Factors dominating 3-dimensional ozone distribution during high tropospheric ozone period

Environmental Pollution xxx (2017) 1e10

Contents lists available at ScienceDirect

Environmental Pollution journal homepage: www.elsevier.com/locate/envpol

Factors dominating 3-dimensional ozone distribution during high tropospheric ozone period* Xiaoyang Chen a, Yiming Liu a, Anqi Lai a, Shuangshuang Han b, c, Qi Fan a, *, Xuemei Wang a, d, Zhenhao Ling a, Fuxiang Huang b, Shaojia Fan a a

School of Atmospheric Sciences/Guangdong Province Key Laboratory for Climate Change and Natural Disaster Studies, Sun Yat-sen University, Guangzhou 510275, China National Satellite Meteorological Center, Beijing 100081, China c School of Earth Sciences and Resources, China University of Geosciences, Beijing 100083, China d Institute for Environmental and Climate Research, Jinan University, Guangzhou 510632, China b

a r t i c l e i n f o

a b s t r a c t

Article history: Received 19 February 2017 Received in revised form 1 September 2017 Accepted 6 September 2017 Available online xxx

Data from an in situ monitoring network and five ozone sondes are analysed during August of 2012, and a high tropospheric ozone episode is observed around the 8th of AUG. The Community Multi-scale Air Quality (CMAQ) model and its process analysis tool were used to study factors and mechanisms for high ozone mixing ratio at different levels of ozone vertical profiles. A sensitive scenario without chemical initial and boundary conditions (ICBCs) from MOZART4-GEOS5 was applied to study the impact of stratosphere-troposphere exchange (STE) on vertical ozone. The simulation results indicated that the first high ozone peak near the tropopause was dominated by STE. Results from process analysis showed that: in the urban area, the second peak at approximately 2 km above ground height was mainly caused by local photochemical production. The third peak (near surface) was mainly caused by the upwind transportation from the suburban/rural areas; in the suburban/rural areas, local photochemical production of ozone dominated the high ozone mixing ratio from the surface to approximately 3 km height. Furthermore, the capability of indicators to distinguish O3-precursor sensitivity along the vertical O3 profiles was investigated. Two sensitive scenarios, which had cut 30% anthropogenic NOX or VOC emissions, showed that O3-precursor indicators, specifically the ratios of O3/NOy, H2O2/HNO3 or H2O2/ NOZ, could partly distinguish the O3-precursor sensitivity between VOCs-sensitive and NOx-sensitive along the vertical profiles. In urban area, the O3-precursor relationship transferred from VOCssensitive within the boundary layer to NOx-sensitive at approximately 1e3 km above ground height, further confirming the dominant roles of transportation and photochemical production in high O3 peaks at the near-ground layer and 2 km above ground height, respectively. © 2017 Published by Elsevier Ltd.

Keywords: Vertical ozone analysis WRF/CMAQ Process analysis Vertical O3-precursor sensitivity

1. Introduction China is widely undergoing continuous ozone (O3) increases in recent years. Research on key regions, such as the Northern China Plain (NCP), Yangtze River Delta (YRD) and the Pearl River Delta (PRD) including Hong Kong, has observed increasing trends of surface ozone (Li et al., 2014a; Ma et al., 2016; Tang et al., 2008; Wang et al., 2009; Xu et al., 2008; Xue et al., 2014). Continuous data, based on a monitoring network set up by the Ministry of

*

This paper has been recommended for acceptance by Dr. Hageman Kimberly Jill. * Corresponding author. E-mail address: [email protected] (Q. Fan).

Environmental Protection (MEP) since 2013, show that mean surface ozone concentrations are increasing, while other major pollutants (SO2, NOX and suspended particulates) are more or less decreasing (www.mep.gov.cn) (Wang et al., 2016). As for the PRD region including Hong Kong, surface ozone concentrations increased at a rate of 0.88 mg/m3 per year from 2006 to 2014 (Guangdong-Hong Kong-Macao Pearl River Delta Regional Air Quality Monitoring Network: A Report of Monitoring Results in 2014). However, annual concentrations of SO2, NO2 and respirable suspended particulates (PM10) decrease during the same period from 47 mg/m3, 46 mg/m3 and 74 mg/m3 to 16 mg/m3, 37 mg/m3 and 56 mg/m3, respectively. The increasing pattern is not only found at the surface but is also observed vertically (Sun et al., 2016). Ozone is

https://doi.org/10.1016/j.envpol.2017.09.017 0269-7491/© 2017 Published by Elsevier Ltd.

Please cite this article in press as: Chen, X., et al., Factors dominating 3-dimensional ozone distribution during high tropospheric ozone period, Environmental Pollution (2017), https://doi.org/10.1016/j.envpol.2017.09.017

2

X. Chen et al. / Environmental Pollution xxx (2017) 1e10

relatively becoming a more severe problem in many regions and may even be a tough problem for China (Wang et al., 2013). Many factors can largely affect the vertical ozone distribution. Evidence has been revealed by observation experiments (Hocking et al., 2007) and numerical studies (Li and Rappenglück, 2014) that stratosphere-troposphere exchange (STE) of ozone plays an important role in mixing the ozone of the upper troposphere and lower stratosphere (UTLS) (Holton, 1995). Typhoons can largely affect the strength of the stratosphere-troposphere exchange, especially in eastern China and the western North Pacific Ocean. Within the typhoon's large-scale circulation, downward propagation of stratospheric ozone to the surface induces an elevated surface ozone mixing ratio (Jiang et al., 2015). As for the upper troposphere and lower stratosphere from approximately 200 hPa to 50 hPa, which lie in the top most levels of regional air quality models like the CMAQ and WRF-CHEM, a typhoon also redistributes the vertical ozone through its convective system (Fu et al., 2013). Therefore, O3 at the surface and above ground can affect each other, suggesting the great necessity for the investigation of vertical O3 distributions. In addition to stratospheretroposphere exchange, other processes including local photochemical formation and atmospheric transportation (horizontal transport and vertical convection) could lead to high O3 mixing ratios along the vertical profiles (Fiore, 2002; Jacob, 1999; Tang et al., 2008; Godowitch et al., 2015; Park et al., 2014), making it a complicated issue in tropospheric O3 studies. The relative contributions of the above influenced factors of vertical O3 distributions in different environments are varied (Wang et al., 2010). By applying some O3-precursor sensitivity indicators, Wang et al. (2011) found that it is usually VOCs-sensitive for megacity area and NOX-sensitive for rural areas. However, research shows that there are limitations and accuracy problems in using indicators to investigate O3-precursor relationships. For example, previous research mainly focused on the surface layer and the transition values of different indicators, i.e., the ratios of O3/ NOz, HCHO/NOy, H2O2/HNO3 (Sillman, 1995), O3/NOy, O3/HNO3 and H2O2/NOz (Sillman et al., 1997). And these indicators are more nez and varied in different areas (Castell et al., 2009; Jime Baldasano, 2004; Tonnesen and Dennis, 2000). In the PRD region, previous research showed that the ratio of H2O2/HNO3 has good performances (Lam et al., 2005; Wang et al., 2011). In this study, we apply some of the indicators to analyse whether they are capable along the vertical direction or not, instead of horizontal dimensions like most other research. The transition values of those indicators are localized for our key region. Accompanied by the process analysis tool of CMAQ (Fan et al., 2015a, 2015b; Liu et al., 2015), we can distinguish vertical layers between O3-producing and O3consuming ones. The top layer of regional chemical transport models is usually set to 50 hPa or 100 hPa, inducing uncertainties in simulating the stratosphere-troposphere exchange effect, especially for regional chemical transport models. Therefore, challenges often remain in simulating the stratosphere-troposphere exchange effect. Ozone near the upper troposphere and lower stratosphere cannot be well simulated by regional chemical transport models. The WRF/CMAQ modelling system, which is used in this research (Byun and Schere, 2006), also faces the same challenge of modelling upper air ozone with its default initial conditions and boundary conditions (ICBCs) (Tang et al., 2008). To improve the model performance of simulating the stratosphere-troposphere exchange of ozone, the results of global chemical models can be applied to our regional air quality model. GEOS-CHEM and MOZART4-GEOS5 have been widely used in recent years. A toolbox called “geos2cmaq” (http://wiki.seas. harvard.edu/geos-chem/index.php/GEOS-Chem_to_CMAQv5.0)

