An important mechanism of regional O 3 transport for summer smog over the Yangtze River Delta in eastern China

Severe ozone (O3) pollution episodes plague a few regions in eastern China at certain times of the year, e.g., the Yangtze River Delta (YRD). However, the formation mechanisms, including meteorological factors, contributing to these severe pollution events remain elusive. A severe summer smog stretched over the YRD region from 22 to 25 August 2016. This event displayed hourly surface O3 concentrations that exceeded 300 μg m−3 on 25 August in Nanjing, an urban area in the western YRD. The weather pattern during this period was characterized by near-surface prevailing easterly winds and continuous high air temperatures. The formation mechanism responsible for this O3 pollution episode over the YRD region, particularly the extreme values over the western YRD, was investigated using observation data and by running simulations with the Weather Research and Forecasting model with Chemistry (WRF-Chem). The results showed that the extremely high surface O3 concentration in the western YRD area on 25 August was largely due to regional O3 transport in the nocturnal residual layer (RL) and the diurnal change in the atmospheric boundary layer. On 24 August, high O3 levels, with peak values of 220 μg m−3, occurred in the daytime mixing layer over the eastern YRD region. During nighttime from 24 to 25 August, a shallow stable boundary layer formed near the surface which decoupled the RL above it from the surface. Ozone in the decoupled RL remained quite constant, which resulted in an O3-rich “reservoir” forming in this layer. This reservoir persisted due to the absence of O3 consumption from nitrogen oxide (NO) titration or dry deposition during nighttime. The prevailing easterly winds in the lower troposphere governed the regional transport of this O3-rich air mass in the nocturnal RL from the eastern to the western YRD. As the regional O3 transport reached the RL over the western YRD, O3 concentrations in the RL accumulated and rose to 200 μg m−3 over the western Nanjing site during the sunrise hours on 25 August. The development of the daytime convective boundary layer after sunrise resulted in the disappearance of the RL, as the vertical mixing in the convective boundary layer uniformly redistributed O3 from the upper levels via the entrainment of O3-rich RL air down to the O3-poor air at the ground. This net downward transport flux reached up to 35 μg m−3 h−1, and contributed a considerable surface O3 accumulation, resulting in severe daytime O3 pollution during the summer smog event on 25 August Published by Copernicus Publications on behalf of the European Geosciences Union. 16240 J. Hu et al.: An important mechanism of regional O3 transport in the western YRD region. The mechanism of regional O3 transport through the nocturnal RL revealed in this study has great implications regarding understanding O3 pollution and air quality change.

eastern YRD region.During nighttime from 24 to 25 August, a shallow stable boundary layer formed near the surface which decoupled the RL above it from the surface.Ozone in the decoupled RL remained quite constant, which resulted in an O 3 -rich "reservoir" forming in this layer.This reservoir persisted due to the absence of O 3 consumption from nitrogen oxide (NO) titration or dry deposition during nighttime.The prevailing easterly winds in the lower troposphere governed the regional transport of this O 3 -rich air mass in the nocturnal RL from the eastern to the western YRD.As the regional O 3 transport reached the RL over the western YRD, O 3 concentrations in the RL accumulated and rose to 200 µg m −3 over the western Nanjing site during the sunrise hours on 25 August.The development of the daytime convective boundary layer after sunrise resulted in the disappearance of the RL, as the vertical mixing in the convective boundary layer uniformly redistributed O 3 from the upper levels via the entrainment of O 3 -rich RL air down to the O 3 -poor air at the ground.This net downward transport flux reached up to 35 µg m −3 h −1 , and contributed a considerable surface O 3 accumulation, resulting in severe daytime O 3 pollution during the summer smog event on 25 August

