Elevated layers of high ozone concentration were observed over the Seoul metropolitan region (SMR) in Korea by ozonesonde measurements during 6–9 June 2003. An analysis of the synoptic-scale meteorological features and backward trajectories revealed that the layers were associated with the long-range transport of ozone from eastern China. Further examination of the long-range transport process responsible for the development of these layers was performed using the Community Multiscale Air Quality (CMAQ) model. CMAQ demonstrated that the upward mixing of ozone by convective activity in eastern China and subsequent horizontal transport aloft in the periphery of a slow-moving high pressure system led to the development of thick ozone layers over the SMR. Through comparative simulation studies, it was found that the surface ozone levels in the SMR can be significantly enhanced by the vertical down-mixing of ozone from the layer aloft with the growing mixed layer. On average, about 25% of the surface peak concentration in a given area during a high-ozone episode was due to the influence of the ozone layer aloft developed by the long-range transport process.
Elevated ozone layers in the lower troposphere have the potential to enhance the ozone levels at the surface by down-mixing processes and likely influence the tropospheric chemistry and the global climate system (McKendry and Lundgren 2000). They are often developed by synoptically driven airflows associated with the regional transport of ozone and its precursors upwind (Daum et al. 1996; Fast and Berkowitz 1996; Blumenthal et al. 1997; Hidy 2000; Fast et al. 2002). Since these layers aloft can contribute significantly to local high-ozone episodes (Solomon et al. 2000; Fast et al. 2002), a clear understanding of their development and mixing processes is fundamental when diagnosing and controlling ozone pollution in specific regions where the regional transport influences are significant.
Recently, the influence of the long-range transport of pollutants of international significance has been examined continuously in the northeast Asia region (Pochanart et al. 1999; Kato et al. 2004; Naja and Akimoto 2004; Itano et al. 2006; Wong et al. 2007). In particular, in response to the rapid increase in emissions as well as the ambient concentrations of NOx, a critical precursor of ozone, in China (Naja and Akimoto 2004; Richter et al. 2005), the long-range transport of ozone-rich air masses passing over the Yellow Sea could make a significant contribution to the ozone levels in the lower atmosphere over downwind areas, especially the Korean Peninsula. The modeling results of Yamaji et al. (2006) suggested that continental air masses with high ozone concentrations were transported over the Yellow Sea and to the Korean Peninsula situated downwind under dominant westerly winds in May–June. A recent enhancement of the background ozone concentrations observed in the western coastal regions of South Korea (MOE 2007) also indicated the influence of such long-range transport. While this influence of transport on the ozone levels at the surface and aloft in South Korea is expected to be significant, it has not been evaluated quantitatively by both ozone measurements and three-dimensional gridded photochemical models.
Ozonesonde measurements were made in the Seoul metropolitan region (SMR) in northwest South Korea from 6 to 9 June 2003. During this time period, distinct elevated ozone layers with peaks exceeding 100 ppb were observed at higher altitudes on the first and second days. These layers aloft offered the opportunity of examining the long-range ozone transport over northeast Asia and its impacts on the local ozone concentrations in the downstream urban region. In this study, an analysis of the vertical ozone profiles, meteorological conditions, and backward trajectories is presented to examine the creation and origin of the elevated ozone layers. We also document for the first time the long-range ozone transport across the Yellow Sea responsible for the development of these elevated layers using the Models-3 Community Multiscale Air Quality (CMAQ) modeling system and quantify the potential impacts of the transported ozone plumes aloft on the surface concentrations in the SMR during a high-ozone episode.
2. Study area and ozonesonde measurements
The SMR is the national capital area located in the central western part of the Korean Peninsula (Fig. 1a). It has a population of 24.5 million (as of 2007) and contains three different administrative districts: the central city of Seoul, the western coastal city of Incheon, and surrounding Gyeonggi Province (Fig. 1b). The SMR occupies a broad area of relatively flat land with low mountains to the east and has many built-up areas, as shown in Fig. 1b.
