Impact Comparison of Synoptic Meteorology and Nationwide/local Emissions on the Seoul Metropolitan Area during High PM Multi-event and Non-event Days

Meteorology and emissions play very important roles in the concentrations of air pollutants during severe haze/smog periods. This study compares the impacts of synoptic meteorology and nationwide/local emissions during high PM10 multi-event and non-event days in the Seoul Metropolitan Area (SMA). The multi-event and non-event cases were selected based on daily mean PM10 concentrations in Seoul from January 2014 to March 2019. The multi-event cases in spring and winter were closely associated with weak synoptic winds, while that in autumn was due to the strong winds at the rear side of a strong cold front, which induced the Asian dust event in northeastern China and Korea. The multi-event case in spring was found to be mainly due to series of migratory anticyclones, while winter case was due to the stagnant system after northerly winds. The surrounding low pressure systems as well as high pressure systems could be important to determine whether the synoptic systems would be stagnant or not. The fractional contributions of SMA emissions to the mean PM10 and PM2.5 concentrations were 24%–35% and 22%–35% for the multi-event cases, respectively. The contributions to the maximum PM10 and PM2.5 concentrations were larger than those to the mean concentrations by 16%–23% and 19%–26% for the multi-event cases, respectively.


INTRODUCTION
The Seoul metropolitan area (SMA) is a mega city with a population of 26.1 million (http://kosis.kr) (Kim et al., 2018).Multi-event days with high PM 10 concentrations in the SMA have been an urgent issue for the South Korean society (Park, 2014;Jo and Kim, 2013).In 2019, the Korean government established "The National Council on Climate and Air Quality", a presidential-level agency, with a focus on reducing PM concentrations as well as building international cooperation and domestic strategies.air quality in this area is impacted by both local emissions and long-range transboundary transport of air pollutants (Seo et al., 2017;Jo and Kim, 2013;Kim et al., 2012;Park et al., 2005).Further knowledge of both local emissions and long-range transboundary processes is needed to develop effective emission mitigation strategies in the SMA as well as throughout Korea (Park et al., 2019(Park et al., , 2018;;Park, 2016).The dominant emission source of PM 10 and NO 2 in the SMA is diesel-powered vehicles (Koo et al., 2008).During the last two decades, there were some studies on the domestic or foreign contributions to the high PM concentration in Korea.Foreign contributions were reported to have ranges from about 30% to about 70% depending on cases and periods (NIER, 2017;Park et al., 2015).According to "The Expert Meeting for Long-range Trans boundary Air Pollutants in Northeast Asia" report, the contributions of China emissions to sulfur and nitrate deposition in Korea were 43% and 45% in January and March 2002, respectively (NIER, 2006).The contribution of foreign or China emission to PM concentration in Korea exhibited larger difference among researchers due to the difference of model, analyzed period, and methodologies.The Korea-US Air Quality (KORUS-AQ) campaign showed that China contributed 34% to PM 2.5 concentrations from 2 May to 12 June 2016 (NIER, 2017), and 68% during May to June 2016 (Choi et al., 2019).According to the summary report of the 4th stage LTP project, China's contribution to major cities in Korea was 32% in 2017.But, the foreign contribution (including China, North Korea, Mongolia) on PM 10 concentration in Korea was reported to reach 60% yearly (Kim et al., 2017), 66% (Daejeon city) in February 2014 ( Jung et al., 2019), 40-65% during 2010-2017 (Bae et al., 2020).
According to the previous studies on high PM 10 concentration events in Korea, meteorological condi tions contributed to high PM 10 concentration more than emissions (Park et al., 2019;Seo et al., 2018Seo et al., , 2017;;Park, 2014).For example, higher PM 10 concentration events in the SMA frequently occurred on warmer days (Park et al., 2019;Park, 2014).Moreover, during the high PM 10 concentrations in spring, the SMA was usually governed by local circulation systems under weak synoptic winds (Park et al., 2019).There was no clear criteria to classify weak and strong synoptic winds.And the criteria can be dependent on the location.Conventionally, the synoptic winds were classified as weak when the wind speed at 850 hPa is less than about 5 m s 1 (Park, 1994).Some-times, the Korean Peninsula was influenced by warm and stagnant high-pressure systems (Park et al., 2019;Seo et al., 2018Seo et al., , 2017;;Park, 2016).Some simulation studies had suggested that severe PM 2.5 episodes were driven by stable synoptic weather conditions rather than an abrupt increase in emissions (Kim et al., 2017;Li et al., 2016).Besides the synoptic meteorology and local circulation, vertical distributions of meteorological and/ or turbulent parameters in the atmospheric boundary layer could also lead to high PM 10 concentrations (Li et al., 2018;Pahlow et al., 2005;Zhou et al., 2005).Day-today PM 10 concentrations in Belgium depended more on meteorology than on anthropogenic emissions (Hooyberghs et al., 2005).Nonetheless, high PM 10 concentration cases tended to be difficult to categorize by their meteorological features in advance.There were many uncertainties and complexities in the meteorological mechanism affecting high PM 10 concentrations.Moreover, there have not been many studies on the multi-day high PM concentration events.
The main objective of this study was to compare the synoptic meteorological characteristics between the high PM 10 multi-event days and the non-event days in the SMA.The fractional contributions of the SMA emission to the PM 10 /PM 2.5 concentrations were quantified for the multi-event and non-event cases by using The Air Pollution Model (TAPM).

