Atmospheric oxidation capacity and ozone pollution mechanism in a coastal city of southeastern China: analysis of a typical photochemical episode by an observation-based model

A typical multi-day ozone (O3) pollution event was chosen to explore the atmospheric oxidation capacity (AOC), OH reactivity, radical chemistry, and O3 pollution mechanism in a coastal city of southeastern China, with an observation-based model coupled to the Master Chemical Mechanism (OBM-MCM). The hydroxyl radical (OH) was the predominant oxidant (90± 25 %) for daytime AOC, while the NO3 radical played an important role in AOC during the nighttime (72± 9 %). Oxygenated volatile organic compounds (OVOCs; 30± 8 %), NO2 (29± 8 %), and CO (25± 5 %) were the dominant contributors to OH reactivity, accelerating the production of O3 and recycling of ROx radicals (ROx =OH+HO2+RO2). Photolysis of nitrous acid (HONO, 33± 14 %), O3 (25± 13 %), formaldehyde (HCHO, 20± 5 %), and other OVOCs (17± 2 %) was a major ROx source, which played an initiation role in atmospheric oxidation processes. Combined with regional transport analysis, the reasons for this O3 episode were the accumulation of local photochemical production and regional transport. The results of sensitivity analysis showed that volatile organic compounds (VOCs) were the limiting factor of radical recycling and O3 formation, and the 5 % reduction of O3 would be achieved by decreasing 20 % anthropogenic VOCs. Controlling emissions of aromatics, alkenes, and alkanes with ≥ 4 carbons was beneficial for ozone pollution mitigation. The findings of this study provide significant guidance for emission reduction and regional collaboration for future photochemical pollution control in the relatively clean coastal cities of China and similar countries. Published by Copernicus Publications on behalf of the European Geosciences Union. 2174 T. Liu et al.: AOC and ozone pollution mechanism in a coastal city of southeastern China


Introduction
Tropospheric ozone (O 3 ) is mainly produced by photochemical reactions of anthropogenic and natural emitted volatile organic compounds (VOCs) and nitrogen oxides (NO x ) and is an important factor resulting in regional air pollution Lu et al., 2018). Elevated O 3 concentrations enhance the atmospheric oxidation capacity (AOC) and have harmful effects on global climate change, ecosystems, and human health (J. Fowler et al., 2009). The formation mechanisms of O 3 pollution are extremely difficult to figure out, due to the complex types and sources of its precursors (Simon et al., 2015). O 3 formation is affected by multiple factors, such as O 3 precursor speciation or level, atmospheric oxidation capacity, meteorological conditions, and regional transport (Gong and Liao, 2019; Chang et al., 2019). To effectively control tropospheric O 3 pollution, exploration of the photochemical mechanism and judgment of the controlling factors of O 3 formation have become extremely important for the scientific community Li et al., 2018).
The atmospheric oxidation capacity reflects the essential driving force in tropospheric chemistry and plays an important place in the loss rates of primary components and production rates of secondary pollutants; thus the key factors to quantify AOC are processes and rates of species being oxidized in the atmosphere (Elshorbany et al., 2009). Atmospheric conditions (such as photolysis rate, meteorology, pollutant concentrations, and regional transport) together influence the AOC levels, and the AOC levels in polluted urban regions are generally much higher than those at the background sites or remote regions due to the dominant limiting factor for the significant differences of pollutant concentrations (Geyer et al., 2001;Xue et al., 2016). RO x radicals, including the hydroxyl radical (OH), hydroperoxy radical (HO 2 ), and organic peroxy radical (RO 2 ), are very important indicators in atmospheric photochemistry and dominate the atmospheric oxidation capacity (Li et al., 2018). Meanwhile, radical chemistry drives the transformation and recycling of O 3 through initiating atmospheric oxidation processes . Among these radicals, the OH radical accounts for the majority of AOC over 90 % during the daytime; thus the OH reactivity (i.e., OH loss) indicates the primary contribution of individual pollutants (H. Mao et al., 2010). Hence, atmospheric oxidation capacity, OH reactivity, and radical chemistry are crucial aspects for understanding the complex atmospheric photochemistry processes (Li et al., 2018). For example, major RO x sources are the photolysis reaction of O 3 , formaldehyde (HCHO), other oxygenated volatile organic compounds (OVOCs), nitrous acid (HONO), and the reactions of O 3 with unsaturated volatile organic compounds (VOCs; Volkamer et al., 2010). The dominant RO x sources at some rural sites were O 3 photolysis and O 3 reactions with VOCs (Li et al., 2018;Martinez et al., 2003), and those at many urban sites were HONO and OVOC photolysis (Xue et al., 2016;Liu et al., 2012;Emmerson et al., 2005). For oil and gas field sites, there were highly abundant VOCs to promote the formation of O 3 , and the contribution of OVOC photolysis was 2-5 times higher than that in urban areas Edwards et al., 2013Edwards et al., , 2014. HONO photolysis was a very important RO x source at the high-altitude or background sites. (Acker et al., 2001;Jiang et al., 2020).