Introduction
Tropospheric ozone (O 3 ) is an important atmospheric component and influences climate change and air quality in different ways.According to the Intergovernmental Panel on Climate Change (Bréon et al., 2013), tropospheric O 3 is one of the most important greenhouse gases affecting global warming.It is also a health hazard for sensitive individuals, reducing lung function and contributing to the exacerbation of asthma symptoms (Bell et al., 2006).Tropospheric O 3 can also alter atmospheric chemistry because its photolysis in the presence of water vapor is the primary source of the hydroxyl radical (OH), which is responsible for the removal of many important trace gases.(Thompson, 1992;Logan et al., 1981).
The spatiotemporal variations of tropospheric O 3 are substantial at global and regional scales.In addition to photochemical reactions associated with O 3 precursor emissions and solar radiation, the atmospheric transport of O 3 and its precursors, including horizontal transport (Wolff et al., 1977;Yienger et al., 2000;Wild and Akimoto, 2001;Lelieveld et al., 2002;Duncan et al., 2008;Liu et al., 2011;Zhu et al., 2017;Han et al., 2018) and vertical transport, e.g., exchange between stratosphere and troposphere (Hu et al., 2010;Jiang et al., 2015), plays an important role in determining the spatiotemporal distribution of this species in the troposphere.
Ambient O 3 levels are strongly influenced by the diurnal variation of the atmospheric boundary layer (BL) structure.The daytime BL, also known as the convective boundary layer (CBL), is directly affected by the solar heating of the Earth's surface.In the major part of the CBL, known as the mixing layer (ML), air pollutant concentrations are almost uniformly distributed due to convective turbulent mixing.The nocturnal BL is often characterized by a stable layer (SL) near the surface and an overlying residual layer (RL).The SL develops due to radiative cooling after sunset.Above the SL, the remnants of the daytime ML form the RL, which initially constitutes uniformly mixed air pollutants remaining from the preceding day (Stull, 1988); O 3 is one of these representative remnant air pollutants in the RL due to the lack of O 3 consumption from NO titration and dry deposition during the night (Sillman, 1999;Xie et al., 2016).
Therefore, nocturnal O 3 in the RL can exert an impact on the ambient O 3 variation the following day (Aneja et al., 2000;Hu et al., 2012;Morris et al., 2010;Neu et al., 1994;Tong et al., 2011;Yorks et al., 2009;Hu et al., 2013;Klein et al., 2014).Locally, O 3 from the RL can potentially contribute to the maximum surface O 3 concentrations the following day and enhance surface O 3 by as much as 10-30 ppb (Hu et al., 2012).A few studies (Zhang et al., 1998;Zhang and Rao, 1999) have investigated O 3 episodes in the BL over the northeastern US based on measurements and 1-D model simulations.These studies suggested that O 3 in the nocturnal RL can be transported to downwind areas by low-level jets over the eastern coast of the US. Lee et al. (2003) found that the daytime upslope flows transported O 3 precursors up into the mountains, while the nocturnal downslope flow transported the O 3 -rich RL air mass downwards into the Phoenix Valley; therefore, it was concluded that the transport, distribution and storage of O 3 are highly impacted by background meteorological conditions.Zhang et al. (2015) found that regional transport within the RL from surrounding urban areas led to a nighttime O 3 peak on a mountaintop in eastern China.However, the regional transport of O 3 in the RL with respect to air pollution is poorly understood, especially for plain regions.
In recent years, ambient O 3 levels have increased over the Yangtze River Delta (YRD) in eastern China, with more frequent photochemical pollution events or summer smog from late May to July (Tang et al., 2013).From 1990 to 2013, the hourly O 3 peaks varied from 140 to 167 ppbv (about 294-350 µg m −3 ) in the YRD region, from 160 to 180 ppbv (about 336-378 µg m −3 ) in the Beijing-Tianjin-Hebei area over the North China Plain and from 200 to 220 ppbv (about 420-462 µg m −3 ) in the Pearl River Delta in southern China (Wang et al., 2017).Coupled with the increases in nitrogen oxide (NO x ) and volatile organic compound (VOC) emissions, O 3 distribution in the lower troposphere is significantly influenced by winds, air temperature, cloud cover and downward shortwave radiation, which all affect the regional transport and chemical formation of O 3 (An et al., 2015;Gao et al., 2016;Xu et al., 2008;Li et al., 2018).O 3 levels have been found to increase at a rate of 4-5 ppb K −1 when temperatures are between 28 and 38 • C (Pu et al., 2017).The prevailing winds driving the transport of air pollutants from the YRD industrialized areas may also contribute to O 3 enhancement (Tang et al., 2013).Heat waves with a maximum temperature of ≥ 32 • C for 3 consecutive days in the YRD, a region with a sunny and strong solar radiation environment, can significantly strengthen photochemical reactions and potentially lead to substantially elevated O 3 in a warmer climate (Tie et al., 2009;Li et al., 2012;Wang et al., 2017;Xie et al., 2016;Pu et al., 2017).In addition, the ambient O 3 level can be affected by the diurnal variation of the atmospheric BL structure over the YRD, with the nighttime stable BL height dropping to 200 m and the daytime BL height reaching up to about 1200 m (Chang et al., 2016).Therefore, it is important to understand the formation mechanisms of O 3 pollution including the meteorological factors influencing O 3 pollution for summer smog over the YRD region.
This study focused on the formation of O 3 pollution during a summer smog episode observed over the YRD in August 2016.We aimed to explore the underlying mechanism of regional O 3 transport over the YRD using observational data and WRF-Chem modeling simulations.The rest of this paper is organized as follows: Section 2 describes the observational data and the O 3 pollution episode.Section 3 presents the WRF-Chem modeling methodology and validation.In Sect.4, a O 3 pollution formation mechanism is revealed involving regional O 3 transport in the RL from the eastern to the western YRD.The conclusions are then summarized in Sect. 5.
2 Observed O 3 pollution episode