DATA AND METHODS
To select the multi-event case, hourly mean PM 10 , PM 2.5 , SO 2 , and NO 2 concentrations, monitored at the Seoul Station from 1 January 2014 to 31 March 2019 (PM 2.5 data were not available in 2014), were used (http://www.airkorea.or.kr;Fig. 1).The Seoul Station is located at the center of Seoul and is surrounded by many line and area emission sources (Fig. 1).Surface meteorological variables such as wind speed and direction, temperature, and global solar radiation were obtained from the Songwol Station (Seoul Automatic Synoptic Observation Station; https://data.kma.go.kr;Fig. 1).The Songwol Station is located 1.2 km northwest of the City Hall.
When the daily mean PM 10 concentration in the Seoul Station was greater than the air quality standard for PM 10 concentration in Korea, 100 μg m −3 , the day was classified as a high PM 10 event day (Park et al., 2019).
The 7 days, including 5 consecutive event days in spring and winter and 2 consecutive event days in autumn, were classified as multi-day event cases (hereafter, multievent cases).Korea had no high PM 10 event days in summer and not so many event days in autumn due to the seasonal dependency of prevailing winds and energy usage with respect to temperature in East Asia.When the daily mean PM 10 concentration was below 50 μg m −3 , the day was classified as a non-event day.The non-event case was chosen as the case of 7 consecutive non-event days in the same month with a multi-event case (Park et al., 2019).
To simulate the 3-dimensional fields of meteorological and air quality concentrations, TAPM was used.TAPM is composed of two prognostic subsections: meteorological and chemical modules (Hurley, 2008).It was set to have three domains with horizontal resolu-tions of 12 km, 4 km, and 2 km.Each domain had the same number of grid (100 × 100 × 25) with the same center (37°33.97N, 126°58.67E).Domain 1 includes South and North Korea, while Domain 3 includes SMA (Seoul, Incheon, and Gyeonggi Province).The Global Assimilation and Prediction System (GASP) data with a resolution at the equator of 0.75° (ftp://ftp.csiro.au/TAPM)were used for the initial boundary field of meteorology (Schulz et al., 2007).It is a spectral model run operationally at every 6 hours by Australia Bureau of Meteorology Research Center.Observation data are assimilated by a generalized statistical interpolation scheme (Steinle, 2005) using an iterative solution similar to that described by (Cohn et al., 1998).Land-use and land-cover data (1 km × 1 km) from the USGS (United States Geological Survey) were used as input parameters.Fig. 2 illustrates the spatial distribution of total nationwide emissions in 2015 (reference year) of PM 10 , NO x , and SO 2 with a resolution of 4 × 4 km 2 .Nationwide emissions imply emission rates from all South Korean boundary, while SMA emissions mean emission rates from Seoul City, Incheon City, and Gyeonggi Province.Emission rates were 232,794 ton y −1 , 1,000,064 ton y −1 , and 351,010 ton y −1 nationwide, and 44,385 ton y −1 , 289,431 ton y −1 , and 37,223 ton y −1 in the SMA for PM 10 , NO x , and SO 2 , respectively (Park et al., 2019).The SMA emissions account for 19.1%, 28.9%, and 10.6% of nationwide emissions for PM 10 , NO x , and SO 2 , respectively.Emission rates of PM 10 and NO x were highly dependent on traffic densities (Park et al., 2019;Park, 2014).Note that emission rates from outside South Korea (such as China, North Korea, and Japan) were not considered in this study.Instead, initial and background concentration and emission rates from South Korea were used to simulate the air pollutant concentrations.