From 6 to 9 June 2003, the vertical profiles of ozone and meteorological parameters were measured at the Seoul Olympic Park (SOP; 37°31′N, 127°07′E, 30 m MSL) in the eastern part of Seoul city (Fig. 1b). The SOP is surrounded by urban residential areas and located approximately 45 km from the western coastline. Ozone, temperature, humidity, and pressure soundings were obtained at a vertical resolution of approximately 5 m on average using electrochemical concentration cell (ECC) ozonesondes (EN-SCI Model 2Z; Environmental Science Corporation) coupled with standard radiosondes (RS-80-15; Vaisala, Inc.) that were released twice per day [around 0400 Korean standard time (KST, where KST = UTC + 9 h) and 1500 KST] during four consecutive days. The ozone data measured using the ECC ozonesondes had precision levels ranging from ±3% to 12% in the troposphere and the corresponding accuracies for the individual ozonesonde soundings were ±6% near the ground (Komhyr et al. 1995). The measurement accuracies for the temperature, humidity, and pressure were reported to be 0.2°C, 2%, and 0.5 hPa, respectively (Vaisala 2004).
3. Description of the modeling system
CMAQ is a comprehensive urban-to-regional-scale Eulerian photochemical modeling system developed by the U.S. Environmental Protection Agency (EPA; Byun and Ching 1999). It was applied successfully in eastern Asia to simulate the tropospheric ozone levels (Zhang et al. 2002; Yamaji et al. 2006). In this study, the fifth-generation Pennsylvania State University–National Center for Atmospheric Research Mesoscale Model (MM5; Grell et al. 1994) was used as the meteorological driver for the CMAQ modeling system. The MM5 was configured in nonhydrostatic mode, having 43 sigma layers up to 100 hPa, with the lowest level being 32 m thick (sigma = 0.996), and with the following four nested grids on the Lambert conformal grid projection (Fig. 1a): a coarse-grid domain (54-km grids, 64 × 49 array) that covers northeast Asia; a regional domain (18-km grids, 130 × 94 array), including eastern China with a high emission density; an intermediate regional domain (6-km grids, 91 × 85 array); and a fine-grid domain (2-km grids, 79 × 64 array) that covers the SMR and its surroundings. The model was run on these grids individually in a one-way nesting mode. The MM5 physical options used for the simulations were the Grell cumulus scheme, the Medium-Range Forecast Model PBL scheme for the PBL parameterization, the Dudhia simple ice moisture scheme, the Rapid Radiative Transfer Model longwave scheme, and the Noah land surface model. The MM5 simulations were carried out using the 24-category land-use data from the U.S. Geological Survey and the initial/lateral boundary conditions generated by interpolating the National Centers for Environmental Prediction (NCEP) Final (FNL) Operational Global Analysis Model fields with a 1° resolution at the standard pressure levels every 6 h, and the nudging of winds was applied to the 54- and 18-km grids.
The generated MM5 fields were processed using the Meteorology-Chemistry Interface Program (MCIP), version 3.0, to produce CMAQ-ready meteorological inputs. The MCIP options used to process the MM5 output fields are as follows: 1) the “pass through” option, where the PBL values estimated by the MM5 were used directly; 2) the radiation fields from the MM5 files were used; and 3) the Model-3/CMAQ dry deposition (M3DDEP) routine was used to calculate the dry deposition velocities.
Hourly gridded emissions from the regions of China and Japan within the modeling domains were generated using the anthropogenic emissions, including the lumped volatile organic compound species for the Statewide Air Pollution Research Center, University of California, Riverside (SAPRC99) mechanism, which were obtained from the Asia emission inventory dataset with 0.5° × 0.5° resolution developed in support of phase B of the Intercontinental Chemical Transport Experiment (INTEX-B) of the National Aeronautics and Space Administration (http://www.cgrer.uiowa.edu/EMISSION_DATA_new/index_16.html; Zhang et al. 2009). To prepare the emission inputs for South Korea, the Sparse Matrix Operator Kernel Emissions system was used to process the domestic anthropogenic emissions inventory developed under the 2004 Clean Air Policy Support System in Korea (Moon et al. 2006). The SAPRC99 chemical mechanism was used to process the emissions for the modeling period. More details on the processing methods and data used to prepare the anthropogenic emissions procedure were reported by Kim et al. (2008a). The Biogenic Emissions Inventory System, version 3.12 (BEIS 3.12), was used to estimate the biogenic emissions with vegetation data over South Korea (Kim et al. 2008b). A set of MM5–MCIP outputs was used to vertically allocate major point sources and calculate the biogenic emissions, depending on the meteorological conditions. Figure 2 shows an example of the distribution of NOx emissions within the 18-km modeling domain at 0000 UTC 2 June.