and its updated version pncglobal2cmaq (Henderson et al., 2014) were developed by the GEOS-CHEM community to fit the GEOSCHEM model outputs for initial and boundary conditions of CMAQ. There is also a linkage tool called “mozart2cmaq” for MOZART included in the CAMx package (www.camx.com). Previous studies found that ozone profiles could be simulated more reasonably by importing either the GEOS-CHEM or MOZART lateral boundary conditions (Li and Rappenglück, 2014; Tang et al., 2008). In this study, the simulation results of the WRF/CMAQ model imported with the above ICBCs further confirm the importance of the stratospheric O3 boundary condition in modelling vertical O3 profiles. Furthermore, the “ozone mixing ratios tongue” in the upper and middle troposphere near 9 km above ground level (AGL), which is probably underestimated in global chemical transport model (Zhang et al., 2012), was well captured in this study. 2. Methodology 2.1. In-situ monitoring data and ozone sondes data Surface meteorological observations, including hourly temperature and wind speed, as well as relative humidity, are obtained by 74 Guangdong Meteorological Observatory (GMO) stations. Environmental monitoring in-situ data are obtained by the Pearl River Delta Regional Air Quality Monitoring Network (PRDRAQMN) which consists of stations in the three regions of Guangdong, Hong Kong and Macao. SO2, NO2, O3 and PM10 concentration data are available from all stations in this monitoring network during the study period - August 2012, Fine Suspended Particulates (PM2.5), NOX and CO are additionally available from monitoring stations in Hong Kong. Five ozone sondes were launched at Kings Park by the Hong Kong Observatory during August 2012, on the 1st, 8th, 14th, 22nd and 29th. The locations of these stations are presented in Fig. S1. In this study, 9 PRDRAQM stations are mainly used: 6 stations in the Guangdong Province are in black, and 3 in Hong Kong are marked in red. 2.2. Models In this study, we use the Weather Research and Forecasting model and the Community Multiscale Air Quality model (WRF/ CMAQ) to simulate such a high ozone period. The domain of the WRF/CMAQ modelling system covers an area from 94.4
E to 133.3
E and 11.9
N to 44.4
N. The spatial resolution is 27 km, and there are 40 vertical levels with a model top height at 50 hPa. Increasing vertical resolution can somehow help the numerical model achieve a better performance in simulating boundary layer structures and meteorological factors. Approximately 16 levels are set within the boundary layer in this study in order to offer a better driven field for CMAQ chemical modelling. In CMAQ version 5.1, some modified chemistry options are updated. The aero6i (ae6i) for aerosol chemistry has been updated from the aero6 (ae6) of former versions. A more detailed option (AQCHEM-KMTI) for aqueous-phase chemistry is updated as well (Pye et al., 2015). In our study, the chemistry mechanism option is set to SAPRC07tic in order to get more detailed gas-phase reaction formulas, which are important for ozone modelling. Other settings for our WRF/CMAQ model are listed in Table S1. In our study, the process analysis (PA) method is used for providing different factors’ contributions to ozone generation and consumption. Vertical advections, horizontal advections, vertical diffusions, horizontal diffusions, cloud effects, chemical reactions and dry depositions are the 7 PAs that account for ozone changes. PA values show how much the concentrations are contributed to by those 7 PAs during the concerned period.

Please cite this article in press as: Chen, X., et al., Factors dominating 3-dimensional ozone distribution during high tropospheric ozone period, Environmental Pollution (2017), https://doi.org/10.1016/j.envpol.2017.09.017

X. Chen et al. / Environmental Pollution xxx (2017) 1e10

Three anthropogenic emission inventories are mainly applied in this study. One is the Multi-resolution Emission Inventory for China for the year 2012 (MEIC2012: www.meicmodel.org) (Li et al., 2014b). As there are no data for Hong Kong in MEIC2012, the Hong Kong Air Pollutant Emission Inventory compiled by The Hong Kong Environmental Protection Department (HKEPD) is applied to modify emissions of the Hong Kong area based on INTEX-B (Zhang et al., 2009). MEGAN 2.1 (Guenther et al., 2012) is applied for biogenic emissions. However, because some chemical mechanism updates are applied in the recent CMAQ version 5.1, ordinary emission inventories cannot be input directly. Therefore, some modifications have to be used: the reallocation of species for the SAPRC07tic mechanism is one such important part (Carter, 2015). Four model scenarios are deployed, including one control run and three sensitive scenarios (Table S2). The BASE-CTLRUN runs under the options mentioned in Table S1. To separate the effect of stratosphere-troposphere ozone exchange, the CASE-noICBC sensitive scenario runs under the same options of BASE-CTLRUN but without importing initial and boundary conditions from MOZART4/GEOS-5. The other two sensitive scenarios (CASE-NOxCUT and CASE-VOCCUT) aim to understand the O3-precursor sensitivity and photochemical reactivity along the vertical layers. They also run under the same settings of BASE-CTLRUN but cut 30% anthropogenic NOx or VOC emissions, respectively, in the whole modelling domain. Cutting anthropogenic NOx or VOC emissions at a percentage ranging from 25% to 35% is widely acceptable for O3-prenez and cursor sensitivity researches (Castell et al., 2009; Jime Baldasano, 2004; Liang et al., 2006; Wang et al., 2010). Accompanied by photochemical indicators, the differences between BASECTLRUN and the two emission-limited scenarios can tell the importance and contributions of photochemical formation of ozone at different layers. 3. Results and discussion 3.1. Variations of tropospheric and surface ozone in August We use ozone columns from the surface to 3 km above ground levels (AGL) to show ozone concentrations at low tropospheric layers. These tropospheric ozone columns and surface ozone concentrations are based on ozone sonde observations and in situ monitoring data, respectively. The sonde on the 8th of AUG