Observation sites and data
Observation data from the YRD urban sites of Nanjing (NJ), Zhenjiang (ZJ), Changzhou (CZ), Wuxi (WX), Suzhou (SZ) and Shanghai (SH) (Fig. 1) were used to study O 3 pollution during the August 2016 summer smog episode.The meteorological data were collected from the China Meteorological Administration (http://www.cma.gov.cn, last access: 11 November 2018), whereas the air quality monitoring data were sourced from Ministry of Ecology and Environment of China (http://www.mee.gov.cn, last access: 11 November 2018).The meteorological data included wind speed (m s −1 ) and direction ( • ) at 10 m above ground level, air temperature ( • C) and relative humidity (%) at 2 m above ground level with a temporal resolution of 3 h, and total radiation irradiance with a time resolution of 1 h.The air quality monitoring data included O 3 and NO 2 concentrations with a temporal resolution of 1 h.The air pollutant concentrations were averaged from measurements from multiple sites for each city, whereas the meteorological variables were only from one site for each city.

Summer smog in a heat wave episode over the YRD
With the western Pacific subtropical high being located over the YRD area, heat wave events with high surface air temperature always occur over this region in summer.The average daily temperature and daily maximum temperature in the YRD were 27.1 and 39.5 • C, respectively, in summer during the 2013-2016 period, whilst the average summer daily temperature was 28.5 • C in NJ.It is generally accepted that high O 3 concentrations are accompanied by high air temperatures and strong photochemical reactions (Filleul et al., 2006;Pu et al., 2017;Seinfeld and Pandis, 1986).During the heat wave episode from 22 to 25 August 2016, a summer smog accompanied by severe O 3 pollution occurred over the YRD region (Table 1); high surface O 3 concentrations with average maximum 8 h running mean values from 141.1 to 204.3 µg m −3 were measured at all six urban sites (NJ, ZJ, CZ, WX, SZ and SH; Table 1).During this summer smog episode, on mostly sunny days controlled by the westward stretching subtropical anticyclone of the Western Pacific, the daily maximum air temperature ranged from 32.8 to 34.0 • C over the YRD region, and the mean air temperature during the periods of the maximum 8 h running mean surface O 3 concentrations ex-