Current studies of atmospheric O 3 photochemical pollution observations have been conducted at urban, suburban, rural, and remote sites around the world (Smith et al., 2006;Eisele et al., 1997;Kanaya et al., 2001;Hofzumahaus et al., 2009;George et al., 1999;Emmerson et al., 2005;Kanaya et al., 2007;Michoud et al., 2012). In China, O 3 photochemical pollution events have been reported in some megacities, such as Beijing, Shanghai, Guangzhou, and Chengdu Tan et al., 2019;Zhu et al., 2020;Wang et al., 2020;Ling et al., 2017). Few studies on O 3 photochemical pollution in cities with low O 3 precursor emissions have been reported, and the air quality in these areas usually depends on the change of meteorological conditions. In a coastal city of southeastern China, the concentrations of O 3 precursors were higher than those in remote sites and background but lower than those in most urban and suburban areas, even lower than those in rural regions (Table S1). In a word, O 3 precursor emissions in our observation site were relatively low. Meanwhile, the southeastern coastal region is influenced by the East Asian monsoon and acts as an important transport path between the Yangtze River Delta (YRD) and the Pearl River Delta (PRD) (T. Liu et al., 2020a, b), which is a good "laboratory" to further explore O 3 photochemical pollution and formation mechanisms with relatively low O 3 precursors and complex meteorological conditions (Zhang et al., 2020b;Hu et al., 2020).
Observation-based models (OBMs) are widely used to investigate O 3 -VOC-NO x relationships and radical chemistry (H. Tan et al., 2019). The O 3 sensitivity revealed the nonlinear relationship between O 3 and its precursors (i.e., VOCs and NO x ), which was established to investigate O 3 formation mechanisms and control strategies . An OBM combined with the Master Chemical Mechanism (V3.3.1) (OBM-MCM) has been applied to explore the O 3 photochemical pollution mechanism in different environmental conditions Li et al., 2018;Xue et al., 2016;. In this study, we chose a typical multi-day O 3 pollution event in the coastal city Xiamen (Fig. S1), when Xiamen was affected by various meteorological conditions, such as typhoons and the West Pacific Subtropical High (WPSH), accompanied by temperature inversion phenomena. Based on the OBM-MCM analyses, the study aims to clarify (1) the pollution characteristics of O 3 and its precursors, (2) the atmospheric oxidation capacity and radical chemistry, and (3) the O 3 formation mechanism and sensitivity analysis. The results are expected to enhance the understanding of the O 3 formation mechanism with low O 3 precursor levels and provide scientific evidence for O 3 pollution control in the coastal cities.

Study area and field observations
Xiamen is a coastal city in the southeastern area of China, to the west coast of the Taiwan Strait. The field campaigns were carried out at the Atmospheric Environment Observation Supersite (24.61 • N, 118.06 • E) on the rooftop of around 70 m high building in the Institute of Urban Environment, Chinese Academy of Sciences. The supersite was equipped with complete monitoring instruments to measure gas and aerosol species compositions, O 3 precursors, meteorological parameters, and photolysis rate. Criteria air pollutants of O 3 , SO 2 , NO-NO 2 -NO x , and CO were monitored by commercial instruments TEI 49i, 43i, 42i, and 48i (Thermo Fisher Scientific, USA), respectively. The meteorological parameters of wind speed (WS), wind direction (WD), air temperature (T ), pressure (P ), and relative humidity (RH) were measured by a weather station with a sonic anemometer (150WX, Airmar, USA). HONO was measured with an analyzer for Monitoring Aerosols and Gases in Ambient Air (MARGA; ADI 2080, Applikon Analytical B.V., the Netherlands). A gas chromatography-mass spectrometer (GC-FID/MS, TH-300B, Wuhan, CN) was used for the monitoring of atmospheric VOC concentrations, involving about 103 species of VOCs with a 1 h time resolution. Photolysis frequencies were measured by a photolysis spectrometer (PFS-100, Focused Photonics Inc., Hangzhou, China). The photolysis rate constants include J (O 1 D), J (NO 2 ), J (H 2 O 2 ), J (HONO), J (HCHO), and J (NO 3 ). Strict quality assurance and quality control were applied, and the detailed descriptions of the monitoring procedures were documented in our previous studies (Zhang et al., 2020b;Wu et al., 2020;Liu et al., 2020a;T. Liu et al., 2020b;Hu et al., 2020).

Observation-based chemical box model
In this study, an observation-based model (OBM) combined with the latest version 3.3.1 of MCM (MCM v3.3.1; http: //mcm.leeds.ac.uk/MCM/, last access: 25 July 2021), involving 142 non-methane VOCs and more than 17 000 elementary reactions of 6700 primary, secondary, and radical species Saunders et al., 2003), was used to explore the atmospheric oxidation processes and O 3 formation mechanisms. The physical process of deposition within the boundary layer height (BLH), which varied from 300 m during nighttime to 1500 m during the daytime in autumn (Li et al., 2018), was considered in the model. Therefore, the dry deposition velocity was utilized to simulate the deposition loss of some reactants in the atmosphere (shown in Table S2), which avoided continuous accumulation of pollutant concentrations in the model (Zhang et al., 2003;Xue et al., 2016).