The CMAQ chemical transport model (CCTM), version 4.5, was run with 23 vertical layers compressed from the MM5’s 43 layers but keeping the first 12 layers up to approximately 1 km AGL (see open circles in Fig. 10). The CMAQ model top is the same as that used by the MM5, 100 hPa (approximately 15 km AGL). The horizontal domains are three grid cells smaller on each side than the nested MM5 domains at 18, 6, and 2 km. The CCTM simulation employs the following: the SAPRC99 gas-phase chemistry mechanism, Regional Acid Deposition Model–type aqueous chemistry and subgrid cloud processes, and the efficient Euler backward iterative solver. The advection scheme chosen for the model was a piecewise parabolic method. The multiscale horizontal diffusion was based on the local wind deformation, and the vertical diffusion was derived from the K-theory eddy diffusivity (Byun and Ching 1999; Byun and Schere 2006). The initial and boundary conditions of ozone and its precursors for the coarse domain (18-km grid) were derived from the default profile distributed with Models-3/CMAQ by the EPA. These conditions can be assumed to represent relatively clean air conditions (Byun and Ching 1999).
4. Results and discussions
a. Vertical ozone profiles from ozonesondes
The vertical profiles of ozone, potential temperature, and humidity from eight ozonesonde soundings made at the SOP from 6 to 9 June 2003 are shown in Fig. 3. To assist in the analysis of the ozone variations, the upper winds observed at Osan, the 850-hPa synoptic charts generated using the global gridded final run (FNL) meteorological data, and the surface meteorological parameters obtained at Seoul meteorological station (SMS) are presented in Figs. 4 –6, respectively.
The ozone profile at 0418 KST 6 June reveals the presence of a thick ozone layer (>100 ppb) with a maximum concentration of 114 ppb in the nighttime residual layer between approximately 1 and 3 km MSL. In the stable layer near the surface, the ozone concentrations exhibit a sharp gradient due to removal processes such as deposition and NO titration near the surface. The wind profiles (Fig. 4) indicate that northwesterly or westerly winds with speeds of roughly 5–7 m s−1 were dominant at the layer altitude from the previous afternoon. These winds were associated with the anticyclonic flow around a high pressure system approaching and passing through the Korean Peninsula (Figs. 5b and 5c). The wind speeds were high enough to advect the locally generated ozone out of the SMR in several hours, implying that the ozone layer aloft was transported from a long distance and that it was not a residual layer of ozone produced by local emissions on the previous afternoon. The observed high humidity of over 50% at around 2 km also suggests that this layer could not be characterized by subsidence from the upper troposphere. In addition, the persistent westerly surface winds with high speeds of up to 5 m s−1 during the previous 2 days (Fig. 6a) give a possible indication of regional-scale transport affecting the layer development. The origin of the layer and transport processes will be investigated in detail later.
From the 1455 KST sounding, a significant decrease in the ozone concentration was found, mostly at 1–3 km, compared with that in the early morning. This was mainly attributed to the large change in winds aloft between 0900 and 1500 KST associated with the passage of a surface cold front. Surface observations (Fig. 6b) showed clear signs of the passage of such a front with a significant temperature drop, an increased amount of clouds, and a small amount of precipitation lasting for 3 h during the early afternoon. These meteorological conditions were not conducive to photochemical production, thereby resulting in a low ozone concentration of about 40 ppb near the surface.
At 0403 KST 7 June, a relatively thin ozone layer with a clear peak of 101 ppb at 2 km developed. It is interesting to compare this vertical profile with the one taken in the afternoon of the same day. Since photochemical reactions do not occur at night, this new layer aloft clearly represents a regional-scale feature. The strong southwesterly upper winds at the layer altitude in advance of an intensified low pressure system (Fig. 5d) and vertical wind shear with height on the previous evening are likely responsible for the development of this thin layer. In the afternoon (1450 KST) of the same day, high ozone concentrations of up to 95 ppb were observed within a well-developed mixed layer to a height of 1.8 km. As the temperature rose above 25°C (Fig. 6b), photochemical reactions and convective mixing processes, in addition to the advection of urban plumes from nearby sources, led to the buildup of ozone in the mixed layer.