A potential role of regional O 3 transport
Surface air temperature and solar radiation deeply affect photochemical production.Therefore, high O 3 concentrations are generally associated with high air temperatures (Rao et al., 1992;Council, 1991).The hourly maximum O 3 concentrations at the NJ site were 256.8 and 317.2 µg m −3 at 16:00 on 24 August and 15:00 on 25 August (all times mentioned in this paper are local times), respectively, with a lag of a few hours being observed between the time of maximum total radiation irradiances and maximum air temperature and the maximum O 3 concentrations (Fig. 2).However, we found from the observation of this particular summer smog episode that the air temperature and the O 3 levels exhibited opposite overall changes in the western YRD area from 24 to 25 August (Fig. 2 and    , 1999;Carnero et al., 2010;Corsmeier et al., 1997;Gangoiti et al., 2002;Godowitch et al., 2011;Tang et al., 2017;Shu et al., 2016).Based on the available observations of gaseous species, it is estimated that the daytime mean surface nitrogen dioxide (NO 2 ) concentrations varied slightly during the 24-25 August period (Table 2), reflecting the lower impact from local photochemical production on the daily enhancement of O 3 .The near-surface 90 • and 111 • easterly winds prevailed with daily averaged wind speeds of 2.4 and 2.6 m s −1 on 24 and 25 August, respectively, at NJ (Table 2); this indicated fewer changes in both wind speed and direction over NJ during these 2 days.Excluding the impact of photochemical production and the changes in horizontal winds, the potential role of regional O 3 transport in association with vertical exchange over the YRD becomes an important mechanism in ambient O 3 pollution for summer smog in the western YRD on 25 August 2016.This is explored using a modeling study in the following sections.
3 Simulation settings and validation

Simulation settings
To investigate regional O 3 transport over the YRD and the underlying mechanism involved, the Weather Research and Forecasting model with Chemistry (WRF-Chem) version 3.8.1 (Grell et al., 2005) was employed in this study.Three nested domains with respective horizontal resolutions of 45, 15 and 5 km cover East Asia, eastern China and the YRD region (Fig. 1) with 32 vertical layers extending from the surface to 100 hPa.The simulation period spans from 21 to 25 August 2016 with 1-hourly model outputs, and the spinup time is the first 24 h.The physical parameterizations include the Noah land surface model (Tewari et al., 2004), the Mesoscale Model (MM5) similarity surface layer, the Yonsei University (YSU) boundary layer scheme (Hong et al., 2006), the Rapid Radiative Transfer Model (RRTM) longwave scheme (Mlawer et al., 1997), the Goddard shortwave scheme (Chou et al., 1998), the Morrison double-moment microphysics scheme (Morrison et al., 2009) and the Kain-Fritsch cumulus parameterization (Kain, 2004).The gasphase chemical mechanism is selected using the Regional Acid Deposition Model, version 2 (RADM2) (Chang et al., 1990;Stockwell et al., 1984)

