 , and SO 2 on high-O 3 -and-PM 2.5 days and high-O 3 days.Compared to high-O 3 days, the high-O 3 -and-PM 2.5 days showed higher concentrations of MDA8 O 3 , PM 2.5 , O X , SO 2 , and NO 2 (p < 0.01).During high-O 3 -and-PM 2.5 days, the MDA8 O 3 and PM 2.5 concentrations were 195.6 ± 2.1 µg m −3 and 47.2 ± 0.9 µg m −3 , respectively, which were approximately 6% and 80% higher than on high-O 3 days (p < 0.01).
Meteorological parameters, which have significant impacts on O 3 and PM 2.5 concentrations, were statistically analyzed and are presented in Table S5.Compared to high-O 3 days, the high-O 3 -and-PM 2.5 days were associated with lower temperatures and wind speeds (p < 0.01), as well as comparable boundary layer height and lower relative humidity (p > 0.05).Since the temperature difference between the two scenarios was not significant (approximately 1 • C), and the high-O 3 -and-PM 2.5 days often occurred close to high-O 3 days, the occurrence of high-O 3 -and-PM 2.5 days might be attributed to inferior pollutant dispersion conditions under lower wind speeds.The synoptic systems during high-O 3and-PM 2.5 days are statistically presented in Table S6.Three synoptic system patterns governed the high-O 3 -and-PM 2.5 days, including the typhoon periphery, uniform pressure system, and subtropical high-pressure system, with occurrence proportions of 65.9%, 18.2%, and 15.9%, respectively.Under the periphery of a typhoon, unfavorable meteorological conditions occur, including downdrafts and low wind speeds, and the transmission of pollution from upstream cannot be neglected [71].The uniform pressure system always results in calm weather, which is unfavorable for the dispersion of air pollutants [70,80].The subtropical high-pressure system was found to play a crucial role in the formation and accumulation of O 3 and secondary aerosols [81,82].Therefore, under these three patterns of synoptic systems, air pollutants are more likely to accumulate and less likely to disperse, increasing the likelihood of the occurrence of high-O 3 -and-PM 2.5 days.