3

captures the highest boundary layer ozone concentration among these 5 ozone sondes observations, with 28.53 Dobson units (DU) (Fig. 1). Although daily average surface ozone concentration synchronously reached highest (226.8 mg/m3) on the 8th of AUG in Hong Kong, obvious differences between the 5 tropospheric ozone columns (in bars) were found. As for the whole PRD region (curve with crossings), surface ozone concentrations are not as varied as the ozone columns’ differences around these 5 launching days. Structures of vertical ozone on the 8th of August need to be further studied. We use our WRF/CMAQ modelling system to simulate August of 2012. The structure of the vertical ozone on the 8th of AUG will be further discussed in later sections. The other four O3 profiles are shown in Fig. S2. O3 profiles can be well reproduced by the WRF/ CMAQ modelling system in general. On the 22nd of AUG, the model overestimates O3 because there is precipitation observed by many meteorological stations in the PRD but the model does not reproduce such precipitation well. Meteorological observations from 74 GMO stations are used in the evaluation of model performances. Observation data from PRDRAQMN stations are used to evaluate the model performances as well (Fig. 2). The WRF/CMAQ modelling system can well reproduce the surface ozone concentrations during the whole of August, especially in Hong Kong. Five high surface ozone periods around the 5 ozone sonde launching days are well captured by the model, especially the 8th of AUG one. Index of agreement (IOA: Willmott, 1981) ranges from 0.90 to 0.92 in 3 Hong Kong stations. Detailed statistics are shown in Table S3, and simulations with the process analysis focusing in such period and region are held for further study. 3.2. High tropospheric ozone period during typhoon HAIKUI The high tropospheric ozone event on the 8th of AUG happened during typhoon HAIKUI entering the East China Sea and then landing on eastern China (Fig. S3b). The PRD and Hong Kong were located in HAIKUI's outer circulation (Fig. S3a). When HAIKUI was landing on eastern China, the PRD was mainly controlled by a north wind, with a southwest wind around Hong Kong. The simulated landing location of the typhoon and surface wind field well match the observation (Fig. S3), suggesting that the model simulation is suitable for further study. Fig. 3 shows the simulated horizontal ozone distribution when HAIKUI was landing.

Fig. 1. Tropospheric ozone concentrations (from surface to 3 km height) observed by 5 Hong Kong ozone sondes in AUG (bar) and the daily highest 8-h average concentrations of 3 Hong Kong stations (plot curve with circles) and all 9 PRD stations (plot curve with crossings).

Please cite this article in press as: Chen, X., et al., Factors dominating 3-dimensional ozone distribution during high tropospheric ozone period, Environmental Pollution (2017), https://doi.org/10.1016/j.envpol.2017.09.017

4

X. Chen et al. / Environmental Pollution xxx (2017) 1e10

Fig. 2. O3 concentrations comparison of model results and observation during August 2012.

Fig. 3. (a) Simulated surface ozone mixing ratios in ppb; (b) surface wind direction in vectors and speed in filled colors. Both (a) and (b) are at 15:00LST on the 7th of AUG. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

High ozone mixing ratios were found in the outer typhoon circulation, while low levels were present inside the typhoon cloud system. The wind field clearly indicated that surface O3 in downwind areas (i.e., Hong Kong) could be influenced by the transport of O3 and its precursors from upwind areas. Therefore, by considering the location signature, 3 sites in Hong Kong (i.e., Tung Chung, Sha Tin and Tap Mun) were selected for further investigation into the influence of transport. Tung Chung is located to the west of downtown Hong Kong (Fig. S1). In this episode, Tung Chung is also located in the upwind direction of the Hong Kong urban area; Sha Tin, an urban site, is located in the north of the Hong Kong urban area and not very far away from the Kings Park ozone sonde station. Tap Mun is located downwind of Tung Chung and Sha Tin, on an island far away from the Hong Kong urban area. These three sites are nearly in line along

the southwest wind during the severe ozone period, providing us ideal representation for in situ observations. 3.3. Surface ozone analysis Fig. 4 shows the time series of surface ozone concentrations and the contributions of different processes to those concentrations at Tung Chung (upwind suburban site), Sha Tin (urban site) and Tap Mun (downwind rural site). During this episode event, there were two phases: on the 6th of AUG, O3 concentrations were low with the maximum under 100 mg/m3 for all 3 sites; as HAIKUI moved and landed on Eastern China, meteorological fields rapidly changed, especially in the wind field, which rotated clockwise from a southeast to west wind (Fig. S4). The temperature increased (not shown) because the PRD was controlled by the outer typhoon

Please cite this article in press as: Chen, X., et al., Factors dominating 3-dimensional ozone distribution during high tropospheric ozone period, Environmental Pollution (2017), https://doi.org/10.1016/j.envpol.2017.09.017

X. Chen et al. / Environmental Pollution xxx (2017) 1e10

5

Fig. 4. Process analysis and ozone time series of Tung Chung, Sha Tin and Tap Mun. Black curves are in situ monitoring ozone values, purple curves are model simulated ozone, and red curves are the total contributions of 7 processes for ozone. Colourful bars stand for different factors' contributions to ozone increasing (positive) and decreasing (negative): ZADV stands for vertical advection, HADV stands for horizontal advection, HDIF stands for horizontal diffusion, VDIF stands for vertical diffusion, CLDS stands for cloud convections, CHEM stands for gas-phase chemical reactions and DDEP stands for dry deposition. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