AIR POLLUTANTS CONCENTRATION
According to the above criteria, the multi-event and non-event cases were chosen as the periods from 28 February to 6 March and from 8 to 14 March 2019 in spring, from 24 to 30 and from 14 to 20 November 2018 in autumn, and from 22 to 28 and from 12 to 18 February 2014 in winter, respectively (Table 1).The chosen multievent cases in spring and winter included 6-consecutive high PM 10 event days from 1 to 6 March 2019 and from 23 to 28 February 2014, respectively.Only 2 cases with 3 or more consecutive event days were found from 2014 to 2019.Fig. 3 shows the time series of PM 10 , PM 2.5 , NO 2 , and SO 2 concentrations for 7 days (from 2 days before to 4 days after the peak value) for multi-event and nonevent cases.PM 10 and PM 2.5 concentrations thro ughout the event period were much higher than those throughout the non-event period, except in autumn (Fig. 3).High PM concentration in autumn was due to Asian dust, not smog/haze.There were not so many differences between the multi-event and non-event cases for NO 2 and SO 2 , except for SO 2 in winter.
Median concentrations of PM 10 , PM 2.5 , NO 2 , and SO 2 at Seoul station during 7 days for the event and nonevent cases were calculated.The median of PM 10 concentration exhibited its highest value (126.5 μg m −3 ) in winter and its lowest value (61.5 μg m −3 ) in autumn for the multi-event case, while for the non-event case its highest value (46 μg m −3 ) was exhibited in spring and its lowest value (27 μg m −3 ) in winter (Table 1).As a result, the PM 10 concentration ratio of the multi-event case to the non-event case was as high as 4.7 in winter and as low as 1.5 in autumn (Table 1).
The ratios of the case mean PM 10 concentration to monthly mean concentration were 1.9, 1.4 and 2.2 for the multi-event cases, and 0.8, 0.8, and 0.6 for the nonevent cases in spring, autumn, and winter, respectively (Table 2).
Because the TAPM did not include the emission from outside South Korea, the initial background concentrations were set as the monthly mean observed ones on the same month for the multi-event and non-event cases (Table 3).