Contributions of VOC Sources
A total of 19 VOC species, along with one trace gas, were applied to PMF for source apportionment.A five-factor resolution was identified to best describe the source of ambient VOCs (Figure 6).Factor 1 was characterized by considerable percentages of C 2 -C 4 hydrocarbons, n/i-pentanes, and n-hexane, indicating its association with vehicle exhaust [36,83,84].Factor 2 was characterized by the dominance of aromatic hydrocarbons, as well as C 3 -C 5 and C 8 -C 11 hydrocarbons, which are consistent with the composition of gasoline and diesel evaporation [85,86].Thus, this factor was identified as fuel evaporation.Factor 3 exhibited high percentages of n-hexane, n-nonane, n-decane, and TEX (toluene, ethylbenzene, and xylenes), while the proportions of other species were relatively low, which defined as solvent usage [36,87].Factor 4 had high loadings of C2-C3 hydrocarbon and CO, which are associated with natural gas and biomass combustion [85,88].This factor is therefore classified as stationary combustion.Factor 5 was classified as biogenic VOCs (BVOCs) due to its exclusive dominance by isoprene, the indicator of biogenic emissions [33,89].
Atmosphere 2023, 14, x FOR PEER REVIEW 10 o VOCs (BVOCs) due to its exclusive dominance by isoprene, the indicator of biogenic em sions [33,89].Table 2 lists the contributions of anthropogenic sources to VOCs, as well as O3 SOA formations on high-O3-and-PM2.5 days.Notably, vehicle emissions, including veh exhaust and fuel evaporation, were the dominant sources of VOCs during high-O3-a PM2.5 days in Shenzhen, accounting for a total contribution of 58.4 ± 1.3%.Compared other cities in China, the contribution of vehicular emissions was higher than that fou at an urban site in Chongqing (45.1%) [90], while it was comparable to that in Bei (57.7%) [91].The contribution of solvent usage to VOCs in this study (20.2 ± 1.3%) w higher than in Wuhan (16.2%) [88] but much lower than that in Hong Kong (54.1%) [ However, the contribution of stationary combustion (21.4 ± 1.3%) was lower in Shenz than in Wuhan (31.5%) [33].It is worth noting that the results of source identification contributions strongly depend on the species and profiles used for source apportionm the study period, and the sampling site.
This study examined the impact of VOCs on the formation of O3 and SOA.Veh emissions (the combination of vehicle exhaust and fuel evaporation) made the largest c tributions to both O3 formation (45.3 ± 0.6%) and SOA formation (46.6 ± 0.4%).Solv usage was found to make a higher contribution (p < 0.05) to O3 formation (24.8 ± 0.5%) SOA formation (41.7 ± 0.3%) compared to its contribution to VOCs (20.2 ± 1.3%).Howe stationary combustion made a greater contribution to O3 formation (30.0 ± 0.8%) bu lower contribution to SOA formation (11.7 ± 0.1%) compared to its contribution to VO (21.4 ± 1.3%).The discrepancies among source contributions to VOCs and O3 and P formations underscored the importance of emphasizing the chemical reactivities of V species.The outcomes imply that VOCs from vehicle emissions were the main caus the generation of O3 and SOA on high-O3-and-PM2.Table 2 lists the contributions of anthropogenic sources to VOCs, as well as O 3 and SOA formations on high-O 3 -and-PM 2.5 days.Notably, vehicle emissions, including vehicle exhaust and fuel evaporation, were the dominant sources of VOCs during high-O 3 -and-PM 2.5 days in Shenzhen, accounting for a total contribution of 58.4 ± 1.3%.Compared to other cities in China, the contribution of vehicular emissions was higher than that found at an urban site in Chongqing (45.1%) [90], while it was comparable to that in Beijing (57.7%) [91].The contribution of solvent usage to VOCs in this study (20.2 ± 1.3%) was higher than in Wuhan (16.2%) [88] but much lower than that in Hong Kong (54.1%) [92].However, the contribution of stationary combustion (21.4 ± 1.3%) was lower in Shenzhen than in Wuhan (31.5%) [33].It is worth noting that the results of source identification and contributions strongly depend on the species and profiles used for source apportionment, the study period, and the sampling site.This study examined the impact of VOCs on the formation of O 3 and SOA.Vehicle emissions (the combination of vehicle exhaust and fuel evaporation) made the largest contributions to both O 3 formation (45.3 ± 0.6%) and SOA formation (46.6 ± 0.4%).Solvent usage was found to make a higher contribution (p < 0.05) to O 3 formation (24.8 ± 0.5%) and SOA formation (41.7 ± 0.3%) compared to its contribution to VOCs (20.2 ± 1.3%).However, stationary combustion made a greater contribution to O 3 formation (30.0 ± 0.8%) but a lower contribution to SOA formation (11.7 ± 0.1%) compared to its contribution to VOCs (21.4 ± 1.3%).The discrepancies among source contributions to VOCs and O 3 and PM 2.5 formations underscored the importance of emphasizing the chemical reactivities of VOC species.The outcomes imply that VOCs from vehicle emissions were the main cause of the generation of O 3 and SOA on high-O 3 -and-PM 2.5 days.The study by Peng et al. [93] reported that vehicle emissions made the largest contribution (31.1%) to PM 2.5 in Shenzhen in 2021.This underscores the significance of controlling vehicle emissions.S7.Over 64.8% of the trajectories originated from inland regions, which could bring polluted air masses to Shenzhen.About 41.5% of the trajectories originated from Guangdong Province, where Shenzhen is located, and showed small-scale and shortdistance air transport features, indicating the impact of nearby cities on Shenzhen's air quality.The O 3 levels for clusters 4 (147.4± 11.3 µg m −3 ) and 5 (133.8 ± 10.2 µg m −3 ) were much higher than the other clusters (p < 0.01).These clusters passed through Guangdong Province and Jiangxi Province.Regarding PM 2.5 , the values corresponding to clusters 3 and 4 were higher than the others (p < 0.01), with average concentrations of 51.1 ± 1.8 µg m -3 and 48.8 ± 2.1 µg m −3 , respectively.Most trajectories in clusters 3 and 4 originated from Guangdong Province.
Figure 8 shows the potential source regions of O 3 and PM 2.    S7.Over 64.8% of the trajectories originated from inland regions, whic polluted air masses to Shenzhen.About 41.5% of the trajectories originated dong Province, where Shenzhen is located, and showed small-scale and sho transport features, indicating the impact of nearby cities on Shenzhen's air q levels for clusters 4 (147.

Health Risk Assessment
Figure 9 shows the premature mortality associated with long-term exposure to O3 and PM2.5 during 2016 and 2022.Among all-cause premature deaths, cardiovascular diseases made the largest contribution, accounting for 51.2%~70.3% of premature deaths between 2016 and 2022.The number of premature deaths due to long-term exposure to PM2.5 was higher than that related to O3 each year, indicating the greater health risk of exposure to PM2.5.This is consistent with the results of an O3-and PM2.5-related health burden study by Zhang et al. [94], which reported 26.1 and 10.