The observation parameters of the gaseous pollutants (i.e., O 3 , CO, NO, NO 2 , HONO, SO 2 , and VOCs), meteorological parameters (i.e., T , P , and RH), and photolysis rate constants (J (O 1 D), J (NO 2 ), J (H 2 O 2 ), J (HONO), J (HCHO), and J (NO 3 )) were input into the OBM-MCM model as constraints. The photolysis rates of other molecules such as OVOCs were parameterized by solar zenith angle and then scaled by the measured J (NO 2 ) . We pre-ran the model for 5 d before the start of the experiment to initialize the unmeasured compounds and radicals (Xue et al., 2014).
OBM-MCM is mainly used to simulate in situ atmospheric photochemical processes and quantify the O 3 production rate, AOC, OH reactivity, and RO x radical budgets. Among them, primary sources of RO x , including the photolysis reactions of O 3 , HONO, formaldehyde (HCHO), and other OVOCs, as well as reactions of VOCs with O 3 and NO 3 radicals, are important (Xue et al., 2016). The termination reactions of RO x are controlled by cross-reactions with NO x (under high-NO x conditions) and RO x (under low-NO x conditions) to form nitric acid, organic nitrates, and peroxides Xue et al., 2016). Table 1 shows the production and destruction reactions and relevant reaction rates of O 3 in our study. The production rate of O 3 (P (O 3 )) includes RO 2 + NO (R1) and HO 2 + NO reactions (R2, Eq. 1), and the destruction of O 3 (D(O 3 )) involves reactions of O 3 photolysis (R3), O 3 + OH (R4), O 3 + HO 2 (R5), NO 2 + OH (R6), O 3 + VOCs (R7), and NO 3 + VOCs (R8, Eq. 2). The net O 3 production rate (P net(O 3 )) is calculated by P (O 3 ) minus D(O 3 ), as in Eq. (3).

Model performance
The index of agreement (IOA) can be used to judge the reliability of the model simulation results, and its equation is (X. Liu et al., 2019) where S i is the simulated value, O i represents the observed values, O denotes the average observed values, and n is the sample number. The IOA range is 0-1, and the higher the IOA value is, the better agreement there is between simulated and observed values. In many studies, when IOA ranges from 0.68 to 0.89 (Y. , the simulation results are reasonable, and the IOA in our research is 0.80. Hence, the performance of the OBM-MCM model was reasonably acceptable.

Meteorological data and back trajectory calculation
The backward trajectories of air masses arriving at the observation site were calculated by MeteoInfo during the episode . The backward trajectories with 72 h were run with a time resolution of 3 h at 100 m height above ground level. The starting time was 00:00 LT, and the ending time was 23:00 LT. Meteorological data were provided by NOAA ARL (ftp://arlftp.arlhq.noaa.gov/pub/archives/gdas1, last access: 3 January 2020). The Final Operational Global Analysis data (FNL) are from the Global Data Assimilation System and analyze results with the model, which is also used by the National Center for Environmental Prediction (NCEP) in the Global Forecast System (GFS) (https://rda. ucar.edu/datasets/ds083.2/, last access: 17 December 2019). The weather charts were conducted using Grid Analysis and Display System (GrADS) with the specific programmed script files. A detailed description of the synoptic information was shown in our previous study (Wu et al., 2019).

Overview of observations
The O 3 pollution events frequently appeared in the coastal city Xiamen during autumn, related to the WPSH, carrying favorable photochemical reaction conditions (high temperature, low RH, and stagnant weather conditions) and encouraging the formation and accumulation of O 3 in the southeastern coastal area (H.   Fig. 1 and Table 2. During this period, the dominant wind direction was northeast, with an average wind speed of 1.8 ± 0.9 m s −1 . The maximum hourly temperature was as high as 35 • C, and the average RH was 56.4 ± 12.6 %. Our previous study showed that particulate pollution was slight in Xiamen, which could affect solar radiation by the light-absorbing component, and the concentrations of particulate matter had not exceeded the National Ambient Air Quality Standard (Class II: 75 µg m −3 )  for a whole year (Hu et al., 2021;Deng et al., 2020). Therefore, solar radiation intensity and J (NO 2 ) were strong, compared to those of the Yellow River Delta , Shanghai , and Hong Kong (Xue et al., 2016). In general, these meteorological parameters were conducive to the production and accumulation of O 3 . In addition, O 3 concentrations at nighttime stayed relatively high ( Fig. 1), indicating the influence of regional transport and little NO titration Wu et al., 2020). Figure S2 shows the 72 h back trajectories at the monitoring site. Among them, 80 % of the air masses came from the Yellow Sea, and the other 20 % air masses originated from the northeast China through long-range transport. Table 3 lists the detailed VOC concentrations during the observation period. Alkanes (6970 ± 2325 pptv) were the predominant components of total VOCs, followed by OVOCs (4246 ± 1263 pptv), aromatics (2131 ± 1236 pptv), halocarbons (1951 ± 572 pptv), alkenes (1205 ± 464 pptv), and acetylene (674 ± 290 pptv). The ratio of ethene / ethane (0.4 ± 0.2) was significantly (p < 0.05) lower than that in Hong Kong (0.7 ± 0.1) with significant aged air masses, indicating that the high O 3 in Xiamen might be partially attributed to the aged air masses (e.g., transport of air from polluted regions or intense atmospheric oxidation) (H. . The concentration of TVOCs in Xiamen (17.2 ± 4.8 ppbv) was much lower than that in the developed areas with large anthropogenic emissions (i.e., Beijing (44.2 ppbv), Lanzhou (45.3 ppbv), Wuhan (30.2 ppbv), Chengdu (36.0 ppbv), Hong Kong (26.9 ppbv), Los Angeles (41.3 ppbv), and Tokyo (43.4 ppbv)) but was higher than that at the background and remote sites (i.e., Mt. Wuyi 4.7 ppbv and Mt. Waliguan 2.6 ppbv) ( Table S1).