SYNOPTIC METEOROLOGICAL CHARACTERISTICS
Local meteorological factors play an important role in the accumulation of both primary pollutants and secondary aerosol precursors as well as secondary formation processes.They are largely controlled by synoptic-scale conditions (Seo et al., 2017;Zheng et al., 2015).To investigate the synoptic meteorology for the multi-event and non-event cases, surface and upper weather charts were used (http://data.kma.go.kr).
Fig. 4 shows the surface weather chart at 00 UTC during the multi-event (1-6 March 2019) and nonevent (9-14 March) cases in spring.During the multi-event case, a weak Siberian high-pressure system and migratory anticyclone systems were located over the Korean Peninsula (Figs. 4a-f).Due to the high-pressure system to the west and low-pressure systems to the east, westerly or northwesterly winds were recorded at the Seoul Station on 28 February and 1 March 2019 (Fig. 4a).Several migratory low-pressure systems generated over southern China or southern Japan on 2-3 March (Figs.4b, c), evolved, and moved to eastern or northern Japan on 5 March (Fig. 4e).The migratory low-pressure systems to the south or east hindered the weak high-pressure system over Korea from moving eastward, making the system stagnant during 2-5 March 2019 (Figs.4b-e).A low-pressure system was centered near eastern Russia, and cloudy conditions were observed over Korea except on 13 March 2019 during the non-event case in spring (Figs.4g-l).There was little precipitation during 11-14 March.Synoptic systems near the Korean Peninsula can move as fast as 50 km h -1 .A low-pressure system, centered over Taiwan on 9 March 2019 (Fig. 4g) evolved, moved to northern Japan on 12 March 2019 (Fig. 4j), and combined with another far eastern-Russiaoriented low-pressure system on 13 March 2019 (Fig. 4k).
The multi-event case on November in 2018 was due to the passage of Asian dust.An array of high-pressure systems extended from southwestern China and moved to eastern Japan on 25 November 2018 during the multievent case (Figs.5a-f ).A low-pressure system, centered over Lake Baikal on 25 November 2018 (Fig. 5a), moved southeastward.A cold front accompanied by a low-pressure system moved over the Korean peninsula on 26-27 November 2018 (Figs.5b, c).An Asian dust event, due to the strong wind at the rear side of the cold front, occurred in Seoul on 26-27 November 2018 (Figs.5b, c).
There was a high-pressure system centered over Mongolia and a low-pressure system located over the Kamchatka peninsula from 15 to 18 November 2018 during the non-event case in autumn (Figs.5g-j).The high-pressure system extended to the Korean peninsula and central Japan during this time.Several weak lowpressure systems passed over the Korean peninsula on 19 November 2018 (Fig. 5k).Trace precipitation (0.2 mm) was recorded in Seoul on 20 November 2018 (Fig. 5l).
There was a multi-event case on late February in 2014 (Fig. 6).A migratory anticyclone, centered over northeastern China on 23 February (Fig. 6a), moved to the Korean peninsula on 25 (Fig. 6b), stayed over it during the next two days (Figs.6c, d), and again moved eastward the next day (Fig. 6e).Two very small and weak low-pressure systems, centered over the northern Korean peninsula and southern Japan, respectively, and passed over the Korean peninsula on 26-27 February (Figs.6c, d).
The boundary between a weak Siberian High-pressure system and a relatively strong low-pressure system, centered between Hokkaido and Kamchatka, was located over Korea throughout the non-event period in winter (Figs.6g-l).A weak low pressure system with a central pressure of 1013 hPa, centered over eastern Taiwan on 13 February 2014 (Fig. 6g), moved northeastward and reached northern Japan on 17 February (Fig. 6k).The low-pressure system evolved into a strong one with a central pressure of 976 hPa on 17 (Fig. 6k).Due to this high-pressure system over northern Korea and the low-pressure system over southern Korea, easterly winds were recorded in Seoul on 13-14 February 2014 (Figs.6g, h).Fig. 7 shows the schematics of synoptic meteorology during the event periods in spring, autumn, and winter.A series of migratory anticyclone systems had formed between two weak cyclone systems located at 30°N and 40°N, and moved into and passed Korean peninsula for spring event case (Fig. 7a).For the autu mn event case, a cold front had passed northeastern China on 27 November 2018 (Fig. 7b).Strong winds and Asian dust were accompanied by the cold front.For the winter event case, a high pressure system had moved from northeastern China to Korea, and had been stagnant during two days (Fig. 7c).As a rule, synoptic winds were relatively weak except for autumn Asian dust event case.Stagnant high-pressure or migratory anticyclone systems were dominant near the Korean peninsula for the multi-event cases in spring and winter.The surrounding low pressure systems were also important to determine whether the systems would be stagnant or not.Weak winds and stagnant synoptic systems blocked the horizontal transport of air pollutants, so air pollutants emitted from local sources could accumulate in the atmosphere.Migratory anticyclone systems did continuously transport air pollutants from the source region of China to Korea.On the contrary, for the non-event cases, the winds were not as weak and lower pressure systems were dominant.Strong winds could easily ventilate the polluted air, then made the concentration of air pollutants low.