At 0404 KST 8 June, a high level of ozone (>90 ppb) still remained between 1.4 and 2.4 km. This was likely attributed to the effects of residual ozone from the previous-day production over the SMR. The weak upper winds from the previous afternoon, as evident in the winds between 800 and 850 hPa, may have been associated with this effect. A sharp decrease in the ozone concentration is also seen within the nocturnal stable layer, similar to those in the early morning of 6 and 7 June. At 1455 KST, a significant buildup of ozone with a maximum of 115 ppb at approximately 1.3 km was observed within the mixed layer. This is due to the local effects under light wind conditions and high temperature of over 30°C (Fig. 6b). At 0609 KST 9 June, there was a substantial decrease in the ozone concentration at all heights under the dominant strong southeasterly flows in association with the increased pressure gradient over the Korean Peninsula (Fig. 5f), which can bring relatively unpolluted air from the upwind rural areas to the east of the SMR. The ozone profile in the afternoon (1522 KST) shows a mainly uniform distribution of approximately 60–70 ppb. Relatively low ozone levels were observed in the mixed layer, in contrast to those observed on 7 and 8 June, even though a higher temperature exceeding 30°C was recorded. This would be due to the strong southeasterly winds during the daytime, which can inject a relatively clean air mass into the region.
b. Backward trajectories
To investigate the major origin of the high ozone layers observed in the early morning of 6 and 7 June, the 72-h isentropic backward trajectories were calculated for three levels (1, 2, and 3 km) using the National Oceanic and Atmospheric Administration’s (NOAA) Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model (Draxler and Rolph 2008) with the FNL meteorological data. All of the trajectories for 6 June (Fig. 7a) illustrated that the air masses arriving at the SOP at 1–3 km were influenced by continental air masses originating from eastern China with high emissions of ozone precursors (Klimont et al. 2002; Wang et al. 2005) and appeared to originate in the boundary layer about 1–2 days earlier, as was evident in their vertical motions. The 2- and 3-km trajectories had a long residence time over these areas before traveling across the Yellow Sea. The Geostationary Operational Environmental Satellite (GOES) IR-satellite image at midday on 4 June (Fig. 8) clearly shows the cloud-free regions in eastern China under the high pressure conditions (see Fig. 5a). This provided favorable conditions for enhancing the ozone concentrations in the lower atmosphere. Shan et al. (2008) reported that high ozone levels exceeding 100 ppb were observed at urban Jinan in Shandong Province on the same day. These air masses ascended gradually before arriving in western Korea due to vertical motions in the periphery region of the high pressure system. This indicates that the ozone-rich air within the boundary layer over eastern China could be transported to the SOP site, pointing to a regional influence on the development of the elevated ozone layer at 0400 KST 6 June.
For the trajectories on 7 June, quite different traveling paths of the air masses depending on their altitudes can be seen in Fig. 7b. The trajectories arriving at the 2-km level show that the air masses, which originated from the coastal areas of eastern China with high emissions (see Fig. 2), traveled over the Yellow sea before arriving at the SOP site. The quite different patterns of the trajectories arriving at 1, 2, and 3 km, which are associated with the advection of air masses of different origins, were mainly attributed to the formation of a thin ozone layer at 2 km on 7 June.
As a consequence of the analyses of the synoptic-sale meteorological features and backward trajectories, the development of ozone layers on 6 and 7 June 2003 can be mainly attributed to the long-range transport of ozone-rich air from eastern China. However, these analyses are insufficient to illustrate the long-range transport processes associated with the development and fate of the ozone layer and to quantify the potential impacts of the ozone layers aloft on the surface concentrations. The long-range transport process and vertical down-mixing effects will be examined in the following sections by using the CMAQ modeling system.
c. Model simulation and evaluation
The spatiotemporal variations of ozone were simulated and examined using the CMAQ modeling system configured with the options described in section 3. Model simulations were carried out over a sequential 8-day run with an initial spinup period of 48 h, beginning at 0000 UTC (0900 KST) on 2 June 2003 and ending at 0000 UTC on 10 June 2003.