Health Risk Assessment
Figure 9 shows the premature mortality associated with long-term exposure to O 3 and PM 2.5 during 2016 and 2022.Among all-cause premature deaths, cardiovascular diseases made the largest contribution, accounting for 51.2%~70.3% of premature deaths between 2016 and 2022.The number of premature deaths due to long-term exposure to PM 2.5 was higher than that related to O 3 each year, indicating the greater health risk of exposure to PM 2.5 .This is consistent with the results of an O 3 -and PM 2.5 -related health burden study by Zhang et al. [94], which reported 26.1 and 10.The economic losses caused by long-term exposure to O3 and PM2.5 were calc and are presented in Figure 10.The economic losses due to long-term PM2.5 expos creased from 2016 to 2019 but decreased from 2020 to 2022.However, the economic related to long-term O3 exposure showed an upward trend from 2016 to 2022.Th economic losses due to long-term O3 and PM2.5 exposure increased from USD 13.3 15.4) billion in 2016 to USD 14.7 (10.7, 18.0) billion in 2022.The total economic loss i was estimated to be as high as USD 20.7 (15.3, 24.6) billion, representing an incre 55.1% compared to 2016.Health economic loss is closely related to the VSL, wh creases year by year with economic and social development.The increasing total eco losses may be attributed to the increase in VSL values under the circumstances that ature deaths due to long-term O3 and PM2.5 exposure decreased from 2016 to 202 share of total economic losses in the city's GDP has shown a decreasing trend, w percentage decreasing from 0.64% in 2016 to 0.45% in 2022.The economic losses caused by long-term exposure to O3 and PM2.5 were calc and are presented in Figure 10.The economic losses due to long-term PM2.5 expos creased from 2016 to 2019 but decreased from 2020 to 2022.However, the economic related to long-term O3 exposure showed an upward trend from 2016 to 2022.Th economic losses due to long-term O3 and PM2.5 exposure increased from USD 13. 15.4) billion in 2016 to USD 14.7 (10.7, 18.0) billion in 2022.The total economic loss was estimated to be as high as USD 20.7 (15.3, 24.6) billion, representing an incr 55.1% compared to 2016.Health economic loss is closely related to the VSL, wh creases year by year with economic and social development.The increasing total eco losses may be attributed to the increase in VSL values under the circumstances that ature deaths due to long-term O3 and PM2.5 exposure decreased from 2016 to 202 share of total economic losses in the city's GDP has shown a decreasing trend, w percentage decreasing from 0.64% in 2016 to 0.45% in 2022.

Conclusions
In this study, the long-term trends of O3, PM2.5, OX, and their precursors (VOC and SO2), the characteristics of concurrent O3 and PM2.5 pollution, and the health r long-term exposure to O3 and PM2.5 were analyzed in Shenzhen.It was found tha

Figure 1 .
Figure 1.Location of the sampling sites.

Figure 1 .
Figure 1.Location of the sampling sites.

Figure 3
Figure 3 illustrates the seasonal variation of MDA8 O3 and PM2.5 from 2016 to 20The MDA8 O3 levels exhibited relatively low values in summer and high values in tumn.This differs from the seasonal variations of O3 in inland cities such as Beijing, Sha hai, Xi'an, and Wuhan, where the highest O3 concentrations consistently occur in the su mer[53,67,68].The levels of O3 in Shenzhen during the summer are influenced by un vorable conditions, such as rainy and humid weather, as well as the dispersion of cl air from the southern sea.Despite the lower temperatures compared to summer (Ta S3), Shenzhen maintains a relatively warm climate during the autumn season, particula in September (28.7 ± 0.02 °C) and October (25.5 ± 0.02 °C).The decrease in relative hum ity and precipitation during autumn, along with unfavorable synoptic systems, such typhoon periphery and subtropical high-pressure systems, contribute to the format and accumulation of O3.The periphery of typhoons and subtropical high-pressure s tems often create favorable conditions, such as intense solar radiation and low w speeds, for the photochemical formation of O3, as well as the accumulation of O3 and precursors[69][70][71].From 2016 to 2022, MDA8 O3 concentrations significantly increased in spring, tumn, and winter (p < 0.05), while showing no significant trend in summer (p > 0.05).rate of increase in autumn is as high as 3.4 µg m −3 year −1 .PM2.5 concentrations were r tively high in winter and low in summer.The long-term trends in all seasons show significant decreases (p < 0.01), with the highest rate of −3.4 µg m −3 year −1 in winter.Si larly, all precursors (i.e., NO2, SO2, and VOCs) exhibited low levels in summer and h in winter (FiguresS3 and S4).NO2 and SO2 levels significantly decreased in all seas between 2016 and 2022 (p < 0.05), whereas VOCs only decreased in spring (p < 0.05) contrast to other pollutants, the long-term trend of OX increased in autumn (0.7 µg

Figure 2 .
Figure 2. Trends of daily average MDA8 O 3 and PM 2.5 concentrations in Shenzhen from 2016 to 2022.