1 Model Evaluation
Fig. 8 shows the time series of temperature, relative humidity, rain, and wind speed and direction observed at the Songwol Station and simulated by TAPM for the multi-event and non-event cases (Fig. 1).Simulated meteorological variables exhibited similar temporal variation with the observed ones, except for the rain.The meteorological variables simulated by TAPM were validated by those observed at the Songwol Station.Statistical performance measures such as the mean bias (MB), root-mean-square error (RMSE), and index of agreement (IOA) were calculated for each case (Table 4).IOA is a frequently used measure of how successful is the simulation of the variable regarding the observed one.The IOA with a value greater than 0.5 is an index of a good simulation (Hurley et al., 2008).
Temperature was underestimated for both cases in spring and the non-event case in winter, while it was overestimated for both cases in autumn.Relative hum-idity was slightly overestimated except for the multievent case in winter.Wind was underestimated for both cases in winter.The IOAs for temperature and relative humidity were higher than those for wind speed and direction (Table 4).
The simulated concentrations for PM 10 and PM 2.5 show strong diurnal patterns and reveal an underestimation for the multi-event cases and an overestimation for the non-event cases in spring and winter (Fig. 9).It might be mainly related to the same background concentration, which was higher than the mean PM concentration during the non-event case and lower than that during the event case.By and large, the IOAs for PM 2.5 concentration were larger than those for PM 10 because TAPM has certain weaknesses in simulating the effect of long-range transported contribution.The simulated NO 2 concentrations were generally in good agreement with the observed ones except for the nonevent cases in spring and autumn (Fig. 9).The SO 2 concentration was more or less overestimated for all cases (Fig. 9).This discrepancy seems to be associated mainly with the assumption of constant emissions.The Table 4. Mean bias (MB), root mean square error (RMSE), and index of agreement (IOA) between simulated and observed temperature (°C), relative humidity (%), wind speed (m s −1 ), wind direction (deg), PM 10 (μg m −3 ), PM 2.5 (μg m −3 ), NO 2 (ppb), and SO 2 (ppb) concentrations for each case.IOAs in this study were found to be better than those in recent similar studies in urban areas except those for the multi-event cases in spring and autumn (Park et al., 2018;Hurley et al., 2008).

2 Wind and Concentration Fields
Fig. 10 shows the horizontal distribution of PM 10 with the wind vector simulated with nationwide emissions and without SMA emission for the multi-event and non-event cases.The patterns are strongly governed by the wind; upwind regions showed the lowest concentrations while downwind regions showed the highest concentrations.The areas with the confluent wind exhibited the highest concentrations while the areas with the divergent wind exhibited the lowest concentrations.
In spring, the overall concentration field for nationwide emissions for the multi-event case showed the horizontal distribution with higher concentrations in the central SMA while the highest one was recorded in the western coastal area.The highest concentration might result from the high density of point sources such as power plants, industrial complexes, and steel production facilities near the area.High concentrations in central SMA might be due to the high density of line sources, and somewhat higher PM 10 concentrations in the northeast might be due to cement manufacturing facilities.For the non-event case, industrial complexes in the southwest were a major contribution to the high PM 10 concentration in the SMA due to the prevailed westerly wind (Park et al., 2019).
In winter, the PM 10 concentration for the multi-event days was higher in the southern area due to the prevailing northwesterly wind.Due to strong winds in winter, the overall concentration field showed a homogeneous horizontal distribution pattern except for the high PM 10 concentrations that peaked in the western coastal area and the SMA.It should be noted that point sources including industrial combustion, electric utility, incineration, cement manufacturing, and residential heating facilities in the satellite cities located to the northeast and east of the SMA were important local sources that account for approximately 50 μg m −3 in the SMA.For the non-event case, point sources in the northeast were major contributions to the high PM 10 concentration in the SMA due to the prevailing northeast wind with the stronger wind (Park et al., 2019).
In autumn, the multi-event case was influenced by an Asian dust storm, and the overall concentration field showed a more homogeneous horizontal distribution except for the high PM 10 concentrations in the western coastal area.For the non-event case, the overall concentration field was similar to the horizontal distribution for the event days except for the high PM 10 concentrations in the SMA.