The O 3 formation process depends on its precursors and related environmental conditions, while the photochemical reactions during the daytime are the basis for O 3 changes. Figure 2 shows the diurnal patterns of major trace gases and meteorological parameters during 20-29 September 2019. The O 3 concentration was maintained at relatively low levels from the night to 07:00 LT then rose and reached its maximum at around 17:00 LT. The O 3 peak in the afternoon was related to the accumulation of both local photochemical reactions and potential regional transport (including O 3 and its precursors in the upwind direction to the observation site), and detailed analysis will be shown in Sect. 3.3.2. The reduction of observed O 3 ( O 3 ) in the early morning rush hour caused by NO titration did not appear, verifying the impacts of regional transport (X. Zeren et al., 2019;Chen et al., 2020). Due to the photochemical reactions, the precursors of CO, NO x , and VOCs were consumed during the daytime and were accumulated during the nighttime with weak solar radiation. The diurnal patterns of VOCs, NO x , and CO were similar, with the highest concentrations at around 08:00 LT, which then decrease during 09:00-16:00 LT and increase at night, which is related to human activity emissions (especially vehicle exhaust) and the variations of the boundary layer (Elshorbany et al., 2009;Hu et al., 2020).

Atmospheric oxidation and radical chemistry
3.2.1 Atmospheric oxidation capacity (AOC) Figure 3 shows the time series of the model-calculated AOC during the O 3 pollution period. The AOC determines the removal rate of primary pollutants and the production rate of secondary pollutants and was the basis for reflecting atmospheric photochemical pollution (Geyer et al., 2001). AOC is calculated as the sum of oxidation rates of various primary pollutants (CO, NO x , and VOCs, etc.) by the major oxidants (i.e., OH, O 3 , NO 3 ) Xue et al., 2016Xue et al., , 2014. In this study, the average daytime AOC was 6.7 × 10 7 molecules cm −3 s −1 , and the daily maximum AOC was 1.3 × 10 8 molecules cm −3 s −1 , which was higher than that at rural sites, with much low pollution emissions in Berlin (1.4 × 10 7 molecules cm −3 s −1 ) and a regional background in Hong Kong (6.2 × 10 7 ), but lower than that in polluted cities, such as Santiago (3.2×10 8 molecules cm −3 s −1 ), due to the main limiting factor of the significant differences of pollutant concentrations among different sites (Li et al., 2018;Xue et al., 2016;Geyer et al., 2001;. In some urban regions, the concentrations of air pollutants were higher than those in Xiamen, but their AOC levels (Hong Kong: 1.3 × 10 8 molecules cm −3 s −1 ; Shanghai: 1.0 × 10 8 molecules cm −3 s −1 ) were comparable to or even lower compared with the AOC in Xiamen, which could be attributed to the relatively high solar radiation (Xue et al., 2016;Zhu et al., 2020) (detailed descriptions shown in Sect. 3.1). The results of AOC characteristics in different regions were decided by the precursor concentrations/types and photochemical environment.
According to the diurnal patterns of the AOC contributed by OH, O 3 , and NO 3 , the predominant oxidant was OH (90 ± 25 %) during the daytime, followed by NO 3 (8 ± 22 %) and O 3 (2 ± 3 %). Meanwhile, the diurnal characteristics of AOC were consistent with the profile of the model-calculated OH (Fig. S3) and the observed photolysis rate constants ( Fig. 1)  . Meanwhile, NO 3 (72 ± 9 %) played the most important role in the oxidant capability during the nighttime, followed by OH (20 ± 12 %) and O 3 (8 ± 1 %). In particular, the contribution of NO 3 to AOC reached a maximum of 80 % at around 18:00 LT, when the concentrations of O 3 and NO 2 were relatively high and the formation of NO 3 was accelerated (Fig. 2). In addition, solar radiation was weak during the nighttime, which resulted in the accumulation of NO 3 due to the ceasing of photolysis of NO 3 (Rollins et al., 2012;Chen et al., 2020). AOC contributed by O 3 was negligible, owing to the relatively low concentration of alkenes at the monitoring site ( Fig. 1 and Table 3), since O 3 contributed to the oxidation capacity through alkene  ozonolysis (Xue et al., 2016). In summary, the OH radical dominated the AOC, and it was necessary to further explore the partitioning of OH reactivity among different precursor groups.
