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 O3 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 O3 and NO2 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.

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 O3 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 O3 pollution occurred over the YRD region (Table 1); high surface O3 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 O3 concentrations exceeded 32.0 ∘C. It is worth noting that the O3 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.

Surface air temperature and solar radiation deeply affect photochemical production. Therefore, high O3 concentrations are generally associated with high air temperatures (Rao et al., 1992; Council, 1991). The hourly maximum O3 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 O3 concentrations (Fig. 2). However, we found from the observation of this particular summer smog episode that the air temperature and the O3 levels exhibited opposite overall changes in the western YRD area from 24 to 25 August (Fig. 2 and Table 2). The maximum 8 h running mean O3 concentrations and the maximum hourly O3 concentrations increased from 230.1 and 256.8 µg m-3 on 24 August to 284.8 and 317.2 µg m-3, on 25 August 2016, respectively, presenting an obvious daily O3 enhancement in the western YRD area. In contrast, the surface maximum total radiation irradiance and maximum air temperature decreased from 896 W m-2 and 34.1 ∘C to 872 W m-2 and 33.9 ∘C, respectively, during this 2-day period. This is noteworthy observational evidence, as the surface O3 concentrations in the ambient air increased while the daytime air temperature and total radiation irradiance decreased from 24 to 25 August in the western YRD area (Figs. 1b–2, Table 2); 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 O3 concentrations (Jacob, 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 (NO2) concentrations varied slightly during the 24–25 August period (Table 2), reflecting the lower impact from local photochemical production on the daily enhancement of O3. 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 O3 transport in association with vertical exchange over the YRD becomes an important mechanism in ambient O3 pollution for summer smog in the western YRD on 25 August 2016. This is explored using a modeling study in the following sections.

To investigate regional O3 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 spin-up 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 gas-phase chemical mechanism is selected using the Regional Acid Deposition Model, version 2 (RADM2) (Chang et al., 1990; Stockwell et al., 1984) 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).

Simulated wind speed, air temperature, relative humidity and O3 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 O3 pollution episode (Fig. 3). The correlation coefficients for the near-surface 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 O3 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 O3 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 O3 is very important in the analysis of the O3 budget; however, it was not possible for us to evaluate the vertical structure of O3 from the simulation. If there were observational data available for the O3 vertical profiles, the validation of the O3 vertical profiles could be carried out in a future study of the O3 budget. In general, the WRF-Chem simulated O3, air temperature, relative humidity and wind speed in the YRD show good agreement with the observations. The simulation also captures the observed changes of O3 and meteorology during the summer smog episode over the YRD reasonably well. Therefore, simulation data could be used to investigate the regional O3 transport and the underlying mechanism over the YRD during the summer smog period, as presented in the following sections.

In order to analyze the development and evolution of the O3 “reservoir” in the RL during the summer smog event, the time–altitude cross sections of the O3 concentrations and potential temperature over the western (NJ site) and eastern (CZ, WX, SZ and SH sites) YRD regions are chosen to present the temporal changes in the vertical structures of the O3 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 O3 profiles from afternoon to midnight on 24 August over the eastern YRD region. On the afternoon of 24 August, especially around 16:00, surface O3 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 O3 vertical mixing over the eastern YRD area (Fig. 4a). After sunset on 24 August, the near-surface O3 concentrations decreased sharply in the absence of photochemical production, the near-surface O3 consumption from nitrogen oxide (NO) titration and dry deposition. This resulted in the formation of a typical O3-poor stable boundary layer and an overlying O3-rich “reservoir” in the nocturnal RL over the eastern YRD region (Fig. 4a).

Figure 4b displays the hourly changes in the vertical profiles of O3 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 O3-rich air mass in the RL from being vertically transported to the surface. This resulted in the accumulation of an O3-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 O3-rich air mass in the RL and the near-surface O3-poor air mass; this redistributed the O3 concentrations in the daytime CBL and enhanced the surface O3 level during the morning (Fig. 4b).