3 Effects of SMA Emissions on the Air
Pollutants Concentrations in the SMA To quantify the effects of SMA emission on the air pollutants concentrations in the SMA for the multi-event and non-event cases, the TAPM model simulated the meteorological variables and air quality concentration with the same initial meteorological field, concentration, and two different modes of emissions for each case (Table 1 and Fig. 2).One mode refers to the nationwide emissions, and the other to the non-SMA emissions (nationwide-SMA).This approach is the same as the Brute-Force Method (BFM) by which we zero-out the emissions in the targeted area (Park et al., 2019;Koo et al., 2008).In order to figure out the effect of local emission on the SMA concentration, the air pollutant concentrations averaged over 4 stations in SMA (Seoul, Incheon, Suwon, and Uijeongbu) were used (Fig. 1).Table 5 shows the mean concentrations in the SMA simulated by nationwide and non-SMA emissions of PM 10 , PM 2.5 , NO 2 , and SO 2 during the multi-event and non-event cases.The difference between mean concentrations and the fractional contribution for each case are listed in Table 5.The fractional contribution was simply calculated by the ratio (%) of concentration simulated with the use of the SMA emission to concentration simulated with the use of the nationwide emission.
For the multi-event cases, the fractional contribution of SMA emissions to the mean PM 10 concentration was as low as 25% in spring, while as high as 32% in winter.The fractional contribution to the mean PM 2.5 concentration was similar to or lower than the one to the PM 10 concentrations.For the non-event cases, the fractional contribution to the mean PM 10 concentration was similar to, larger than, and lower than those for the multievent cases in spring, autumn, and winter, respectively.The fractional contribution to the mean NO 2 concentration was as high as 46%-51% for both cases in spring, while as low as 32%-36% for the multi-event case in autumn and the non-event case in winter.High fractional contribution to the NO 2 concentration might be related to the domestic dominant cases rather than long-range transported cases.The fractional contribution to the mean SO 2 concentration was 49%-53%, except for the event case (61%) in winter.These fractional contributions were similar to those in the previous study (Park et al., 2019).
PM 10 and PM 2.5 mean concentrations simulated with the use of nationwide and SMA emission for event cases were higher than those for non-event cases, except for autumn.The fractional contributions to the PM 10 and PM 2.5 concentrations for event cases were similar to those for non-event cases in all seasons.The concentrations of air pollutants in the SMA were strongly influenced by the outside area of the SMA as well as inside.