The MM5 model performance in depicting the synoptic conditions was evaluated based on the observed meteorological data collected at 38 upper-air sounding sites (see Fig. 1a) from the Integrated Global Radiosonde Archive (IGRA) of the National Climatic Data Center. An overview of the IGRA was reported by Durre et al. (2006). Table 1 shows a statistical summary of the comparison of the simulated temperature, wind speed, and u- and υ-wind components for the 18-km grid simulations with the observations at three levels of the atmosphere (1000, 850, and 700 hPa) at 0000 and 1200 UTC for all 38 sites. The simulated values showed good correlation with the observations for all statistical measures. The improved agreement between the simulated and observed winds at both 850 and 700 hPa is encouraging because the modeling accuracy of the long-range transport of pollutants depends heavily on the upper winds. An example of the observed and simulated winds at the Osan site is shown in Fig. 9. There are some discrepancies, but overall the model reproduced the temporal variations of winds aloft well. The simulated thermal structure of the lower troposphere was also evaluated with the air soundings at the SOP site. Figure 10a shows the comparisons between the simulated and observed profiles of the potential temperatures at ozonesonde launch times. The vertical profile features and the evolution of the boundary layer structure are in good agreement with the observations. However, there are somewhat larger differences at the upper layer at 1500 KST 6 June. This would be due to the discrepancy in wind aloft, as evident in Fig. 9.
The reproducibility of the ozone concentration simulated by CMAQ was confirmed by comparing it with the observation data at the surface and aloft. Figure 11 shows a comparison of the surface ozone concentrations derived from the 18-km grid simulations with the observations at two background monitoring sites (Taean and Gosan) located at the western boundary of Korea (see Fig. 2). Overall, the simulated variations at both sites showed reasonable agreement with the observations. However, model simulations could not reproduce the sharp temporal variations of the observed ozone, which are likely attributed to mixing effects for the large grid spacing. Figure 10b shows the vertical distributions of the simulated and observed ozone concentrations over the SOP. With a few exceptions, the day-to-day variations and vertical distributions of ozone are well reproduced by the CMAQ model. In particular, the peak concentrations in the ozone layers aloft at 0400 KST on 6 and 7 June agree well with the observations. However, some underpredictions are observed at the upper layers, particularly at 0400 KST on 6 June. This might be caused by uncertainties in the vertical mixing processes between the middle and upper troposphere. The surface ozone concentrations in the SMR were also evaluated by comparing the simulated values at the lowest model layer (approximately 16 m AGL) with the measurements collected at 69 sites (see Fig. 1b) located within a nested fine 2-km grid domain. Figure 12 shows that the diurnal and multiday evolution of ozone and the range of values in the SMR were well captured with a strong mean correlation coefficient of 0.81 and a reasonable bias (mean bias = −2.0 ppb) and error (RMSE = 15.1 ppb), although the afternoon concentrations were somewhat lower than the observations. This reasonable agreement between the simulations and observations suggests that the multiscale emissions, wind fields, and transport processes were treated reasonably well in the CMAQ modeling system.
d. Simulation of long-range ozone transport and elevated layer development
To confirm the long-range ozone transport responsible for the observed ozone layers and to identify the extent of the ozone plumes during transport, the horizontal distributions of the simulated ozone and wind vectors at approximately 2 km (model layer 15), where ozone peaks in the layer aloft were observed at the SOP, are shown in Fig. 13 for at 0400 and 1500 KST from 4 to 9 June. The vertical cross sections indicated in Fig. 13b are also presented in Figs. 14a–d to identify the vertical behavior of the ozone during its transport.
On 4 June, eastern China, having high emissions of ozone precursors, was under the influence of a slow-moving high pressure system. High ozone concentrations (>80 ppb) were simulated in and around Shandong Province at 1500 KST (Fig. 13b). This resulted from both the upward transport of ozone that had effectively accumulated within the daytime convective boundary layer and its horizontal advection by higher winds aloft to the outer edges of the high pressure system. Figure 14a illustrates that a high level of ozone was mixed up to the top of the deep boundary layer, in which the vertical potential temperature gradient is very small. Subsequently, ozone-rich air in the upper region of the boundary layer was advected toward the Yellow Sea by the strong westerly synoptic flow as the high pressure system moved eastward. At 0400 KST 5 June, the region of high ozone concentration extended across the Yellow Sea, and the ozone plumes that had been transported hundreds of kilometers across the Yellow Sea reached the west coast of the SMR (Fig. 13c). Figure 14b shows that a thick layer of ozone was formed over the sea during its transport. This layer was clearly decoupled from the stable marine layer and was confined mainly to two levels of isentropic surfaces (300 and 304 K), where high ozone concentrations were present within the mixed layer over the land during the afternoon of 4 June.