The OH reactivity is an indicator for the OH chemical loss frequency, computed as the reaction rates of OH with CO, NO x , SO 2 , HONO, HNO 3 , HO 2 NO 2 , and VOCs (Whalley et al., 2016;Chen et al., 2020). Time series and diurnal patterns of model-calculated OH reactivity as well as its partitioning to the major reactants during the episode are shown in Figs. 4 and 5. The OH reactivity reached a peak (18.6 ± 4.8 s −1 ) at around 08:00 LT, mainly caused by the reaction of OH with NO x , since vehicles exhausted large amounts of NO x during rush hour. The average daily OH reactivity was 14.4 ± 3.83 s −1 , which was much lower than that in some polluted regions in Santiago (42 s −1 ) and the PRD (50 s −1 ), comparable to that at a rural site in Nashville (11 s −1 ) but higher than that at a mountain site in Pennsylva-nia (6 s −1 ) (Elshorbany et al., 2009;Lou et al., 2010;Kovacs et al., 2003;Ren et al., 2005). Figure 5 shows the diurnal variations and percentage of model-calculated OH reactivity to the major reactants during the episode. The OH reactivity exhibited a morning peak caused by the reactions of NO with OH, which should be ascribed to the freshly emitted urban plumes. Furthermore, OVOCs showed high fractions at around 12:00-18:00 LT, which were mainly owing to the transport of the regional air masses containing the abundant OVOCs, as well as the oxidation effect by strong photochemical processes. As shown in Fig. 5b, OVOCs (30 ± 8 %) , NO 2 (29 ± 8 %), and CO (25 ± 5 %) were the dominant contributors to OH reactivity, followed by alkanes (5 ± 3 %), aromatics (3 ± 2 %), alkenes (3 ± 1 %), and NO (2 ± 4 %). The high fraction of OVOCs and NO 2 in OH reactivity indicated the high degree of aged air mass and the intensive NO x emissions during the observation period, respectively (Li et al., 2018). However, the fraction of CO to OH reac-  tivity at our observation site was higher than that at an urban site in Los Angeles (Hansen et al., 2021), a rural site in Hong Kong (Li et al., 2018), and a mountain site in Colorado (Nakashima et al., 2014), comparable to that at the urban site of Shanghai (K. Zhang et al., 2021), which could be attributed to the abundant CO in our observation site. CO mainly comes from vehicle exhaust and the combustion of fossil fuels, and the observation site is a city with high density vehicles. Meanwhile, this pollution event was under the influence of the WPSH, which promoted the formation and accumulation of pollutants. The partitioning of OH reactivity elucidated the inherent photochemical processes and major reactants in southeastern China. High OH reactivity of OVOCs, NO 2 , and CO promotes the production of RO x radicals. Therefore, the investigation of detailed chemical budget of the RO x , recycling, and termination reaction is meaningful to figure out the complex atmospheric photochemistry (Li et al., 2018;Lou et al., 2010).

Radical chemistry
With the influence of NO x and VOCs, RO x radicals (OH, HO 2 , and RO 2 ) undergo efficient recycling and produce secondary pollutants, such as O 3 and OVOCs (Sheehy et al., 2010). Figure 6 shows the model-simulated OH, HO 2 , and RO 2 concentrations and their primary sources. The detailed time series of RO x concentrations and chemical budget are summarized in Fig. S3. Figure 6a shows the diurnal variations of the simulated OH, HO 2 , and RO 2 . The maximum daily values of OH, HO 2 , and RO 2 concentrations were 2.4×10 7 , 7.9×10 8 , and 4.7×10 8 molecules cm −3 , with daytime average concentrations of 7.4 × 10 6 , 2.4 × 10 8 , and 1.7 × 10 8 molecules cm −3 , respectively. Modelpredicted concentrations of OH in Xiamen were higher than that in the Yellow River Delta (an oil field with high VOC emission), while the concentrations of HO 2 and RO 2 showed reverse trends . The RO x recycling of OH → RO 2 was mainly controlled by the reaction of OH + VOCs, and the RO 2 → HO 2 and HO 2 → OH depended on the reactions with NO (Fig. 7). Combined with the ratio of VOCs / NO x (1.1 ± 0.4), it was established that NO x would not be the limiting factor in the radical recycling processes. Hence, efficient conversions of radical propagation of RO 2 + NO → HO 2 and HO 2 + NO → OH were expected, and the OH + VOCs → RO 2 reaction was the rate-dependent step of the radical recycling in our study. The detailed radical chemistry is further discussed in the following.  Figure 6b shows the daytime average diurnal variations of primary OH sources. HONO photolysis reached a maximum of 2.7 ppb h −1 at around 08:00 LT, which occupied 56 ± 19 % of the total OH primary production rates. The second source of OH primary production was O 3 photolysis (42 ± 21 %), and the percentages of O 3 + VOC and OVOC photolysis were minor. The highest HONO photolysis rate appeared in the morning rush hour, suggesting the influence of vehicle emissions and nocturnal accumulation of HONO . Considering the radical recycling, the reaction of HO 2 + NO (8.0 ± 6.2 ppb h −1 ) dominated the total production of OH (Fig. S3a). Meanwhile, OH-initiated oxidations of VOCs (4.9 ± 3.3 ppb h −1 ) consumed OH most during the daytime, followed by OH + CO (2.6 ± 1.9 ppb h −1 ), OH + NO 2 (2.4 ± 1.1 ppb h −1 ), OH + NO (0.6 ± 0.3 ppb h −1 ), and OH + O 3 (0.2 ± 0.1 ppb h −1 ).