circulation and a clear sky. Such changes in meteorological fields induced a high ozone episode on the 7th-8th of August. Ozone concentrations around the entire Hong Kong region were extremely high, with the highest levels exceeding 300 mg/m3 in both rural/suburban and urban sites. Previous studies have claimed that gas phase chemical reactions make different contributions to ambient O3 in suburban/rural and urban areas due to variations of precursors and meteorological conditions (Li et al., 2013; Yu et al., 2014). Particularly in the PRD and Hong Kong, when anthropogenic emissions are always large in urban areas, gas phase chemical reactions will make negative contributions to ozone formation caused by NO titration (Guo et al., 2013; Wang et al., 2015). Thus, although similar high O3 levels were found at different sites, the mechanisms may be different because of the variations of precursors and meteorological conditions between urban and suburban/rural sites. For Tung Chung, it is clear that the gas phase chemical reactions produce ozone and cause ozone concentrations to rapidly increase during the daytime. During the nighttime, ozone is consumed more than it is produced by the total effect of chemical reactions due to a lack of solar radiation. However, as for the urban site, Sha Tin, gas phase chemical reactions always make negative contributions to the surface ozone because of high NOX emissions around the central urban area. Vertical diffusion and horizontal advection dominate the high ozone peak during the daytime, indicating that such a surface high ozone event of the Hong Kong urban area is mainly due to ozone from upper air and upwind suburban/rural areas like Tung Chung. Fig. 5 shows the contribution of total gas-phase chemical reactions to surface ozone (Fig. 5a) and the average mixing rations of OX (OX ¼ O3 þ NO2) in the PRD region (Fig. 5b). It is evident that local-precursor-produced O3 has decreased in recent years in Hong Kong. However, regional background precursors and O3 have increased due to the rapid development of Pearl River Delta metropolitans (Xue et al., 2014). Tung Chung is affected considerably by

such cross boundary transportation (Fig. 5). It is obvious that Tung Chung is largely affected by upwind pollutant and precursor transportation from central PRD regions (such as Guangzhou, Foshan and Zhongshan). Fig. 5a shows that chemical reactions contribute more than 60 ppb in Tung Chung to the change in the ozone mixing ratio from the 6th-18:00 LST to the 8th-18:00 LST. Meanwhile, high emissions of NO in central PRD cities will be partly oxidized into NO2. It forms ozone-consuming areas in the upwind direction of Tung Chung like Guangzhou and Zhongshan. Produced NO2, as well as emitted NO2, transports to downwind of Tung Chung, making Tung Chung a precursor-rich environment for ozone production. Therefore, large amounts of O3 are produced in Tung Chung by precursors coming from central PRD regions, causing a high ozone and OX mixing ratio near the surface around the Tung Chung site (Fig. 5b). Such ambient air with high ozone mixing ratios is transported downwind to other areas, the Sha Tin and Tap Mun sites in this case, causing synchronous ozone peaks in Sha Tin and Tap Mun but with a delay of 1 h and 3 h to Tung Chung, respectively (Fig. 4). 3.4. Dominant factors leading to ozone peaks at different vertical levels Surface ozone concentrations, as well as tropospheric ozone columns, reach high levels during 5 sonde periods in August. However, Section 3.1 shows that differences between tropospheric ozone columns are more obvious than those between surface ozone concentrations. It is necessary and interesting to understand the formation mechanism of the vertical ozone profile on the 8th of August, when both tropospheric ozone columns and surface ozone levels are highest in all of August. There are around 3 peaks along the ozone profile observed by ozone radiosonde (Fig. 6): one is around 8 kme10 km above ground level near the tropopause, the second is around 2 km above ground, and the last is around 500 m

Please cite this article in press as: Chen, X., et al., Factors dominating 3-dimensional ozone distribution during high tropospheric ozone period, Environmental Pollution (2017), https://doi.org/10.1016/j.envpol.2017.09.017

6

X. Chen et al. / Environmental Pollution xxx (2017) 1e10

Fig. 5. From the 6th of AUG-18:00 LST to the 8th of AUG-18:00 LST: (a) total gas-phase chemical reaction contribution to surface ozone and (b) average surface OX (O3 and NO2) mixing ratios. Average surface winds are shown in vectors. The three triangles are Tung Chung, Sha Tin, and Tap Mun, from left to right, respectively.

Fig. 6. Ozone mixing ratio profiles on the 8th of AUG: ozone sonde observation (black curve); the blue curve is the profile at 15:00 LST from the CMAQ BASE scenario, and the red curve is the profile at 15:00 LST from the CMAQ noICBC CASE scenario. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

AGL within the boundary layer. The following part of this section is to clarify the dominant factors that induce these three peaks. A sensitivity scenario (CASE-noICBC) is carried out to evaluate the UTLS exchange and entrainment effects on the ozone profiles. When the MOZART chemical initial conditions (ICs) and boundary conditions (BCs) are not applied to the CASE-noICBC scenario, large differences between BASE-CTLRUN and CASE-noICBC occur from approximately 5 km height to the model top (approximately 20 km height). The ozone peak around 10 km height near the tropopause cannot be reproduced correctly. It indicates that the peak around the tropopause and the bottom of the stratosphere is mainly caused by entrainment, mixing air in the troposphere with extremely high ozone in the stratosphere. In addition, during this period, Hong Kong may be affected by HAIKUI's subsidence and that downward stream can enhance the strength of entrainment and help stratospheric ozone to sink into the troposphere (Jiang et al., 2015). Although the simulated profile at 15:00 LST can well match the sounding observation in general, there are still differences, especially within the boundary layer. This is due to the rapid change and strong mixing effects of the boundary layer. Precise simulations of the boundary layer are still tough challenges for numerical models. Even though we increase our model's vertical resolution to 16 levels within the boundary layer, some extreme factors, such as the ozone mixing ratio tongue near 0.5 km in Fig. 6, may not be captured well. Additionally, the model may overestimate the turbulence strength

in the boundary layer so that the difference in ozone mixing ratios between 0.5 km and 2 km is not as large as observing facts. However, the general trend within 3 km above ground height is well reproduced by the model: vertical ozone mixing ratios increase from the surface and reach peak values near 0.5 km in height and then decrease and increase again to reach the second peak value near 2 km in height. To analyse the dominant factors for ozone profiles below 3 km, including two peaks of the mixing ratio, results from process analysis are mainly used. The hourly average PAs contributions from the 6th of AUG 18:00 LST to the 8th of AUG 18:00 LST (Fig. 7) can indicate factors that lead to a change in vertical ozone. Changes in the ozone profile from the 6th 18:00 LST (purple curves) to the 8th 18:00 LST (black curves) are equal to 48-h-accumulated effects of PAs contributions (coloured bars). We use model PAs profiles of Sha Tin to represent Kings Park's, considering the consistency with surface analysis and that Sha Tin's not far from Kings Park (Fig. S1: approximately 6 km). Since there are abundant precursors around the Tung Chung suburban and Tap Mun rural areas from the PRD and Hong Kong downtown, respectively, chemical reactions usually make positive contributions to ozone production from the surface layer to the lower troposphere (approximately 3 km in height). An interesting modelling result is that the cloud convection effect can transport air with a high ozone mixing ratio from 2 km to 3 km into the boundary layer and even the surface at the suburban Tung