Figure 3
Figure 3 illustrates the seasonal variation of MDA8 O 3 and PM 2.5 from 2016 to 2022.The MDA8 O 3 levels exhibited relatively low values in summer and high values in autumn.This differs from the seasonal variations of O 3 in inland cities such as Beijing, Shanghai, Xi'an, and Wuhan, where the highest O 3 concentrations consistently occur in the summer[53,67,68].The levels of O 3 in Shenzhen during the summer are influenced by unfavorable conditions, such as rainy and humid weather, as well as the dispersion of clean air from the southern sea.Despite the lower temperatures compared to summer (TableS3), Shenzhen maintains a relatively warm climate during the autumn season, particularly in September (28.7 ± 0.02 • C) and October (25.5 ± 0.02 • C).The decrease in relative humidity and precipitation during autumn, along with unfavorable synoptic systems, such as typhoon periphery and subtropical high-pressure systems, contribute to the formation and accumulation of O 3 .The periphery of typhoons and subtropical high-pressure systems often create favorable conditions, such as intense solar radiation and low wind speeds, for the photochemical formation of O 3 , as well as the accumulation of O 3 and its precursors[69][70][71].From 2016 to 2022, MDA8 O 3 concentrations significantly increased in spring, autumn, and winter (p < 0.05), while showing no significant trend in summer (p > 0.05).The rate of increase in autumn is as high as 3.4 µg m −3 year −1 .PM 2.5 concentrations were relatively high in winter and low in summer.The long-term trends in all seasons showed significant decreases (p < 0.01), with the highest rate of −3.4 µg m −3 year −1 in winter.Similarly, all precursors (i.e., NO 2 , SO 2 , and VOCs) exhibited low levels in summer and high in winter (FiguresS3 and S4).NO 2 and SO 2 levels significantly decreased in all seasons between 2016 and 2022 (p < 0.05), whereas VOCs only decreased in spring (p < 0.05).In contrast to other pollutants, the long-term trend of O X increased in autumn (0.7 µg m −3 year −1 , p < 0.05) and decreased in other seasons (p < 0.05).The relatively high levels of O X in autumn indicate a more oxidative atmosphere, which favors the formation of O 3 and PM 2.5 .
Figure 3 illustrates the seasonal variation of MDA8 O 3 and PM 2.5 from 2016 to 2022.The MDA8 O 3 levels exhibited relatively low values in summer and high values in autumn.This differs from the seasonal variations of O 3 in inland cities such as Beijing, Shanghai, Xi'an, and Wuhan, where the highest O 3 concentrations consistently occur in the summer[53,67,68].The levels of O 3 in Shenzhen during the summer are influenced by unfavorable conditions, such as rainy and humid weather, as well as the dispersion of clean air from the southern sea.Despite the lower temperatures compared to summer (TableS3), Shenzhen maintains a relatively warm climate during the autumn season, particularly in September (28.7 ± 0.02 • C) and October (25.5 ± 0.02 • C).The decrease in relative humidity and precipitation during autumn, along with unfavorable synoptic systems, such as typhoon periphery and subtropical high-pressure systems, contribute to the formation and accumulation of O 3 .The periphery of typhoons and subtropical high-pressure systems often create favorable conditions, such as intense solar radiation and low wind speeds, for the photochemical formation of O 3 , as well as the accumulation of O 3 and its precursors[69][70][71].From 2016 to 2022, MDA8 O 3 concentrations significantly increased in spring, autumn, and winter (p < 0.05), while showing no significant trend in summer (p > 0.05).The rate of increase in autumn is as high as 3.4 µg m −3 year −1 .PM 2.5 concentrations were relatively high in winter and low in summer.The long-term trends in all seasons showed significant decreases (p < 0.01), with the highest rate of −3.4 µg m −3 year −1 in winter.Similarly, all precursors (i.e., NO 2 , SO 2 , and VOCs) exhibited low levels in summer and high in winter (FiguresS3 and S4).NO 2 and SO 2 levels significantly decreased in all seasons between 2016 and 2022 (p < 0.05), whereas VOCs only decreased in spring (p < 0.05).In contrast to other pollutants, the long-term trend of O X increased in autumn (0.7 µg m −3 year −1 , p < 0.05) and decreased in other seasons (p < 0.05).The relatively high levels of O X in autumn indicate a more oxidative atmosphere, which favors the formation of O 3 and PM 2.5 .