In this study, HCHO photolysis was identified as the most important source for HO 2 primary formation, with an average production rate of 1.1 ± 0.6 ppb h −1 (Fig. 6c), followed by the other OVOC photolysis (0.4 ± 0.2 ppb h −1 ). The rate of OVOC photolysis in Xiamen was much lower than that in some megacities, such as Beijing  and Hong Kong (Xue et al., 2016). The reactions of OH + CO (2.6 ± 2.2 ppb h −1 ) and RO 2 + NO (2.5 ± 1.5 ppb h −1 ) were also important sources of HO 2 (Fig. S3b). The main sink of HO 2 was HO 2 + NO (7.9 ± 6.2 ppb h −1 ), while the loss rates of HO 2 + HO 2 and HO 2 + RO 2 were negligible.
In Fig. 6d, OVOC photolysis contributed most to primary RO 2 production with a rate of 0.5 ± 0.2 ppb h −1 , accounting for 85 ± 20 % of total RO 2 primary production. The reaction of unsaturated VOCs and NO 3 was the second important source, accounting for 11 ± 18 % of the total primary RO 2 . The radical recycling rate of OH + VOCs was 8.4 times higher than the sum of RO 2 primary production. The consumption reaction of RO 2 was mainly caused by RO 2 + NO (3.7 ± 2.9 ppb h −1 ), and the cross-reactions by RO x themselves were limited.
The daytime average RO x budget and its recycling are also demonstrated (Fig. 7). For the RO x primary sources, the photolysis of HONO (33 ± 14 %), O 3 (25 ± 13 %), HCHO (20 ± 5 %), and other OVOCs (17 ± 2 %) was a major contributor. For RO x recycling, CO and VOCs reacted with OH, producing HO 2 and RO 2 , with average rates of 4.0 and 4.4 ppbv h −1 , respectively. RO 2 + NO and HO 2 + NO enhanced the production of RO (3.6 ppbv h −1 ) and OH (7.9 ppbv h −1 ), with O 3 formed as a by-product. For the termination processes, the reaction rates of RO x and NO x were approximately 2-5 times faster than the cross-reaction rates of RO x .

Chemical budget and sensitivity analysis of O 3
production The in situ O 3 production mechanism was examined, and the detailed reaction weights are shown in Fig. 8. The daytime rate of HO 2 + NO was 7.9 ± 6.2 ppb h −1 , accounting for 68 ± 4 % of the total O 3 production. This result was consistent with that in Sect. 3.2.2. The OH radical was the initiator of O 3 photochemical formation, and the source of OH from HO 2 + NO was also the dominant pathway to produce O 3 (Liu et al., 2020c). The second pathway of O 3 production was RO 2 + NO (3.6 ± 2.0 ppb h −1 ). The reaction of RO 2 + NO contained more than 1000 types of RO 2 radicals, and the pathway of CH 3 O 2 + NO (34 ± 6 %) contributed most among them. In contrast, the contributors of O 3 destruction were OH+NO 2 (61 ± 18 %), followed by O 3 photolysis (18 ± 9 %), RO 2 + NO 2 (9 ± 10 %), O 3 + HO 2 (4 ± 4 %), and O 3 + OH (4 ± 2 %), while the other pathways of O 3 + VOCs as well as NO 3 + VOCs contributed limitedly. In addition, the net O 3 production (9.1 ± 5.7 ppb h −1 ) in Xiamen was ∼ 2-5 times lower than that derived from the metropolis of Shanghai (26 ppb h −1 ), Lanzhou (23 ppb h −1 ), and Guangzhou (50 ppb h −1 ), reflecting the influence of O 3 precursor emissions and photochemical conditions (Xue et al., 2014).