Please cite this article in press as: Chen, X., et al., Factors dominating 3-dimensional ozone distribution during high tropospheric ozone period, Environmental Pollution (2017), https://doi.org/10.1016/j.envpol.2017.09.017

X. Chen et al. / Environmental Pollution xxx (2017) 1e10

7

Fig. 7. Ozone profiles and process analysis of vertical levels in Tung Chung, Sha Tin and Tap Mun. Black and purple curves on the black x-axis are O3 profiles at 18:00 LST on the 8th of AUG and the 6th of AUG, respectively. Colourful bars on the red x-axis stand for different factors' hourly average contributions to ozone increasing (positive) and decreasing (negative). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Chung site. In addition, there are oscillations observed in surface ozone concentrations from 18:00 LST to 23:00 LST on the 7th of AUG (Fig. 7), while process analysis indicates the existence of the cloud convection effect. This effect, probably caused by active typhoon outer circulation and typhoon outer cloud system influence, may help to increase the ozone concentration within the boundary layer for some suburban areas. Therefore, in this case, such boundary layer ozone will be transported to the downwind urban areas (such as Sha Tin), leading to a high ozone event in urban areas as well. In Sha Tin, the urban area, chemical reactions usually consume ozone within the boundary layer but make positive contributions over the boundary layer top (~1.5 kme3 km). This is because the turbulence effect can mix air pollutants well within the boundary layer, filling the whole boundary layer of NO that was largely emitted in the urban area. As a result, photochemical reactions usually have negative contributions to near surface ozone (from the surface to approximately 300 m). However, as the altitude goes over the top of the boundary layer, concentrations of NO decrease rapidly. In addition, the photolysis rate increases along with altitude. Ozone is produced more by photochemical reactions than is consumed from approximately 800 m to near 3 km in height. In general, as for the ozone profiles in the urban area Sha Tin, two peaks around 500 m and 2 km are caused by upwind transportation and self-production, respectively. The other peak around 9 km in height near the tropopause is due to the entrainment of stratospheric ozone. 3.5. Vertical O3-precursor sensitivity Two sensitive scenarios, with 30% cuts in anthropogenic NOX or VOC emissions, are applied for distinguishing vertical areas between NOX-sensitive and VOCs-sensitive. In this study, we use a method that developed by previous researchers (Liang et al., 2006) to quantify O3-precursor sensitivity: if O3 concentrations decrease more in the VOCs-reduced(NOX-reduced) sensitive scenario than those in the NOX-reduced(VOCs-reduced) scenario, it tends to be

VOCs-sensitive(NOX-sensitive). Furthermore, accompanied by process analysis, we can not only tell whether O3 increases in sensitive scenarios or not but can also acquire the changes in chemical reactions’ contributions to O3 production and consumption. The process analysis tool can help to make a more precise estimation of O3-precursor sensitivities: if photochemical reactions contribute more (less) to O3 production by reducing NOX emissions, then the area is VOCs-sensitive (NOXsensitive). The difference in 7 processes' contributions to O3 between the NOX-CUT sensitive scenario and the control base scenario (PAs_NOxCUT minus PAs_CTLRUN) is shown in the upper half of Fig. 8. Situations are different in suburban/rural Tung Chung and urban Sha Tin. Chemical reactions tend to make fewer contributions to O3 production with less NOX in Tung Chung. While in Sha Tin, the difference of chemical reactions’ contributions (△PA_CHEM) is positive. It indicates that the dominant O3-precursor sensitivities are varied from Tung Chung (NOX-sensitive) to Sha Tin (VOCssensitive). For Tung Chung, reductions of O3 in the NOX-CUT sensitive scenario (△NEGO3_NOxCUT ¼ negative part of: O3_NOxCUT minus O3_CTLRUN) are obviously larger than those in the VOC-CUT scenario (△NEG O3_VOCCUT ¼ negative part of: O3_VOCCUT minus O3_CTLRUN), except for the surface layer (Fig. 8). It also proves our conclusion that Tung Chung is mainly NOX-sensitive along the vertical direction. Meanwhile, at the surface layer, Tung Chung is VOCs-sensitive and located downwind of Guangdong metropolitans (according to Section 3.3), having good consistency with former research which indicates that surface VOCs-sensitive zones are usually located downwind of metropolitans. At the Sha Tin urban site, situations are more uniform. It is usually VOCs-sensitive from the surface to 1 km in height and then gradually transferring to NOX-sensitive above. Such a conclusion can be demonstrated in Fig. 8 by the consistent results of process analysis and two sensitive scenarios: Firstly, △PA_CHEM remains positive beneath 1 km and then gradually transfers to negative until 3 km in the CASE-NOXCUT scenario. Second, △NEGO3_NOxCUT is always larger

Please cite this article in press as: Chen, X., et al., Factors dominating 3-dimensional ozone distribution during high tropospheric ozone period, Environmental Pollution (2017), https://doi.org/10.1016/j.envpol.2017.09.017

8

X. Chen et al. / Environmental Pollution xxx (2017) 1e10

Fig. 8. Difference in 7 processes and O3 profiles in Tung Chung and Sha Tin between the NOX-CUT and VOC-CUT sensitive scenarios and the control run base are shown in coloured bars and black curves, respectively: (left) the 7th of AUG-13:00 LST; (right) the 7th of AUG-17:00 LST. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

than △NEGO3_VOCCUT from the surface to 1 km, and then △NEGO3_NOxCUT increases (from approximately 6 ppb to 5 ppb at 13:00 LST and from approximately 9 ppb to 18 ppb at 17:00 LST) and △NEGO3_VOCCUT decrease (from approximately 25 ppb to 0 at 13:00 LST and from approximately 17 ppb to 0 at 17:00 LST). Finally, △NEGO3_NOxCUT goes beyond △NEGO3_VOCCUT within the layers of 1 kme3 km. This is because in urban Sha Tin, NOX emissions are always high enough. Through the mixing effects of turbulence within the boundary layer, emitted NOX is well mixed from the surface to the boundary layer top. On the other hand, actinic flux usually increases with increasing altitude. So the transition zone will overlay an urban area from approximately 1 km above ground level. In such a zone, O3-precursor sensitivity may likely transfer from VOCs-sensitive to NOX-sensitive. The sensitivities of O3 to its two main precursors (NOX and VOCs) can be directly reflected by EKMA diagrams, which represent their nonlinear relationships. Such variations of NOX-sensitive and VOCssensitive regimes are based on the status of odd hydrogen radicals in ambient air (Sillman and He, 2002). As shown by the cycles of ozone mechanism in Wang et al. (2016), the cycles are terminated by different reactions in high- or low-NOX environments. In high-