Modeling validation
Simulated wind speed, air temperature, relative humidity and O 3 concentrations are compared with the observations from six sites in the YRD (Fig. 1b) during the period from 22 to 25 August 2016 for the previously mentioned O 3 pollution episode (Fig. 3).The correlation coefficients for the nearsurface air temperature and relative humidity reach up to about 0.9, with only a slight overestimation of relative humidity.Over the NJ, CZ, WX, SZ and SH sites, the correlation coefficients between the observed and simulated wind speeds exceed 0.6.The high correlation coefficients between the observed and simulated O 3 concentrations range between 0.8 and 0.9 with small standard deviations, and their normalized root-mean-square (NRMS) errors are lower than 0.6.All of the O 3 and meteorological correlations between simulations and observations are significant at a 0.001 level (except for the wind speed over ZJ, which is significant at a 0.05 level).The validation of the vertical structures of O 3 is very important in the analysis of the O 3 budget; however, it was not possible for us to evaluate the vertical structure of O 3 from the simulation.If there were observational data available for the O 3 vertical profiles, the validation of the O 3 vertical profiles could be carried out in a future study of the O 3 budget.In general, the WRF-Chem simulated O 3 , air temperature, relative humidity and wind speed in the YRD show good agreement with the observations.The simulation also captures the observed changes of O 3 and meteorology during the summer smog episode over the YRD reasonably well.Therefore, simulation data could be used to investigate the regional O 3 transport and the underlying mechanism over the YRD during the summer smog period, as presented in the following sections.to present the temporal changes in the vertical structures of the O 3 concentrations and atmospheric boundary layer from 24 to 25 August 2016 based on the WRF-Chem simulation (Fig. 4). Figure 4a presents the hourly changes in the vertical O 3 profiles from afternoon to midnight on 24 August over the eastern YRD region.On the afternoon of 24 August, especially around 16:00, surface O 3 reached peak concentrations of about 220 µg m −3 in association with the maximum air temperature during the heat wave.The weak vertical gradients of potential temperature represented the well-developed mixing layer up to a height of about 1.5 km above the surface, which resulted in strong O 3 vertical mixing over the eastern YRD area (Fig. 4a).After sunset on 24 August, the nearsurface O 3 concentrations decreased sharply in the absence of photochemical production, the near-surface O 3 consumption from nitrogen oxide (NO) titration and dry deposition.This resulted in the formation of a typical O 3 -poor stable boundary layer and an overlying O 3 -rich "reservoir" in the nocturnal RL over the eastern YRD region (Fig. 4a).
Figure 4b displays the hourly changes in the vertical profiles of O 3 and potential temperature over the western YRD region on the morning of 25 August 2016.Reflected by the strong vertical gradients of potential temperature, the existence of the stable boundary layer up a height of 0.1 km over the surface at nighttime prevented the O 3 -rich air mass in the RL from being vertically transported to the surface.This resulted in the accumulation of an O 3 -rich "reservoir" in the RL at altitudes from 0.1 to 1 km over the western YRD area (Fig. 4b).After sunrise on 25 August, the stable boundary layer and the RL vanished due to the development of the CBL, which triggered the vertical mixing of the O 3 -rich air mass in the RL and the near-surface O 3 -poor air mass; this redistributed the O 3 concentrations in the daytime CBL and enhanced the surface O 3 level during the morning (Fig. 4b).
We compare the temporal changes of the O 3 "reservoir" in the nocturnal RL over the eastern and western YRD areas in Fig. 4a and b.It is interesting to note that the O 3 concentrations in the eastern O 3 "reservoir" obviously decreased ("leaked") overnight on 24 August (Fig. 4a), while the western O 3 "reservoir" was gradually strengthened and formed a high O 3 center that exceeded 200 µg m −3 over the western NJ site at around 06:00 (at sunrise) on 25 August (Fig. 4b).Considering the prevailing easterly winds in the lower troposphere over the YRD region during the summer smog period, we speculate that the regional O 3 transport in the nocturnal RL could connect the eastern decreases and western increases in the overnight O 3 "reservoir" within the RL (Fig. 4a and b).Therefore, we further investigate the regional O 3 transport in the nocturnal RL over this area to interpret the observational evidence regarding the exacerbated O 3 pollution and weakened photochemical production on 25 August in the western YRD (Figs. 1b-2, Table 2).