Figure 3 .
Figure 3. Trends of monthly average MDA8 O3 and PM2.5 concentrations in different seasons Shenzhen from 2016 to 2022.The error bar represents the 95% CI of the monthly averages.

Figure 4
Figure4shows the frequency of MDA8 O3 and PM2.5 daily values exceeding Grade NAAQS of China (160 µg m −3 for MDA8 O3, and 75 µg m −3 for PM2.5) at the 13 BAQM from 2016 to 2022.The frequency of MDA8 O3 concentrations exceeding the NAAQS di played an upward trend, increasing from 147 days in 2016 to 380 days in 2022, highligh ing the severity of O3 pollution in Shenzhen.The number of MDA8 O3 exceedance days autumn was significantly higher than in other seasons, highlighting the importance controlling O3 pollution during this time of year.In contrast to the trend for MDA8 O3, th frequency of PM2.5 levels exceeding the NAAQS showed a significant downward tren declining from 38 days to 0 days.The frequency of PM2.5 exceedance days is much high in winter than in other seasons.Although PM2.5 levels in Shenzhen are below China NAAQS limit, they are significantly higher than the guideline level recommended by th WHO (5 µg m −3 ).Furthermore, there is still a significant difference compared to the leve observed in internationally advanced cities.In 2022, the annual average PM2.5 concentr tion in Shenzhen was 16 µg m −3 , which is notably higher than Tokyo (9.0 µg m −3 ), Londo (9.6 µg m −3 ), New York (9.9 µg m −3 ), and other advanced international cities[72].Therefor it is essential to simultaneously mitigate O3 and PM2.5 to improve air quality in Shenzhe

Figure 3 .
Figure 3. Trends of monthly average MDA8 O 3 and PM 2.5 concentrations in different seasons in Shenzhen from 2016 to 2022.The error bar represents the 95% CI of the monthly averages.

Figure 4
Figure 4 shows the frequency of MDA8 O 3 and PM 2.5 daily values exceeding Grade II NAAQS of China (160 µg m −3 for MDA8 O 3 , and 75 µg m −3 for PM 2.5 ) at the 13 BAQMS from 2016 to 2022.The frequency of MDA8 O 3 concentrations exceeding the NAAQS displayed an upward trend, increasing from 147 days in 2016 to 380 days in 2022, highlighting the severity of O 3 pollution in Shenzhen.The number of MDA8 O 3 exceedance days in autumn was significantly higher than in other seasons, highlighting the importance of controlling O 3 pollution during this time of year.In contrast to the trend for MDA8 O 3 , the frequency of PM 2.5 levels exceeding the NAAQS showed a significant downward trend, declining from 38 days to 0 days.The frequency of PM 2.5 exceedance days is much higher in winter than in other seasons.Although PM 2.5 levels in Shenzhen are below China's NAAQS limit, they are significantly higher than the guideline level recommended by the WHO (5 µg m −3 ).Furthermore, there is still a significant difference compared to the levels observed in internationally advanced cities.In 2022, the annual average PM 2.5 concentration in Shenzhen was 16 µg m −3 , which is notably higher than Tokyo (9.0 µg m −3 ), London (9.6 µg m −3 ), New York (9.9 µg m −3 ), and other advanced international cities[72].Therefore, it is essential to simultaneously mitigate O 3 and PM 2.5 to improve air quality in Shenzhen.

Atmosphere 2023 ,
14,  x FOR PEER REVIEW 7 of 18 year −1 , p < 0.05) and decreased in other seasons (p < 0.05).The relatively high levels of OX in autumn indicate a more oxidative atmosphere, which favors the formation of O3 and PM2.5.

Figure 3 .
Figure 3. Trends of monthly average MDA8 O3 and PM2.5 concentrations in different seasons in Shenzhen from 2016 to 2022.The error bar represents the 95% CI of the monthly averages.

Figure 4
Figure4shows the frequency of MDA8 O3 and PM2.5 daily values exceeding Grade II NAAQS of China (160 µg m −3 for MDA8 O3, and 75 µg m −3 for PM2.5) at the 13 BAQMS from 2016 to 2022.The frequency of MDA8 O3 concentrations exceeding the NAAQS displayed an upward trend, increasing from 147 days in 2016 to 380 days in 2022, highlighting the severity of O3 pollution in Shenzhen.The number of MDA8 O3 exceedance days in autumn was significantly higher than in other seasons, highlighting the importance of controlling O3 pollution during this time of year.In contrast to the trend for MDA8 O3, the frequency of PM2.5 levels exceeding the NAAQS showed a significant downward trend, declining from 38 days to 0 days.The frequency of PM2.5 exceedance days is much higher in winter than in other seasons.Although PM2.5 levels in Shenzhen are below China's NAAQS limit, they are significantly higher than the guideline level recommended by the WHO (5 µg m −3 ).Furthermore, there is still a significant difference compared to the levels observed in internationally advanced cities.In 2022, the annual average PM2.5 concentration in Shenzhen was 16 µg m −3 , which is notably higher than Tokyo (9.0 µg m −3 ), London (9.6 µg m −3 ), New York (9.9 µg m −3 ), and other advanced international cities[72].Therefore, it is essential to simultaneously mitigate O3 and PM2.5 to improve air quality in Shenzhen.