NOX conditions, terminal reactions are NO2 reacting with RO2 and OH, forming HNO3. On the other hand, in low-NOX conditions, terminal reactions are HO2 reacting with HO2 itself and RO2, forming peroxides (including H2O2). Additionally, formaldehyde (HCHO) is an important intermediate VOC that can produce HO2 through photochemical reactions. Therefore, HNO3, NOz, H2O2 and HCHO are some of the key species that can indicate the relative reaction rate of two types of terminations. And to quantify such rates, indicators can be constructed by ratios of peroxide-related species to NOX-related species. In this case, larger values of indicators stand for a higher probability of NOX-sensitive conditions. We choose 6 indicators for further analysis. Previous studies have tested the transition values for some of these indicators (Sillman, 1995; Sillman et al., 1997; Castell et al., 2009; Lam et al., 2005). The results vary in different regions. In this research, we chose localized values for the PRD (Table S4) (Fan, 2014). These transition values are used in research of horizontal O3-precursor sensitivities (Lam et al., 2005; Wang et al., 2010; Li et al., 2013). Few researchers try to apply them in analysing vertical O3-precursor sensitivities. The more the indicators are smaller than their transition values, the more O3 tends to increase in the NOX-CUT scenario

Please cite this article in press as: Chen, X., et al., Factors dominating 3-dimensional ozone distribution during high tropospheric ozone period, Environmental Pollution (2017), https://doi.org/10.1016/j.envpol.2017.09.017

X. Chen et al. / Environmental Pollution xxx (2017) 1e10

9

Fig. 9. Profiles of six indicators in Tung Chung and Sha Tin: (left) the 7th of AUG-13:00 LST; and (right) the 7th of AUG-17:00 LST.

and decrease in the VOC-CUT scenario. However, when indicators go beyond their transition values, the tendency reverses, but only O3 will remain the same with BASE-CTLRUN in the VOC-CUT scenario. On the other hand, indicators can tell O3-precursor sensitivity from NOX-sensitive to VOC-sensitive by the deviation from transition values. We normalize the mixing ratio of indicators by each of their own transition values. Indicators now range from 100% to positive infinity. This is for clearly showing how much they depart from their transition values. The closer to 0 the normalized indicators are, the closer to transition values the ordinary indicators are. Negative (positive) percentages mean that ordinary indicators are lower (higher) than transition values, respectively. According to the former analysis, the surface layer in Tung Chung and the 0e1 km layers in Sha Tin are VOCs-sensitive. Values of indicators are far from transition values in the negative direction (Fig. 9), showing a good consistency and capability of indicating the VOCs-sensitive situation. As time goes from early afternoon (13:00 LST) to late afternoon (17:00 LST), O3-precursor sensitivity in the surface layer of Tung Chung shows a transition from VOCs-sensitive to NOX-sensitive. Meanwhile, all 6 indicators get closer to their transition values also presenting good performance in the dimension of time, although the HCHO/NOy (red curve) is close to 40% deviation and O3/NOz (purple curve) and O3/HNO3 (cyan curve) are close to þ30% deviation. Another important performance should be whether the indicators can catch the transition of O3-precursor sensitivity along the vertical dimension or not. As can be seen in Fig. 9, no matter early afternoon or late afternoon, indicators start to swiftly increase from 1 km in height at Sha Tin. Indicators can well match the transiting trend along the vertical layers, especially in late afternoon, when such transitions are more obvious in both Figs. 8 and 9. However, it is also evident that transition values of indicators are too high for Tung Chung in early afternoon. When it ought to be NOX-sensitive except for surface layers, the values of indicators drop in the negative side to transition values. Indicators do not perform very well in such occasions. In general, O3/NOy and indicators using hydrogen peroxide (H2O2/HNO3 and H2O2/NOZ) show better consistency in O3-precursor sensitivity identifications in suburban Tung Chung and urban Sha Tin sites. The results of two emission-limited scenarios and O3-precursor sensitivity indicators show good consistency with former analyses in Section 3.5.

4. Conclusions To distinguish factors inducing different high peaks in the vertical ozone profile, WRF/CMAQ models and the process analysis tool are used to simulate a high tropospheric ozone episode around the PRD and Hong Kong areas under the circulation situation of Typhoon HAIKUI. Three sensitive scenarios are performed along with the control run (BASE-CTLRUN) synchronously: one runs the BASE test without any initial and boundary conditions from the MOZART global model (CASE-noICBC) in order to evaluate the entrainment of the stratospheric ozone; the other two run the BASE but cut 30% of anthropogenic NOx or VOC emissions (CASE-NOXCUT and CASE-VOCCUT, respectively) to clarify O3-precursor sensitivity at different vertical levels above suburban/rural and urban areas. The CASE-noICBC sensitive scenario indicates that the peak of the ozone profile near the tropopause is induced by the entrainment of stratospheric ozone. Somehow, whether the strong entrainment is dominated by Typhoon HAIKUI or not remains for future research. O3-precursor sensitivity generally varies from suburban/rural areas to urban areas and from ground level to the lower troposphere. Dominant O3-precursor sensitivities are NOx-sensitive in suburban/ rural areas and VOCs-sensitive in urban areas. O3-precursor sensitivity in suburban areas will transit from VOCs-sensitive in early afternoon to transition status in late afternoon at the surface layer. In urban areas, there is a transition zone within 1e3 km in height. In such a transition zone, O3-precursor sensitivity transits from VOCssensitive to NOx-sensitive because surface emitted NOX can hardly be transported over top of the boundary layer by turbulence, making concentrations of NOX decrease rapidly. The results from process analysis, the NOXCUT and VOCCUT sensitive scenarios and O3-precursor sensitivity indicators show good consistency to demonstrate the dominating factors of two lower tropospheric ozone peaks. The occurrence of the high ozone “tongues” around 500 m and 2 km above ground level are mainly caused by upwind transportation and self-production of ozone, respectively. Acknowledgements This work is supported by the National Key R&D Program of China [2016YFC0202206 and 2017YFC0210105]; National Natural

Please cite this article in press as: Chen, X., et al., Factors dominating 3-dimensional ozone distribution during high tropospheric ozone period, Environmental Pollution (2017), https://doi.org/10.1016/j.envpol.2017.09.017