In order to investigate the O 3 control strategies during this multi-day O 3 pollution event, the scenario analysis with reduction by 0 %-100 % at intervals of 5 % for the reduction of anthropogenic VOCs ( S(VOCs) / S(VOCs)) and NO x ( S(NO x ) / S(NO x )) was conducted using the OBM-MCM. According to the Empirical Kinetic Modeling Approach (EKMA) and scenario analysis, O 3 formation was in the NO titration regime (Fig. 10), in accordance with the RIR analysis results, which meant VOCs should be reduced to effectively control O 3 during the O 3 pollution event. The maximum value of MDA8h O 3 during the monitor-ing period was 85 ppbv, exceeding the national air quality standard of 75 ppbv for O 3 by 13 %. Hence, the O 3 reductions of 5 %, 10 %, and 15 % were set to discuss the reduction schemes of anthropogenic VOCs and NO x . As shown in Fig. 10b, the 5 % control target is achieved as follows: (1) S(VOCs) are reduced by 15 %, while S(NO x ) remain unchanged; (2) S(VOCs) are reduced by more than 35 %; and (3) S(NO x ) reduction is higher than 60 %. The first scenario of just reducing VOC emission was the most cost-efficient way for short-term or emergency control of O 3 . However, NO x , as important precursors of PM 2.5 , need to be reduced according to the long-term multi-pollutant control air quality improvement plan in China; thus the second scenario is a more practical and reasonable way to control air pollution. The 10 % O 3 control target was achieved by the 45 % reduction of S(VOCs) and the S(NO x ) keeping the original emission. In view of the long-term control strategy of NO x and VOCs, reduction of S(VOCs) by 55 % and 80 % could decrease O 3 concentrations by 10 % and 15 %, respectively. Although VOC and NO x control measures were drastically implemented, it is still challenging to achieve the 15 % O 3 control goals in urban areas with relatively low precursor emissions. As the episode is a typical pollution process in the coastal region, the research results might act as a reference for the policy makers. Meanwhile, as the O 3 sensitivity changed under the implementation of control measures, it is necessary to adjust the reduction of VOC and NO x policies in a timely manner.

O 3 from local photochemical production and regional transport
Previous studies have found that the variation of O 3 mixing ratios was mainly influenced by chemical and physical processes (Xue et al., 2014;Tan et al., 2018). Figure 11 shows the time series of O 3 accumulation and contributions from local photochemical production and regional transport. The observed rate of change in O 3 (R meas ) was calculated by the derivative of the observed O 3 concentrations (R meas = d(O 3 )/dt). The local O 3 production (R chem ) was calculated by Eq. (3) and computed hourly by the OBM as described in Sect. 2.2. The physical processes (R tran ) were calculated by the equation of R tran = R meas − R chem , including horizontal and/or vertical transport and dry deposition dilution mixing. Many studies showed that the impacts of dry deposition were minor; thus the differences between observed O 3 changes and local O 3 production were mainly caused by the regional transport (note that the effect of atmospheric mixing was also included in this term) and could reasonably quantify the contributions of regional transport at our observation site Chen et al., 2020). The positive values of R tran represented the O 3 import of regional transport, while the negative values indicated the O 3 export and deposition. We quantified the contributions of local photochemical for-  . O 3 accumulation and contributions from local photochemical production and regional transport. R chem , R tran , and R meas in the legend represent local O 3 photochemical production, regional transport, and observed O 3 formation rate, respectively. mation and regional transport to the observed O 3 and figured out the reasons for the O 3 pollution process.
As shown in Fig. 11, two regular O 3 import phenomena with positive values of R tran were observed, and the curve of the R meas showed the "M" trend during the daytime. The first transient intense O 3 import happened in the early morning (at around 06:00-09:00), leading to a rapid increase in O 3 concentration, which was mainly attributed to the residual ozone from the day before. The O 3 export was remarkable at around 10:00-16:00, indicating the potential impacts on air quality in downwind areas. Generally, the maximum daily value of O 3 at this observation site appeared at around 15:00 LT without regional transport (Wu et al., 2019). In Fig. 11, we found that the O 3 concentrations showed two peaks at around 15:00 and 17:00 LT, and O 3 concentrations rose slowly or even decreased firstly and then increased between the two peaks. Under these circumstances, the local photochemical production kept producing O 3 , but the decreased O 3 concentrations could be attributed to the favorable atmospheric conditions in diluting pollutants (O 3 export). When the nearsurface wind direction changed from northeast to southeast, the second O 3 import phenomenon occurred in the afternoon (16:00-19:00 LT) on 4 d (20,25,27,and 29 September). Due to the persistence of R tran in the afternoon, the daily maximum O 3 values appeared at around 17:00 LT. The conditions of a southeast wind direction in the downtown area with a high density of vehicles would cause O 3 and its precursors to be transmitted to our observation site, consistent with the diurnal patterns of NO 2 , OVOCs, alkanes, and aromatics in the early morning and afternoon (Fig. 2) and matching with the "M" trend of R meas . This result indicated that the sudden changes of near-surface winds corresponded to the variation in the transport of the urban plume.