Figure 4 . 18 3. 2 . 5 Figure 5
Figure 4. Frequency of MDA8 O 3 and PM 2.5 levels exceeding the NAAQS of China in different seasons at 13 BAQM sites from 2016 to 2022.Red bars in the graph indicate days when MDA8 O 3 levels exceeded NAAQS, while purple bars represent days when PM 2.5 levels exceeded NAAQS.

Figure 4 . 5 3. 2 . 1 . 5 Figure 5
Figure 4. Frequency of MDA8 O3 and PM2.5 levels exceeding the NAAQS of China in d sons at 13 BAQM sites from 2016 to 2022.Red bars in the graph indicate days when MDA exceeded NAAQS, while purple bars represent days when PM2.5 levels exceeded NAAQ

Figure 5 .
Figure 5. Variation of average O 3 concentrations with PM 2.5 concentrations.

Figure 6 .
Figure 6.Source profiles of ambient VOCs at the LH site during high-O3-and-PM2.5 days from 2 to 2022.
5 days.The study by Peng et al. reported that vehicle emissions made the largest contribution (31.1%) to PM2.5 in Shenz in 2021.This underscores the significance of controlling vehicle emissions.

Figure 6 .
Figure 6.Source profiles of ambient VOCs at the LH site during high-O 3 -and-PM 2.5 days from 2019 to 2022.

3. 4 .
Potential Source Regions of O 3 and PM 2.5

Figure 7
Figure 7 depicts the 24 h backward trajectories on 58 high-O 3 -and-PM 2.5 days in Shenzhen from 2016 to 2022.The trajectories were classified into seven clusters based on their directions and traveled regions.The proportions of trajectories for each cluster, the traveled regions, and the corresponding mean O 3 and PM 2.5 concentrations are presented in TableS7.Over 64.8% of the trajectories originated from inland regions, which could bring polluted air masses to Shenzhen.About 41.5% of the trajectories originated from Guangdong Province, where Shenzhen is located, and showed small-scale and shortdistance air transport features, indicating the impact of nearby cities on Shenzhen's air quality.The O 3 levels for clusters 4 (147.4± 11.3 µg m −3 ) and 5 (133.8 ± 10.2 µg m −3 ) were much higher than the other clusters (p < 0.01).These clusters passed through Guangdong Province and Jiangxi Province.Regarding PM 2.5 , the values corresponding to clusters 3 and 4 were higher than the others (p < 0.01), with average concentrations of 51.1 ± 1.8 µg m -3 and 48.8 ± 2.1 µg m −3 , respectively.Most trajectories in clusters 3 and 4 originated from Guangdong Province.Figure8shows the potential source regions of O 3 and PM 2.5 calculated via the PSCF and CWT models.The areas with elevated PSCF (PSCF > 0.6) and CWT values (CWT > 140 µg m −3 for O 3 and >45 µg m −3 for PM 2.5 ) were all situated in cities within Guangdong Province.The cities of Dongguan, Huizhou, and Guangzhou were identified as potential source areas of O 3 via PSCF and CWT models.For PM 2.5 , the potential source areas were identified as Dongguan, Huizhou, Guangzhou, Foshan, and Jiangmen.The potential source areas of O 3 and PM 2.5 suggest the importance of coordinated prevention and control of air pollution in Guangdong Province to mitigate O 3 and PM 2.5 .Collaborating on air pollution control with Guangzhou, Dongguan, and Huizhou could effectively reduce O 3 and PM 2.5 levels in Shenzhen simultaneously.
Figure 7 depicts the 24 h backward trajectories on 58 high-O 3 -and-PM 2.5 days in Shenzhen from 2016 to 2022.The trajectories were classified into seven clusters based on their directions and traveled regions.The proportions of trajectories for each cluster, the traveled regions, and the corresponding mean O 3 and PM 2.5 concentrations are presented in TableS7.Over 64.8% of the trajectories originated from inland regions, which could bring polluted air masses to Shenzhen.About 41.5% of the trajectories originated from Guangdong Province, where Shenzhen is located, and showed small-scale and shortdistance air transport features, indicating the impact of nearby cities on Shenzhen's air quality.The O 3 levels for clusters 4 (147.4± 11.3 µg m −3 ) and 5 (133.8 ± 10.2 µg m −3 ) were much higher than the other clusters (p < 0.01).These clusters passed through Guangdong Province and Jiangxi Province.Regarding PM 2.5 , the values corresponding to clusters 3 and 4 were higher than the others (p < 0.01), with average concentrations of 51.1 ± 1.8 µg m -3 and 48.8 ± 2.1 µg m −3 , respectively.Most trajectories in clusters 3 and 4 originated from Guangdong Province.Figure8shows the potential source regions of O 3 and PM 2.5 calculated via the PSCF and CWT models.The areas with elevated PSCF (PSCF > 0.6) and CWT values (CWT > 140 µg m −3 for O 3 and >45 µg m −3 for PM 2.5 ) were all situated in cities within Guangdong Province.The cities of Dongguan, Huizhou, and Guangzhou were identified as potential source areas of O 3 via PSCF and CWT models.For PM 2.5 , the potential source areas were identified as Dongguan, Huizhou, Guangzhou, Foshan, and Jiangmen.The potential source areas of O 3 and PM 2.5 suggest the importance of coordinated prevention and control of air pollution in Guangdong Province to mitigate O 3 and PM 2.5 .Collaborating on air pollution control with Guangzhou, Dongguan, and Huizhou could effectively reduce O 3 and PM 2.5 levels in Shenzhen simultaneously.