10

X. Chen et al. / Environmental Pollution xxx (2017) 1e10

Science Foundation [91544102]; the China Special Fund for Meteorological Research in the Public Interest [GYHY201406031]; the Science and Technology Planning Project of Guangdong Province, China [2014B020216003]; the Science and Technology Planning Project of Guangzhou [201604020028]; the Science and Technology Planning Project of China [2014BAC21B02]; the National Key R&D Program of China [2016YFC0203600 and 2016YFC0203305] and National Nature Science Fund for Distinguished Young Scholars [41425020]. This work is also partly supported by the highperformance grid-computing platform of Sun Yat-sen University. Appendix A. Supplementary data Supplementary data related to this article can be found at https://doi.org/10.1016/j.envpol.2017.09.017. References Byun, D., Schere, K.L., 2006. Review of the governing equations, computational algorithms, and other components of the models-3 community multiscale air quality (CMAQ) modeling system. Appl. Mech. Rev. 59, 51e77. Carter, W.P., 2015. Development of a database for chemical mechanism assignments for volatile organic emissions. J. Air Waste Manag. Assoc. 65, 1171e1184. Castell, N., Stein, A.F., Mantilla, E., Salvador, R., Mill an, M., 2009. Evaluation of the use of photochemical indicators to assess ozonedNOxdVOC sensitivity in the Southwestern Iberian Peninsula. J. Atmos. Chem. 63, 73e91. Fan, S., 2014. Distinguish the Most Suitable Photochemical Indicators for Ozone Sensitivity Identification in the Pearl River Delta Region in Autumn. School of Environmental Science and Engineering. Sun Yat-sen University, Guangzhou. Fan, Q., Lan, J., Liu, Y., Wang, X., Chan, P., Fan, S., Hong, Y., Liu, Y., Zeng, Y., Liang, G., Feng, Y., 2015a. Diagnostic analysis of the sulfate aerosol pollution in spring over Pearl River Delta, China. Aerosol Air Qual. Res. 15, 46e57. Fan, Q., Lan, J., Liu, Y., Wang, X., Chan, P., Hong, Y., Feng, Y., Liu, Y., Zeng, Y., Liang, G., 2015b. Process analysis of regional aerosol pollution during spring in the Pearl River Delta region, China. Atmos. Environ. 122, 829e838. Fiore, A.M., 2002. Background ozone over the United States in summer: origin, trend, and contribution to pollution episodes. J. Geophys. Res. 107. Fu, Y., Xian, T., Lu, D., Liu, G., Heng, Z., Sun, L., Liu, Q., Wang, Y., Yang, Y., 2013. Ozone vertical variations during a typhoon derived from the OMI observations and reanalysis data. Chin. Sci. Bull. 58, 3890e3894. Godowitch, J.M., Gilliam, R.C., Roselle, S.J., 2015. Investigating the impact on modeled ozone concentrations using meteorological fields from WRF with an updated fouredimensional data assimilation approach. Atmos. Pollut. Res. 6, 305e311. Guenther, A.B., Jiang, X., Heald, C.L., Sakulyanontvittaya, T., Duhl, T., Emmons, L.K., Wang, X., 2012. The Model of Emissions of Gases and Aerosols from Nature version 2.1 (MEGAN2.1): an extended and updated framework for modeling biogenic emissions. Geosci. Model Dev. 5, 1471e1492. Guo, H., Ling, Z.H., Cheung, K., Jiang, F., Wang, D.W., Simpson, I.J., Barletta, B., Meinardi, S., Wang, T.J., Wang, X.M., Saunders, S.M., Blake, D.R., 2013. Characterization of photochemical pollution at different elevations in mountainous areas in Hong Kong. Atmos. Chem. Phys. 13, 3881e3898. Henderson, B.H., Akhtar, F., Pye, H.O.T., Napelenok, S.L., Hutzell, W.T., 2014. A database and tool for boundary conditions for regional air quality modeling: description and evaluation. Geosci. Model Dev. 7, 339e360. Hocking, W.K., Carey-Smith, T., Tarasick, D.W., Argall, P.S., Strong, K., Rochon, Y., Zawadzki, I., Taylor, P.A., 2007. Detection of stratospheric ozone intrusions by windprofiler radars. Nature 450, 281e284. Holton, 1995. Stratosphere-troposphere exchange. Rev. Geophys. 33 (4), 403e439. Jacob, D.J., 1999. Introduction to Atmospheric Chemistry. Princeton University Press, Princeton, New Jersey. Jiang, Y.C., Zhao, T.L., Liu, J., Xu, X.D., Tan, C.H., Cheng, X.H., Bi, X.Y., Gan, J.B., You, J.F., Zhao, S.Z., 2015. Why does surface ozone peak before a typhoon landing in southeast China? Atmos. Chem. Phys. 15, 13331e13338.  Jimenez, P., Baldasano, J.M., 2004. Ozone response to precursor controls in very complex terrains: use of photochemical indicators to assess O3-NOx-VOC sensitivity in the northeastern Iberian Peninsula. J. Geophys. Res. Atmos. 109 n/ a-n/a. Lam, K., Wang, T., Wu, C., Li, Y., 2005. Study on an ozone episode in hot season in Hong Kong and transboundary air pollution over Pearl River Delta region of China. Atmos. Environ. 39, 1967e1977. Li, X., Rappenglück, B., 2014. A WRFeCMAQ study on spring time vertical ozone structure in Southeast Texas. Atmos. Environ. 97, 363e385. Li, Y., Lau, A.K.H., Fung, J.C.H., Zheng, J., Liu, S., 2013. Importance of NOx control for peak ozone reduction in the Pearl River Delta region. J. Geophys. Res. Atmos. 118, 9428e9443. Li, J., Lu, K., Lv, W., Li, J., Zhong, L., Ou, Y., Chen, D., Huang, X., Zhang, Y., 2014a. Fast increasing of surface ozone concentrations in Pearl River Delta characterized by a regional air quality monitoring network during 2006-2011. J. Environ. Sci. 26,