We compare the temporal changes of the O3 “reservoir” in the nocturnal RL over the eastern and western YRD areas in Fig. 4a and b. It is interesting to note that the O3 concentrations in the eastern O3 “reservoir” obviously decreased (“leaked”) overnight on 24 August (Fig. 4a), while the western O3 “reservoir” was gradually strengthened and formed a high O3 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 O3 transport in the nocturnal RL could connect the eastern decreases and western increases in the overnight O3 “reservoir” within the RL (Fig. 4a and b). Therefore, we further investigate the regional O3 transport in the nocturnal RL over this area to interpret the observational evidence regarding the exacerbated O3 pollution and weakened photochemical production on 25 August in the western YRD (Figs. 1b–2, Table 2).

It is worth discussing why the nocturnal O3 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 O3 transport over the YRD that contributed to the O3 enhancement in the nocturnal RL over the western YRD area, Figure 5 presents the variations in the O3 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 O3 transport from the eastern to the western YRD region during the night from 24 to 25 August. This confirms our speculation that O3 transport in the nocturnal RL over the YRD connects the overnight changes in the O3 levels between the eastern and the western YRD sites (Fig. 4a and b). It is noteworthy that the regional O3 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 O3 concentrations from moving further west and converged O3 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 O3 concentrations and atmospheric circulation along the regional O3 transport route over the YRD region in order to further explore the mechanism of regional O3 transport during summer smog in this area. The vertical O3 concentration distributions were strongly controlled by the diurnal change of the BL structure. The daytime O3 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 O3-rich RL after sunset. Under the guidance of the prevailing easterly winds, the O3 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 O3 transport in the nocturnal RL over the YRD region. The regional O3 transport in the RL from the eastern YRD to the western YRD NJ site increased the O3 concentrations to 200 µg m-3 within the RL over the western NJ site during the sunrise hours on 25 August (Fig. 4b). The O3 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 O3 horizontal transport in RL to the O3 pollution on 25 August over the western YRD. The RL containing the O3-rich air mass over the western area, which was contributed by nocturnal O3 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 O3 level occurred around 10:00 in the morning and comprised downwards vertical mixing from the upper levels which caused the entrainment of O3-rich RL air to the O3-poor air mass at the ground (Figs. 6f and 7).

As discussed in Sect. 4.1 and 4.2, an O3-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, the O3-rich air mass can be entrained downwards to the surface (Mcelroy and Smith, 1993; Venkatram, 1977) and can contribute to the surface O3 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 O3 in the western urban NJ site from 24 to 25 August. Vertical mixing is initiated by convective and turbulent processes during the development of the daytime convective boundary layer, and chemical reactions constitute the net output of all O3 chemical reactions (Gao et al., 2016). The positive and negative contribution rates indicate the respective gain and loss of surface O3 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 O3 vertical mixing was significantly enhanced on 25 August with the largest contribution of O3, about 35 µg m-3 h-1, being twice as high as on 24 August (Fig. 7). Tropospheric O3 mainly results from photochemical reactions during the day (Seinfeld and Pandis, 1986); although the largest O3 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 O3 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 O3 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 O3-rich RL air mass, which was transported in the nocturnal RL from the eastern to the western YRD. Therefore, regional O3 transport in the nocturnal RL in combination with the diurnal changes of the boundary layer are revealed to be an important mechanism of regional O3 transport in eastern China.

Based on the simulated dry deposition, we calculated hourly changes in O3 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 O3 varied little over these 2 days, with only a slight enhancement on 25 August. This reflects the fact that O3 dry deposition exerted little impact on surface O3 changes during the 24–25 August period. Furthermore, the contribution of O3 dry deposition to tropospheric O3 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 NO2 concentrations and the total radiation irradiance (TRI) to examine the change in photochemical production rates. There were no apparent changes in NO2 or TRI from 24 to 25 August, which indicated that photochemical production exerted little impact on the high O3 level on 25 August compared with regional O3 transport in the nocturnal RL. The analysis of simulation results revealed that vertical mixing of the upper O3-rich RL with the daytime surface layer was a large contributor to O3 enhancement for summer smog in the western YRD on 25 August 2016 (Fig. 7).