Figure 7
Figure 7 depicts the 24 h backward trajectories on 58 high-O3-and-PM2.5 zhen from 2016 to 2022.The trajectories were classified into seven clusters b directions and traveled regions.The proportions of trajectories for each clu eled regions, and the corresponding mean O3 and PM2.5 concentrations are TableS7.Over 64.8% of the trajectories originated from inland regions, whic polluted air masses to Shenzhen.About 41.5% of the trajectories originated dong Province, where Shenzhen is located, and showed small-scale and sho transport features, indicating the impact of nearby cities on Shenzhen's air q levels for clusters 4 (147.4± 11.3 µg m −3 ) and 5 (133.8 ± 10.2 µg m −3 ) were muc the other clusters (p < 0.01).These clusters passed through Guangdong Jiangxi Province.Regarding PM2.5, the values corresponding to clusters higher than the others (p < 0.01), with average concentrations of 51.1 ± 1.8 µ ± 2.1 µg m −3 , respectively.Most trajectories in clusters 3 and 4 originated from Province.
Figure 7 depicts the 24 h backward trajectories on 58 high-O3-and-PM2.5 zhen from 2016 to 2022.The trajectories were classified into seven clusters b directions and traveled regions.The proportions of trajectories for each clu eled regions, and the corresponding mean O3 and PM2.5 concentrations are TableS7.Over 64.8% of the trajectories originated from inland regions, whic polluted air masses to Shenzhen.About 41.5% of the trajectories originated dong Province, where Shenzhen is located, and showed small-scale and sho transport features, indicating the impact of nearby cities on Shenzhen's air q levels for clusters 4 (147.4± 11.3 µg m −3 ) and 5 (133.8 ± 10.2 µg m −3 ) were muc the other clusters (p < 0.01).These clusters passed through Guangdong Jiangxi Province.Regarding PM2.5, the values corresponding to clusters higher than the others (p < 0.01), with average concentrations of 51.1 ± 1.8 µ ± 2.1 µg m −3 , respectively.Most trajectories in clusters 3 and 4 originated from Province.
Figure9shows the premature mortality associated with long-term exposure to O3 and PM2.5 during 2016 and 2022.Among all-cause premature deaths, cardiovascular diseases made the largest contribution, accounting for 51.2%~70.3% of premature deaths between 2016 and 2022.The number of premature deaths due to long-term exposure to PM2.5 was higher than that related to O3 each year, indicating the greater health risk of exposure to PM2.5.This is consistent with the results of an O3-and PM2.5-related health burden study by Zhang et al.[94], which reported 26.1 and 10.3 thousand premature deaths related to long-term PM2.5 and O3 exposure in Guangdong Province in 2020, respectively.From 2016 to 2022, the number of premature deaths attributed to long-term PM2.5 exposure decreased by 36.1%, from 2068 (95% CI: 1600, 2292) in 2016 to 1322 (1013, 1529) in 2022.During the same period, premature deaths due to long-term O3 exposure increased from 521 (338, 699) in 2016 to 747 (486, 1001) in 2022, representing an increase of 43.4%.Notably, the number of premature deaths due to PM2.5 exposure reached its highest value in 2019.The highest level of O3 in 2019 was the primary factor contributing to the highest number of premature deaths related to O3 exposure in that year.The high premature mortality in 2019 can be attributed to a slight decrease in PM2.5 levels and a significant increase in population from 2016 to 2019.The average concentration of PM2.5 decreased by 1.6 µg m −3 in 2019 compared to 2016, while the population increased by about 2.09 million during this period.An increase in the population exposed to PM2.5 leads to an elevation in premature mortality.Despite the increase in O3-related premature deaths, the total number of premature deaths due to O3 and PM2.5 exposure showed a downward trend, decreasing from 2589(1937, 2991) in 2016 to 2068 (1498, 2530) in 2022.This outcome indicates the improvement in air quality and reduction in health impacts in Shenzhen.Given that premature deaths related to long-term O3 and PM2.5 exposure remain at high levels in Shenzhen, it is imperative to implement more effective O3 and PM2.5 control strategies in the future.