23e26. Li, M., Zhang, Q., Streets, D.G., He, K.B., Cheng, Y.F., Emmons, L.K., Huo, H., Kang, S.C., Lu, Z., Shao, M., Su, H., Yu, X., Zhang, Y., 2014b. Mapping Asian anthropogenic emissions of non-methane volatile organic compounds to multiple chemical mechanisms. Atmos. Chem. Phys. 14, 5617e5638. Liang, J., Jackson, B., Kaduwela, A., 2006. Evaluation of the ability of indicator species ratios to determine the sensitivity of ozone to reductions in emissions of volatile organic compounds and oxides of nitrogen in northern California. Atmos. Environ. 40, 5156e5166. Liu, Y., Zhang, S., Fan, Q., Wu, D., Chan, P., Wang, X., Fan, S., Feng, Y., Hong, Y., 2015. Accessing the impact of sea-salt emissions on aerosol chemical formation and deposition over Pearl River Delta, China. Aerosol Air Qual. Res. 15, 2232e2245. Ma, Z., Xu, J., Quan, W., Zhang, Z., Lin, W., Xu, X., 2016. Significant increase of surface ozone at a rural site, north of eastern China. Atmos. Chem. Phys. 16, 3969e3977. Park, S.-Y., Lee, S.-H., Lee, H.W., 2014. Assimilation of wind profiler observations and its impact on three-dimensional transport of ozone over the Southeast Korean Peninsula. Atmos. Environ. 99, 660e672. Pye, H.O., Luecken, D.J., Xu, L., Boyd, C.M., Ng, N.L., Baker, K.R., Ayres, B.R., Bash, J.O., Baumann, K., Carter, W.P., Edgerton, E., Fry, J.L., Hutzell, W.T., Schwede, D.B., Shepson, P.B., 2015. Modeling the current and future roles of particulate organic nitrates in the Southeastern United States. Environ. Sci. Technol. 49, 14195e14203. Sillman, S., 1995. The use of NO y, H2O2, and HNO3 as indicators for ozone-NO x -hydrocarbon sensitivity in urban locations. J. Geophys. Res. Atmos. 100, 14175e14188. Sillman, S., He, D., 2002. Some theoretical results concerning O3-NOx-VOC chemistry and NOx-VOC indicators. J. Geophys. Res. 107. Sillman, S., He, D., Cardelino, C., Imhoff, R.E., 1997. The use of photochemical indicators to evaluate ozone-NOx-hydrocarbon sensitivity: case studies from Atlanta, New York, and Los Angeles. J. Air Waste Manag. Assoc. 47, 1030e1040. Sun, L., Xue, L., Wang, T., Gao, J., Ding, A., Cooper, O.R., Lin, M., Xu, P., Wang, Z., Wang, X., Wen, L., Zhu, Y., Chen, T., Yang, L., Wang, Y., Chen, J., Wang, W., 2016. Significant increase of summertime ozone at mount Tai in central eastern China. Atmos. Chem. Phys. 16, 10637e10650. Tang, Y., Lee, P., Tsidulko, M., Huang, H.-C., McQueen, J.T., DiMego, G.J., Emmons, L.K., Pierce, R.B., Thompson, A.M., Lin, H.-M., Kang, D., Tong, D., Yu, S., Mathur, R., Pleim, J.E., Otte, T.L., Pouliot, G., Young, J.O., Schere, K.L., Davidson, P.M., Stajner, I., 2008. The impact of chemical lateral boundary conditions on CMAQ predictions of tropospheric ozone over the continental United States. Environ. Fluid Mech. 9, 43e58. Tonnesen, G.S., Dennis, R.L., 2000. Analysis of radical propagation efficiency to assess ozone sensitivity to hydrocarbons and NOx: 2. Long-lived species as indicators of ozone concentration sensitivity. J. Geophys. Res. Atmos. 105, 9227e9241. Wang, T., Wei, L.X., Ding, A.J., Poon, C.N., Lam, K.S., Li, Y.S., Chan, L.Y., Anson, M., 2009. Increasing surface ozone concentrations in the background atmosphere of Southern China, 1994-2007. Atmos. Chem. Phys. 9, 6217e6227. Wang, X., Zhang, Y., Hu, Y., Zhou, W., Lu, K., Zhong, L., Zeng, L., Shao, M., Hu, M., Russell, A.G., 2010. Process analysis and sensitivity study of regional ozone formation over the Pearl River Delta, China, during the PRIDE-PRD2004 campaign using the Community Multiscale Air Quality modeling system. Atmos. Chem. Phys. 10, 4423e4437. Wang, X., Zhang, Y., Hu, Y., Zhou, W., Zeng, L., Hu, M., Cohan, D.S., Russell, A.G., 2011. Decoupled direct sensitivity analysis of regional ozone pollution over the Pearl River Delta during the PRIDE-PRD2004 campaign. Atmos. Environ. 45, 4941e4949. Wang, Y., Shen, L., Wu, S., Mickley, L., He, J., Hao, J., 2013. Sensitivity of surface ozone over China to 2000e2050 global changes of climate and emissions. Atmos. Environ. 75, 374e382. Wang, N., Guo, H., Jiang, F., Ling, Z.H., Wang, T., 2015. Simulation of ozone formation at different elevations in mountainous area of Hong Kong using WRF-CMAQ model. Sci. total Environ. 505, 939e951. Wang, T., Xue, L., Brimblecombe, P., Lam, Y.F., Li, L., Zhang, L., 2016. Ozone pollution in China: a review of concentrations, meteorological influences, chemical precursors, and effects. Sci. total Environ. 575, 1582e1596. Willmott, Cort J., 1981. On the validation of models. Phys. Geogr. 2 (2), 184e194. Xu, X., Lin, W., Wang, T., Yan, P., Tang, J., Meng, Z., Wang, Y., 2008. Long-term trend of surface ozone at a regional background station in eastern China 1991-2006: enhanced variability. Atmos. Chem. Phys. 8, 2595e2607. Xue, L., Wang, T., Louie, P.K., Luk, C.W., Blake, D.R., Xu, Z., 2014. Increasing external effects negate local efforts to control ozone air pollution: a case study of Hong Kong and implications for other Chinese cities. Environ. Sci. Technol. 48, 10769e10775. Yu, X., Yuan, Z., Fung, J.C.H., Xue, J., Li, Y., Zheng, J., Lau, A.K.H., 2014. Ozone changes in response to the heavy-duty diesel truck control in the Pearl River Delta. Atmos. Environ. 88, 269e274. Zhang, Q., Streets, D.G., Carmichael, G.R., He, K.B., Huo, H., Kannari, A., Klimont, Z., Park, I.S., Reddy, S., Fu, J.S., Chen, D., Duan, L., Lei, Y., Wang, L.T., Yao, Z.L., 2009. Asian emissions in 2006 for the NASA INTEX-B mission. Atmos. Chem. Phys. 9, 5131e5153. Zhang, Y., Liu, H., Crawford, J.H., Considine, D.B., Chan, C., Oltmans, S.J., Thouret, V., 2012. Distribution, variability and sources of tropospheric ozone over south China in spring: intensive ozonesonde measurements at five locations and modeling analysis. J. Geophys. Res. Atmos. 117, D12304.

Please cite this article in press as: Chen, X., et al., Factors dominating 3-dimensional ozone distribution during high tropospheric ozone period, Environmental Pollution (2017), https://doi.org/10.1016/j.envpol.2017.09.017