O 3 transport in the RL
It is worth discussing why the nocturnal O 3 concentrations in the RL increased by about 40 µg m −3 over the western YRD region from 03:00 to 06:00 on 25 August (Fig. 4b).To investigate the regional O 3 transport over the YRD that contributed to the O 3 enhancement in the nocturnal RL over the western YRD area, Figure 5 presents the variations in the O 3 concentrations and wind streamlines at an altitude of about 900 m in the RL on the morning of 25 August.It is clearly seen from Figure 5 that the prevailing easterly winds drove the O 3 transport from the eastern to the western YRD region during the night from 24 to 25 August.This confirms our speculation that O 3 transport in the nocturnal RL over the YRD connects the overnight changes in the O 3 levels between the eastern and the western YRD sites (Fig. 4a and b).It is noteworthy that the regional O 3 transport in the nocturnal RL reached westward over the western urban NJ site before sunrise (around 06:00) on 25 August; this was also associated with the stagnation of cyclone circulation over NJ from 03:00 to 10:00, which prevented high O 3 concentrations from moving further west and converged O 3 into the RL over the western YRD region (the NJ and ZJ sites) after sunrise (around 09:00) on 25 August 2016 (Fig. 5).
Figure 6 presents the temporal evolution of vertical sections of the O 3 concentrations and atmospheric circulation along the regional O 3 transport route over the YRD region in order to further explore the mechanism of regional O 3 transport during summer smog in this area.The vertical O 3 concentration distributions were strongly controlled by the diurnal change of the BL structure.The daytime O 3 concentrations were vertically and uniformly distributed in the mixing layer (ML), the major part of the CBL over the YRD (Fig. 6a), which forms the O 3 -rich RL after sunset.Under the guidance of the prevailing easterly winds, the O 3 transport from the eastern to the western YRD region persisted during the night from 24 to 25 August (Fig. 6b-e); this further confirmed our speculation regarding regional O 3 transport in the nocturnal RL over the YRD region.The regional O 3 transport in the RL from the eastern YRD to the western YRD NJ site increased the O 3 concentrations to 200 µg m −3 within the RL over the western NJ site during the sunrise hours on 25 August (Fig. 4b).The O 3 horizontal transport flux in the RL averaged over the nighttime from 20:00 on 24 August to 08:00 on 25 August was 541 µg m −2 s −1 at the western NJ site which was 119 µg m −2 s −1 stronger than the flux during the preceding night.This reflects the large contribution of O 3 horizontal transport in RL to the O 3 pollution on 25 August over the western YRD.The RL containing the O 3 -rich air mass over the western area, which was contributed by nocturnal O 3 transport over the YRD, was destroyed during the development of the daytime CBL due to strong vertical mixing after sunrise on 25 August (Fig. 6f).The large daily contribution of vertical mixing to the surface O 3 level occurred around 10:00 in the morning and comprised downwards vertical mixing from the upper levels which caused the entrainment of O 3 -rich RL air to the O 3 -poor air mass at the ground (Figs.6f and 7).  the O 3 -rich air mass can be entrained downwards to the surface (Mcelroy and Smith, 1993;Venkatram, 1977) and can contribute to the surface O 3 concentrations in the morning (Fig. 6f).
Based on the WRF-Chem simulation, Fig. 7 presents the hourly changes in the contribution rates of vertical mixing and local chemical reactions to surface O 3 in the western urban NJ site from 24 to 25 August.Vertical mixing is initiated by convective and turbulent processes during the develop-ment of the daytime convective boundary layer, and chemical reactions constitute the net output of all O 3 chemical reactions (Gao et al., 2016).The positive and negative contribution rates indicate the respective gain and loss of surface O 3 via vertical mixing and local chemical reactions.The daily totals of the positive contribution of vertical mixing and local chemical reactions on 24 and 25 August are given in Table 3. Relative to 24 August, the positive contribution of O 3 vertical mixing was significantly enhanced on 25 August with the largest contribution of O 3 , about 35 µg m −3 h −1 , being twice as high as on 24 August (Fig. 7).Tropospheric O 3 mainly results from photochemical reactions during the day (Seinfeld and Pandis, 1986); although the largest O 3 contributions from chemical reactions reached up to 38 and 44 µg m −3 h −1 in the western YRD area on the afternoon of 24 and 25 August, respectively (Fig. 7), the daily totals of the positive contributions of chemical reactions were estimated to be lower on 25 August (238 µg m −3 ) than during the day on 24 August (240 µg m −3 ).The daily totals of the positive O 3 contribution from vertical mixing rose sharply to 115 µg m −3 on 25 August, which was a large increase of 52 µg m −3 compared with 24 August (Table 3).The high O 3 levels in ambient air for summer smog in the western YRD on 25 August comprised a significant contribution from the downward vertical mixing of the O 3 -rich RL air mass, which was transported in the nocturnal RL from the eastern to the western YRD.Therefore, regional O 3 transport in the nocturnal RL in combination with the diurnal changes of the boundary layer are revealed to be an important mechanism of regional O 3 transport in eastern China.
Based on the simulated dry deposition, we calculated hourly changes in O 3 dry deposition and estimated daily average dry deposition rates of about 0.42 and 0.49 µg m −2 s −1 for 24 and 25 August, respectively.The dry depositions of O 3 varied little over these 2 days, with only a slight enhancement on 25 August.This reflects the fact that O 3 dry deposition exerted little impact on surface O 3 changes during the 24-25 August period.Furthermore, the contribution of O 3 dry deposition to tropospheric O 3 changes was trivial compared to vertical mixing and chemical reactions (Wang et al., 1998;Fowler et al., 1999;Zaveri et al., 2003).
Considering weak changes in local emissions over short periods of time, the WRF-Chem simulation, run with hourly emissions of chemical species, over the YRD remained unchanged from 24 to 25 August.To analyze the impact of photochemical production, we used the observed surface NO 2 concentrations and the total radiation irradiance (TRI) to examine the change in photochemical production rates.There were no apparent changes in NO 2 or TRI from 24 to 25 August, which indicated that photochemical production exerted little impact on the high O 3 level on 25 August compared with regional O 3 transport in the nocturnal RL.The analysis of simulation results revealed that vertical mixing of the upper O 3 -rich RL with the daytime surface layer was a large contributor to O 3 enhancement for summer smog in the western YRD on 25 August 2016 (Fig. 7).