Figure 8 .
Figure 8. PSCF and CWT maps for O 3 and PM 2.5 during high-O 3 -and-PM 2.5 days in Shenzhen from 2016 to 2022.
Figure9shows the premature mortality associated with long-term exposure to O 3 and PM 2.5 during 2016 and 2022.Among all-cause premature deaths, cardiovascular diseases made the largest contribution, accounting for 51.2%~70.3% of premature deaths between 2016 and 2022.The number of premature deaths due to long-term exposure to PM 2.5 was higher than that related to O 3 each year, indicating the greater health risk of exposure to PM 2.5 .This is consistent with the results of an O 3 -and PM 2.5 -related health burden study by Zhang et al.[94], which reported 26.1 and 10.3 thousand premature deaths related to long-term PM 2.5 and O 3 exposure in Guangdong Province in 2020, respectively.From 2016 to 2022, the number of premature deaths attributed to long-term PM 2.5 exposure decreased by 36.1%, from 2068 (95% CI: 1600, 2292) in 2016 to 1322 (1013, 1529) in 2022.During the same period, premature deaths due to long-term O 3 exposure increased from 521 (338, 699) in 2016 to 747 (486, 1001) in 2022, representing an increase of 43.4%.Notably, the number of premature deaths due to PM 2.5 exposure reached its highest value in 2019.The highest level of O 3 in 2019 was the primary factor contributing to the highest number of premature deaths related to O 3 exposure in that year.The high premature mortality in 2019 can be attributed to a slight decrease in PM 2.5 levels and a significant increase in population from 2016 to 2019.The average concentration of PM 2.5 decreased by 1.6 µg m −3 in 2019 compared to 2016, while the population increased by about 2.09 million during this period.An increase in the population exposed to PM 2.5 leads to an elevation in premature mortality.Despite the increase in O 3 -related premature deaths, the total number of premature deaths due to O 3 and PM 2.5 exposure showed a downward trend, decreasing from 2589 (1937, 2991) in 2016 to 2068 (1498, 2530) in 2022.This outcome indicates the improvement in air quality and reduction in health impacts in Shenzhen.Given that premature deaths related to long-term O 3 and PM 2.5 exposure remain at high levels in Shenzhen, it is imperative to implement more effective O 3 and PM 2.5 control strategies in the future.

Figure 9 .Figure 9 .
Figure 9. Premature mortalities attributable to long-term exposure to O 3 and PM 2.5 .The economic losses caused by long-term exposure to O 3 and PM 2.5 were calculated and are presented in Figure 10.The economic losses due to long-term PM 2.5 exposure increased from 2016 to 2019 but decreased from 2020 to 2022.However, the economic losses related to long-term O 3 exposure showed an upward trend from 2016 to 2022.The total economic losses due to long-term O 3 and PM 2.5 exposure increased from USD 13.3 (10.0, 15.4) billion in 2016 to USD 14.7 (10.7, 18.0) billion in 2022.The total economic loss in 2019 was estimated to be as high as USD 20.7 (15.3, 24.6) billion, representing an increase of 55.1% compared to 2016.Health economic loss is closely related to the VSL, which increases year by year with economic and social development.The increasing total economic losses may be attributed to the increase in VSL values under the circumstances that premature deaths due to long-term O 3 and PM 2.5 exposure decreased from 2016 to 2022.The share of total economic losses in the city's GDP has shown a decreasing trend, with the percentage decreasing from 0.64% in 2016 to 0.45% in 2022.

Figure 10 .
Figure 10.Economic losses attributed to long-term exposure to O 3 and PM 2.5 .

Table 1 .
The Spearman correlation coefficients of daily MDA8 O 3 , O X , and PM 2.5 concentrations in different seasons from 2016 to 2022.