SUMMARY AND DISCUSSION
The impacts of synoptic meteorology and nationwide/local emissions in the SMA during the high PM 10 multi-event and non-event days were com pared.The multi-event cases were selected amo ng the consecutive high PM 10 event days with a daily mean concentration ≥100 μg m −3 in spring, aut u mn, and winter.The nonevent cases were selected among the consecutive days with a daily mean PM 10 concentration <50 μg m −3 in the same month as the event cases.The cases of 5 consecu- tive high PM 10 event days in spring and winter, and 2 consecutive days in autumn were chosen as the multi-event cases.The high PM concentration of the latter was mainly due to the Asian dust (natural mineral dust) and not the anthropogenic smog and haze.The synoptic meteorological conditions played an important role in high concentrations.For the multievent cases in spring and winter, high PM 10 concentrations were accompanied by weak migratory anticyclone systems over the Korean Peninsula.The horizontal range of surface pressure near the Korean Peninsula was very small, implying that the synoptic system near the SMA was weak and stagnant.On the contrary, synoptic winds for the non-event cases were not so weak as those in the multi-event cases.
The horizontal distribution of meteorological variables and concentration of air pollutants were simulated using the TAPM.The horizontal distributions of concentration were strongly governed by the confluence and divergence of wind as well as upwind or downwind of the major emission sources.To quantify the effects of SMA emissions on the air pollutants' concentration in the SMA, TAPM simulated the same initial background concentration and meteorological boundary condition, and two different scales of emissions (nationwide emissions and non-SMA emissions) for the multi-event and non-event cases.The fractional contribution of SMA emissions to the mean PM 10 concentrations was as low as 25% in spring, while as high as 33% in autumn and winter for the multi-event cases.The contribution to the mean PM 2.5 concentration was similar to or lower than those to the mean PM 10 concentration.Moreover, the contribution to the mean NO 2 (SO 2 ) concentration was much higher than the one to the mean PM concentration, except for in autumn.These contributions for the multi-event cases were much higher than those for the single-event cases in the previous studies.As the high PM concentration period becomes longer, the local emissions contribute more to the formation of high concentrations.
Because the SMA in Korea is surrounded by mountains on three sides, except the west where the Yellow Sea is located, the air quality is influenced by the local circulation such as sea-land breeze and mountain-valley breeze (Park and Chae, 2018;Park, 2018;Oh et al., 2015).This implies that, apart from synoptic meteorology, the vertical structure of meteorological variables in the atmospheric boundary layer can also be an important factor that enhances the concentration of air pollutants.A more detailed analysis of the local circulation and vertical transport is needed to interpret the mechanism of high PM concentrations more accurately, especially in the SMA.

Fig. 1 .
Fig. 1.The innermost model domain and location of air quality monitoring (Seoul Station, cross), and surface meteorological observation (Songwol Station, diamond) stations, Suwon, Incheon and Uijeongbu cities (red circle) used in the study area.

Fig. 2 .
Fig.2.Total emissions per 4 × 4 km 2 grid of (a) PM 10 , (b) NO x , and (c) SO 2 ; area source emissions of (d) PM 10 , (e) NO x , and (f) SO 2 ; line source emissions of (g) PM 10 , (h) NO x , and (i) SO 2 ; point source emissions of (j) PM 10 , (k) NO x , and (l) SO 2 nationwide.Black dotted-lines in the figures (a)-(c) denote the Seoul Metropolitan Area.

Fig. 3 .
Fig.3.Time series of PM 10 , PM 2.5 , NO 2 , and SO 2 concentrations in the SMA for the multi-event and non-event cases.The PM 2.5 data for winter cases in 2014 were not available.Major ticks for x-axis were drawn at 1200 LST on the day.

Fig. 7 .
Fig. 7. Location of high and low pressure center with its sea level pressure and moving paths (blue dashed line: high pressure, red dashed line: low pressure) near Korean Peninsula during the event periods in (a) spring, (b) autumn, and (c) winter.Cold fronts are indicated by blue dash dot lines in (b).

Fig. 8 .
Fig. 8. Time series of temperature (first row), relative humidity (second row), rain (third row), wind speed (fourth row), and wind direction (fifth row) during the event and non-event periods in spring (first, fourth columns), autumn (second, fifth columns), and winter (third, sixth columns), respectively.

Fig. 10 .
Fig. 10.Horizontal distribution of PM 10 concentration with wind vector considering nationwide (first and third column) and without SMA (second and fourth column) emissions on event (first and second column) and non-event (third and fourth column) periods in spring (first row), autumn (second row), and winter (third row), respectively.The center of Seoul is indicated by a red circle.

Table 1 .
The median concentration of observed PM 10 , PM 2.5 , NO 2 , and SO 2 in the SMA for the multi-event and non-event cases.

Table 2 .
The ratio of mean concentration to the monthly mean concentration of PM 10 , PM 2.5 , NO 2 , and SO 2 in the SMA for the multievent and non-event cases.

Table 3 .
Initial background concentrations for TAPM simulation.

Table 5 .
Mean concentration in the SMA simulated by applying the nationwide and non-SMA emissions for the multi-event and nonevent cases.Fraction indicates the ratio of the SMA contribution to the nationwide emission in %.