Conclusions
By analyzing observational data of gaseous species and meteorological variables during severe summer smog over the YRD in eastern China in August 2016, we found noteworthy observational evidence of increasing daytime surface O 3 levels.However, even though daytime temperatures were lower, and photochemical production and solar radiation were weaker on 25 August, daytime O 3 still increased from 24 to 25 August in the western YRD.Therefore, regional O 3 transport over the YRD is believed to play a pivotal role in the ambient O 3 pollution on 25 August in the western YRD.
By combining environmental and meteorological observation data with air quality modeling, the formation mechanism of the O 3 pollution episode over the YRD area, particularly the severe pollution over western YRD, was investigated.On 24 August, (high) O 3 levels peaked at about 220 µg m −3 in the daytime mixing layer over the eastern YRD area.During nighttime, a shallow stable boundary layer formed near the surface, which decoupled the RL above it from the surface and resulted in the development of an O 3 -rich "reservoir".Governed by prevailing easterly winds in the lower troposphere, the O 3 -rich air mass in the nocturnal RL shifted from the eastern to the western YRD with a horizontal transport flux of 541 µg m −2 s −1 .Consequently, the O 3 concentrations in the RL over the western YRD area increased to 200 µg m −3 around sunrise on 25 August 2016.The disappearance of the RL after sunrise was accompanied by vertical mixing, which was initiated by convective and turbulent processes during the establishment of daytime CBL.The vertical mixing in the CBL occurred from the upper levels to the ground, with the net downward transport flux reaching up to 35 µg m −3 h −1 during daytime on 25 August.This flux contributed considerably to the surface O 3 accumulation in summer smog over the western YRD region on 25 August 2016, which is of great importance with respect to the formation of severe O 3 pollution in the western YRD region.
This study revealed regional O 3 transport through the nocturnal RL (from upstream to downstream areas driven by the prevailing winds in the lower troposphere in close association with the diurnal change in the atmospheric boundary layer) to be an important mechanism, which can be depicted using a conceptual model (Fig. 8).This regional O 3 transport mechanism has substantial implications regarding understanding urban O 3 pollution and air quality change.
Regional O 3 transport in the atmospheric boundary layer during this particular case of summer smog in the YRD, eastern China is to be further studied using more comprehensive observations of meteorology and environment, including the impact of changes in biogenic VOCs on O 3 concentrations and better modeling of the atmospheric boundary layer.

Figure 1 .
Figure 1.(a) The three nesting domains for the WRF-Chem simulation, (b) the topography with respect to altitude (m) and the locations of the six observation sites over the YRD region in the innermost domain, d03.