According to the synoptic situations and meteorological parameters (Figs. 1, S4 and 12), the environmental conditions also favored the O 3 pollution process during the observation periods. The contribution of R chem (daily maximum: ranged from 10.2 to 19.1 ppb h −1 ) during the day-  time was observed (Fig. 11). In Fig. S4a-c, the monitoring site was continuously affected by the northerly airflow with high O 3 and its precursors (from an industrial city adjacent to Xiamen, Quanzhou, or polluted regions of the Yangtze River Delta), due to Typhoon Tapah from 20 to 22 September 2019. The transport of O 3 import appeared on 21 September (7.1 ± 7.0 ppb h −1 ), which resulted in the accumulation of O 3 (MDA8h O 3 : 85 ppbv) on 22 September. When the influence of typhoon disappeared, the direction of airflow turned from northerly into southwesterly with humid and warm conditions at 500 hPa (Fig. S4d), and the surface wind on 23 September was affected by the control of the cold northerly airflow (Fig. S4e and f). Meteorological conditions including continental high pressure during 23 to 27 September were favorable to the accumulation of air pollutants (Fig. 12). The isoline of 5880 gpm moving from north to the Yangtze River ( Fig. 12a and b) indicated the strengthened subtropical high pressure during 23-27 September 2019, which carried high temperature, low RH, and stagnant weather conditions, and the transport rate of O 3 export (5.4 ± 3.4 ppb h −1 ) on 24-26 September was lower than that on other days (6.3 ± 4.0 ppb h −1 ). Favorable meteorological conditions significantly affected the formation and accumulation of O 3 , and we chose five meteorological parameters (i.e. UV, T , RH, P and WS) to quantify the complex nonlinear relationships between O 3 and its influencing factors based on a generalized additive model (GAM) (Hua et al., 2021). Table S3 showed that the factors had significant nonlinear impacts on O 3 concentration changes at the level of P value < 0.01 and degrees of freedom > 1, indicating that each influencing factor has statistical significance as an explanatory variable. According to the F values reflecting the importance of the influencing factors, the order of the explanatory variables was RH (40.1) > WS (26.9) > T (10.9) > P (3.9) > UV (3.0). Response curves of O 3 concentration to explanatory factors are presented in Fig. 13. The O 3 concentration showed a remarkable upward trend until the UV increased to 17 W m −2 , then it changed little with the fluctuation of UV (Fig. 13a). In previous studies, UV had a significant positive correlation with O 3 concentrations (Ma et al., 2020), and these results showed the regional transport impacts on O 3 formation in our study. The RH and T had negative and positive correlations with O 3 concentrations, respectively ( Fig. 13b and c). The increase of wind speed was favorable for O 3 regional transport (Fig. 13d). The influence of atmospheric pressure on O 3 seemed to be irregular and minor, which could be ignored (Fig. 13e). Hence, under the combined effects of favorable photochemical reaction conditions and strengthened WPSH, the MDA8h O 3 exceeded the standard of 75 ppbv during 24-26 September. Previous studies had found that severe multi-day O 3 pollution appeared under the WPSH control (H. . Overall, the results indicate that the three conditions of local photochemical production, synoptic situations, and regional transport played very important roles in the pollution event.

Conclusions
In the present study, we analyzed a typical high-O 3 event during 20-29 September 2019 in a coastal city of southeastern China. We clarified the characteristics of AOC, OH reactivity, and radical chemistry, as well as O 3 formation mechanisms using the OBM-MCM model. The predominant oxidant for AOC during the daytime and nighttime was the OH and NO 3 , respectively. During the period of O 3 pollution process, OVOCs, NO 2, , and CO consumed OH the most. Meanwhile, the photolysis of HONO, O 3 , HCHO, and other OVOCs was a major source of RO x , which played the initiation role in atmospheric oxidation processes. The radical termination reactions were governed by cross-reactions between RO x and NO x . The RIR and EKMA results showed that the O 3 formation in autumn in the coastal city was VOCs-sensitive, and the VOCs were the limited factor of radical recycling and O 3 formation. The reduced emissions of aromatics, alkenes, and alkanes with ≥ 4 carbons were benefit for ozone pollution control. The three conditions of local photochemical production, synoptic situations, and regional transport played very important roles in the pollution event.
Overall, the results clarified the O 3 pollution process with relatively low local precursor emissions and implied the fact that O 3 pollution control in coastal cities needs to be studied further.
Code and data availability. The observation data at this site are available from the authors upon request.
Author contributions. TL and YH contributed equally to this work. JC and LikX designed and revised the manuscript. TL collected the data and contributed to the data analysis. TL and YH performed chemical modeling analyses of OBM-MCM and wrote the paper. JC supported funding of observation and research. LinX, ML, CY, YD, YZ, and MZ contributed to discussions of results. ZH and HW provided meteorological conditions in Xiamen.

Competing interests.
The contact author has declared that neither they nor their co-authors have any competing interests.
Disclaimer. Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
of Sciences (XDPB1903), the FJIRSM&IUE Joint Research Fund (RHZX-2019-006), the Center for Excellence in Regional Atmospheric Environment, CAS (E0L1B20201), the Xiamen Youth Innovation Fund Project (3502Z20206094), the foreign cooperation project of Fujian Province (2020I0038), and Xiamen Atmospheric Environment Observation and Research Station of Fujian Province.
Financial support. This research has been supported by the CAS Center for Excellence in Regional Atmospheric Environment Review statement. This paper was edited by Timothy Bertram and reviewed by two anonymous referees.