A broad region of high-level ozone was developed at 1500 KST 5 June in the periphery of a high pressure system (Fig. 8d). The obvious increase in the concentrations of ozone over the east coast of China would be initially responsible for the photochemical ozone production with the daytime deep convections. As the high pressure moved gradually toward the Korean Peninsula, the ozone plumes aloft that had reached the western coast of the SMR in the early morning were transported slightly farther inland by the westerly synoptic flow. Figure 14c shows that the ozone layer over the sea became thicker due to the warm advection aloft carrying additional ozone-rich air masses from the east coast of China. Over the SMR, the vertical mixing of ozone from the transported layer aloft to the surface can be seen as the mixed layer deepens over the region. This indicates that the long-range transported ozone can contribute to the surface concentrations by a downward mixing process. A more detailed analysis of this process will be discussed later.
The transport of ozone-rich air to the SMR by anticyclonic flow was persistent by the morning of 6 June. At 0400 KST 6 June (Fig. 13e), high concentrations of about 100 ppb were simulated over most of the SMR. Some of the ozone plumes that had traveled over the Yellow Sea were advected to North Korea by strong southeasterly winds in advance of the cold front approaching the Korean Peninsula. Figure 14d confirms the development of a well-defined ozone layer over the SMR, corresponding well to the ozonesonde measurements at the SOP. On the other hand, some layers of ozone in the upper region over the Yellow Sea would be caused by the synoptic-scale lifting process in advance of the cold front. In contrast with this evolution of the ozone layer (Figs. 14a–d), the ozone transport and its layer development are not found from a control simulation with zero emissions in China (Figs. 14e–h). This makes it clearer that the long-range transport from China is responsible for the ozone-layer development over the SMR.
A large decrease in the ozone concentration occurred over the SMR at 1500 KST 6 June (Fig. 13f), which is due to the large wind shift associated with the passage of a surface cold front during the afternoon of this day. After this front passed through the SMR, the following upper low pressure system extended to the Korean Peninsula (see Fig. 4d). The resulting southwesterly winds in advance of the system carried ozone-rich air in the narrow region over the Yellow Sea into the SMR. At 0400 KST 7 June, a high ozone concentration of over 80 ppb was simulated over the SMR, as shown in Fig. 13g.
As the trough that had affected the Korean Peninsula weakened and the high pressure system over the Yellow Sea moved northeastward, the transport of the high ozone plumes from eastern China to North Korea was simulated (Figs. 13h and 13i). The distinct change of the synoptic flows that occurred from midday on 8 June caused the ozone concentrations aloft over the SMR to gradually decrease (Figs. 13j–l), because the strong southeasterly wind carrying a relatively clean air mass continued to affect the region. On the other hand, increased ozone levels were found offshore near the coast of western Korea at 0400 and 1500 KST 9 June, due to the influence of the ozone plumes transported from the inland areas of South Korea.
These simulation results illustrate that vertical mixing processes over the parts of eastern China situated upwind and subsequent long-range transport aloft across the Yellow Sea led to the development of ozone layers over the SMR and that the synoptic flow conditions were closely associated with the regional patterns of ozone aloft.
e. Contribution to surface concentration
A layer with a high ozone concentration has the potential to mix downward to the surface via a deepening mixed layer (Blumenthal et al. 1978; McElroy and Smith 1993; Zhang and Rao 1998; Hidy 2000; Solomon et al. 2000; McKendry and Lundgren 2000). The 5 June episode provides a good example of the vertical down-mixing of ozone from the layer aloft created by long-range transport based on the existence of both the ozone layer aloft and convective activity over the SMR during daytime (see Figs. 13d and 14c). Although high-ozone layers were observed over the SMR in the early morning on 6 and 7 June, the down-mixing effects on the surface concentrations on both days are expected to be insignificant, since the layers aloft did not persist during the day under the influence of the passage of a cold front on 6 June and strong southwesterly flow on 7 June.