Figure 2 .
Figure 2. Time series of surface O 3 concentrations (O 3 ), 2 m air temperature (Temp) and surface total radiation irradiance (TRI) observed at the NJ site.
and includes 158 chemical reactions among 36 species.The NCEP Final Global Forecast System Operational Analysis (FNL) data are used to provide the initial and boundary conditions of the meteorological variables for the WRF-Chem simulation.The chemical initial and lateral boundary conditions are extracted from the global chemical transport Model for Ozone And Related Tracers (MOZART)(Emmons et al., 2010;Horowitz et al., 2003).The Multi-resolution Emission Inventory for China (MEIC) (http://www.meicmodel.org/,last access: 11 November 2018) from 2016 is applied for the anthropocentric pollutant emissions, and the biogenic emissions are generated by the Model of Emissions of Gas and Aerosols from Nature (MEGAN)(Guenther et al., 2006).

Figure 3 .
Figure 3. Modeling validations in Taylor plots displaying the standard deviations and correlation coefficients of the simulated and observed meteorological elements and surface O 3 concentrations at the six YRD sites (Fig.1).The azimuthal angle represents the correlation coefficient, the radial distance denotes the ratio of standard deviation between simulations and observations, and the semicircles centered at the "REF" marker represents the normalized root-meansquare (NRMS) error(Taylor, 2001).

4
Analysis of regional O 3 transport 4.1 O 3 "reservoir" in the RL In order to analyze the development and evolution of the O 3 "reservoir" in the RL during the summer smog event, the time-altitude cross sections of the O 3 concentrations and potential temperature over the western (NJ site) and eastern (CZ, WX, SZ and SH sites) YRD regions are chosen www.atmos-chem-phys.net/18/16239/2018/Atmos.Chem.Phys., 18, 16239-16251, 2018

Figure 4 .
Figure 4. Time-altitude cross sections of O 3 concentrations and potential temperature over (a) the eastern YRD area covering the CZ, WX, SZ and SH sites from 14:00 on 24 August to 00:00 on 25 August, and (b) the western YRD NJ site from 00:00 to 12:00 on 25 August 2016.

Figure 5 .
Figure 5.The spatial distribution of O 3 concentrations and wind streamlines at a height of about 900 m at (a) 00:00, (b) 03:00, (c) 06:00 and (d) 09:00 on 25 August 2016.The blue thick lines with arrows represent the major regional O 3 transport routes from the eastern to the western YRD.

4. 3
Contribution of O 3 vertical mixing from the RL As discussed in Sect.4.1 and 4.2, an O 3 -rich air mass can be transported from east to west in the nocturnal RL over the YRD.Following the establishment of the CBL after sunrise, www.atmos-chem-phys.net/18/16239/2018/Atmos.Chem.Phys., 18, 16239-16251, 2018

Figure 6 .
Figure 6.Vertical sections of O 3 concentrations (contours) and atmospheric circulation (wind vectors) along the regional O 3 transport route (red solid line in Fig. 1b) from east to west over the YRD region.The W and E box columns mark the western YRD area surrounding NJ and the eastern YRD area covering CZ, WX, SZ and SH, respectively, from 24 to 25 August 2016.The vertical wind velocities are multiplied by 50 for the illustration of vertical circulations.

Figure 7 .
Figure 7.The contribution rates of vertical mixing (VMIX) and chemical reactions (CHEM) to surface O 3 concentrations at the NJ site during the 24-25 August 2016 period.

Table 3 .Figure 8 .
Figure 8.A diagram of the regional O 3 transport mechanism proposed in this study.

Table 1 .
Averages (mean) of maximum 8 h running mean surface O 3 concentrations (µg m −3 ), mean air temperature ( • C) over the periods of maximum 8 h running mean surface O 3 concentrations and daily maximum air temperature ( • C) with their standard deviations (SD) over the 22-25 August 2016 period observed at the NJ site in the western YRD.
• C. It is worth noting that the O 3 concentrations over the western YRD (NJ site) were 10-63 µg m −3 higher than those averaged over the eastern YRD region (CZ, WX, SZ and SH sites) during this period.

Table 2
the western YRD area (Figs.1b-2, Table2); these results could be difficult to interpreted in relation to the photochemical production.Both strong local photochemical production and atmospheric transport lead to high surface O 3 concentrations (Ja-

Table 2 .
Meteorological and environmental elements observed at the NJ site in the western YRD from 24 to 25 August 2016 and their daily differences (δx).