Indirect observational evidence for the vertical mixing of ozone was obtained from the temporal variations of the surface concentrations on 5 June. Figure 15 presents the hourly ozone measurements on 5 June at nine monitoring sites, denoted as circles in Fig. 16, where the ozone level exceeded the 1-h Korean ambient air quality standard of 100 ppb. This evidence shows that the ozone concentrations were observed to rise sharply at most sites during the period from midday to early afternoon, especially from 1300 to 1400 KST when they increased at about 28 ppb h−1 on average (maximum: 48 ppb h−1). This rapid increase suggests that the vertical down-mixing processes were associated with high ozone concentrations at the surface.
To estimate quantitatively this effect, an additional 4-km simulation (simulation B) was performed using modified boundary conditions for ozone during the 24-h period on 5 June and was compared with the base-case simulation (simulation A) described in section 3. Simulation B employed the boundary conditions for ozone extracted from the coarse grid simulations without China’s emissions between approximately 0.6 and 3 km. This roughly corresponded to the height between the two isentropic surfaces of 297 and 306 K, where ozone-rich air masses entered into the 4-km domain from the early morning on 5 June. The concentrations of the other species at the boundaries were kept identical to those of simulation A. In this way, the influence of the long-range transport of ozone aloft was eliminated, except for the transport of precursors. Thus, simulations A and B can be compared to estimate the contribution of the ozone layer aloft generated by long-range transport to the surface ozone levels.
Figures 16a–d show the horizontal distributions of the simulated ozone concentrations and wind vectors at the surface and aloft (approximately 1.6 km) at 1400 KST 5 June, when the difference in the surface concentrations between the two simulations is the largest. Figure 16a shows that high concentrations (>80 ppb) were distributed in most of the domain, especially in the western boundary, reflecting the influence of the ozone transport aloft. Relatively low concentrations are also seen in Seoul, because the strong convective activity in the mixed layer (Fig. 16f) led to the upward mixing, from the surface, of air with low ozone and high NO concentrations. This distribution is clearly compared with that from simulation B (Fig. 16b) using the background boundaries for ozone. The relative high ozone levels in the southeastern part of the domain in both simulations are due to the vertical motion of locally produced ozone within the mixed layer.
At the surface (Figs. 16c and 16d), high ozone levels are seen in the downwind eastern SMR. They are mainly responsible for the combined effects of the local production and urban-scale transport of ozone and its precursors from the upwind coastal regions. As expected, the ozone levels in simulation A are relatively high compared with those in simulation B, as the ozone aloft was entrained into the growing mixed layer. Large concentration differences with a maximum of 32.1 ppb are found in Seoul city and around it (Fig. 16e), roughly corresponding to an area of high mixed-layer heights (Fig. 16f). Overall, the area with large differences (>18 ppb) encompasses all nine sites (circles in Fig. 16e) at which the observed ozone concentrations exceeded 100 ppb during the afternoon. This indicates that the vertical down-mixing of ozone aloft contributed significantly to the high-ozone occurrences at the surface.
To better understand the vertical mixing processes associated with the deepening of the mixed layer, the cross sections indicated in Fig. 16c are presented in Fig. 17. At 1000 KST, the ozone layer aloft passing over the SMR, which is separated from the stable marine boundary layer and shallow convective layer inland, is clearly shown in simulation A. Here, the lowest level of the ozone layer is roughly indicated by the thick line corresponding to an isentropic surface of 297 K in Fig. 17. In contrast with this, simulation B shows a relatively uniform and low concentration of approximately 40–50 ppb in the upper region. As the ozone layer passed through the SMR and the top of the mixed layer inland reached approximately 1 km at 1200 KST, the 297-K isentrope was merged into the surface-based mixed layer, indicating that the ozone-rich air aloft apparently began to enter into the mixed layer and led to the increased ozone concentration near the surface shown in simulation A. Significant differences between the two simulations of over 25 ppb are found in the inland mixed layer. With the further growth of the mixed layer at 1400 KST, most of the ozone layer aloft intersected the mixed layer in the eastern region of the cross section, leading to a maximum difference in concentration between two simulations of about 31 ppb near the surface.
To further quantify the contribution of this vertical mixing effect to the enhancement of the surface ozone concentration, the averaged surface concentrations within 30 km of the SOP obtained from simulations A and B and their differences are shown in Fig. 18. The concentrations from two simulations are almost identical by early morning because the ozone layer aloft has been decoupled from the surface. As the mixed layer developed starting from the late morning, this difference gradually increased and reached a maximum of about 22 ppb at 1400 KST. The difference in the peak concentrations between simulations A and B is 20.9 ppb at 1500 KST, indicating that the surface peak concentration over the SMR is enhanced by up to 25% by the vertical down-mixing processes.
The vertical down-mixing of ozone precursors from the layer aloft can also contribute to the surface ozone levels when the transported air masses are not completely aged (Fast et al. 2002). However, in our simulations, this contribution was small enough to be ignored, because the long-range-transported ozone-rich air masses would be well aged photochemically, since they were not exposed to fresh emissions over the Yellow Sea for periods of 24 h or more. A comparison of the results obtained from simulation B and a test run using the same boundary conditions as those in simulation B, except for holding the precursors aloft at the background concentrations (not shown), revealed that the test run produced ozone concentrations only 1–2 ppb less than those of simulation B for the entire domain. This demonstrates that most of the increase in the ozone concentrations brought about by the vertical mixing process is due to the down-mixing of ozone loft.
Elevated layers of ozone that were transported over distances of up to 1000 km downwind of the source regions were observed over the SMR in early June 2003. An analysis of the synoptic meteorological features and backward trajectories suggested that the layers aloft were created by the long-range transport of ozone originating in the boundary layer over eastern China about 1–2 days earlier. CMAQ simulations over northeast Asia demonstrated that ozone-rich air masses emanating from the well-developed mixed layer over eastern China were transported aloft across the Yellow Sea by synoptic-scale winds in the periphery of a slow-moving high pressure system, which led to the development of thick ozone layers between 1 and 3 km over the SMR from early morning on 5 June to midday on 6 June. The simulations also showed that the southeasterly flow in advance of the upper low pressure system transported ozone-rich air over the Yellow Sea to the SMR, leading to the development of a relative thin layer of ozone in the early morning of 7 June.
To quantify the potential impacts of the transported ozone plumes aloft on the surface concentrations in the SMR, various 2-km grid simulations (simulations A and B) with and without considering the ozone plumes aloft entering the model boundaries were performed during a period of 24 h on 5 June. A comparison of simulations A and B showed that significant differences (maximum of 32.1 ppb) in the surface ozone concentrations occurred along the deepening of the mixed layer, which explained well the observations showing a high rate of increase in ozone concentration in the early afternoon. The vertical distributions clearly illustrated that the ozone layer aloft was intercepted by the growing mixed layer and thereby contributed to the surface concentrations. On average, the maximum difference of the afternoon peak ozone concentrations between simulations A and B in a given area is 20.9 ppb, indicating that the surface peak concentration is enhanced by up to 25% by the vertical down-mixing process.
In this study, the observational and modeling analysis provided evidence that the long-range transport of ozone from eastern China can create ozone layers over the Yellow Sea and the downwind coastal urban area in South Korea under certain meteorological conditions. Furthermore, comparative simulations revealed that the transported ozone plumes aloft are responsible for a large fraction of the peak surface ozone concentrations in the case of a high-ozone event. While we have significantly expanded our understanding of the development of the ozone layer arising from the long-range transport across the Yellow Sea, our knowledge of the physical and chemical processes of ozone production and transport aloft over northeast Asia remains limited due to the lack of observations of key trace species over the region and to the uncertainties in the model input data, especially for the boundary conditions and emission inventories. More research is required to explore these uncertainties and improve the modeling and interpretation of long-range transport processes.
This work was funded by the Korea Meteorological Administration Research and Development Program under Grant CATER 2009-3308.
Corresponding author address: Prof. Yoo-Keun Kim, Division of Earth Environmental System, Pusan National University, Busan 609-735, South Korea. Email: email@example.com