Regional Characteristics of NO2 Column Densities from Pandora Observations during the MAPS-Seoul Campaign

Vertical column density (VCD) of nitrogen dioxide (NO2) was measured using Pandora spectrometers at six sites on the Korean Peninsula during the Megacity Air Pollution Studies-Seoul (MAPS-Seoul) campaign from May to June 2015. To estimate the tropospheric NO2 VCD, the stratospheric NO2 VCD from the Ozone Monitoring Instrument (OMI) was subtracted from the total NO2 VCD from Pandora. European Centre for Medium-Range Weather Forecasts (ECMWF) reanalysis wind data was used to analyze variations in tropospheric NO2 VCD caused by wind patterns at each site. The Yonsei/SEO site was found to have the largest tropospheric NO2 VCD (1.49 DU on average) from a statistical analysis of hourly tropospheric NO2 VCD measurements. At rural sites, remarkably low NO2 VCDs were observed. However, a wind field analysis showed that trans-boundary transport and emissions from domestic sources lead to an increase in tropospheric NO2 VCD at NIER/BYI and KMA/AMY, respectively. At urban sites, high NO2 VCD values were observed under conditions of low wind speed, which were influenced by local urban emissions. Tropospheric NO2 VCD at HUFS/Yongin increases under conditions of significant transport from urban area of Seoul according to a correlation analysis that considers the transport time lag. Significant diurnal variations were found at urban sites during the MAPS-Seoul campaign, but not at rural sites, indicating that it is associated with diurnal patterns of NO2 emissions from dense traffic.


INTRODUCTION
Nitrogen dioxide (NO 2 ) is an important chemical species in both tropospheric and stratospheric chemistry (e.g., Crutzen, 1979;Brasseur et al., 1998).NO 2 mainly affects air quality in the troposphere through its role in chemical processes with ozone (O 3 ) and other trace gases (e.g., Seinfeld, 1988;Solomon et al., 1999;IPCC, 2007;Choi et al., 2008Choi et al., , 2009)).In addition, long-term exposure to high concentrations of NO 2 causes respiratory and cardiovascular diseases (e.g., Chitano et al., 1995;Bayram et al., 2001;von Klot et al., 2005).Emission sources of NO 2 differ between the troposphere and stratosphere (Lee et al., 1997;Barton and Atwater, 2002;Galloway et al., 2004).In the stratosphere, the main source of NO 2 is oxidation of nitrous oxide (N 2 O) and other sources include lightning and biomass burning (e.g., Liley et al., 2000;Barthe et al., 2007;Allen et al., 2010;Bucsela et al., 2013).In the troposphere, anthropogenic activities, including fossil fuel combustion and vehicle emissions, are the dominant sources of NO 2 , along with some contributions from soil emissions and lightning (Zhang et al., 2003;Hudman et al., 2007;Choi et al., 2008Choi et al., , 2009)).Because of the diversity of nitrogen oxides chemical reactions and emission sources, spatial distributions of NO 2 typically show regional differences.For these reasons, accurate observations of the spatial distribution of NO 2 are important.
Several measurement platforms for tropospheric gases, including NO 2 , have been developed in recent decades.In particular, tropospheric column NO 2 is widely monitored using ground-and satellite-based platforms that make use of the spectral absorption features of NO 2 .For example, the Brewer spectrophotometer can estimate total column NO 2 using direct solar radiation at several specific wavelengths (e.g., Brewer et al., 1973;Kerr et al., 1988;Kerr, 1989;Cede et al., 2006;Diemoz et al., 2014).In addition, several instruments have been used to measure NO 2 column densities using differential optical absorption spectroscopy (DOAS) methods (e.g., Boersma et al., 2004;Castellanos et al., 2015;Chimot et al., 2016) applied to sunlight scattered along the zenith (Van Roozendael et al., 1997;Liley et al., 2000) or scattered in multiple directions (e.g., Sinreich et al., 2005;Irie et al., 2008;Lee et al., 2009a, b).
Recently, the Pandora spectrometer was developed by Goddard Space Flight Center (GSFC) at the National Aeronautics and Space Administration (NASA) for the estimation of trace gases using direct sunlight (Herman et al., 2009;Cede, 2011).The Pandora system is an array detector spectrometer with a temperature control system.A solar tracking system is also included to allow constant direct-sunlight observations.Using this optical system, the Pandora instrument can continuously observe direct radiance with high temporal resolution.This instrument has been used to obtain the total column amount of trace gases, including O 3 and NO 2 (Herman et al., 2009;Tzortziou et al., 2012).
In South Korea, the first two Pandora instruments were installed in 2012 at Yonsei University in Seoul and Pusan National University in Busan for trace gas observations during the Distributed Regional Aerosol Gridded Observation Networks-North East Asia (DRAGON NE-Asia) campaign.Since the initial installations, several instruments were installed additionally in different regions in Korea.The main purpose of Pandora network in Korea was to conduct observations for the Megacity Air Pollution Studies-Seoul (MAPS-Seoul; May to June, 2015) and Korea-United States Air Quality Study (KORUS-AQ; May to June, 2016), field campaigns in Korea aimed to monitor and understand air quality (https://espo.nasa.gov/korus-aq/content/KORUS-AQ_Science_Overview_0).It also supports studies associated with the planned satellite mission for the Geostationary Environmental Monitoring Spectrometer (GEMS; Kim, 2012), designed to study atmospheric chemical composition and reaction mechanisms over East Asia (Kim et al., 2017).Measurements from Pandora instruments in South Korea have been utilized for multi-year analyses and the validation of total O 3 measurements (Baek et al., 2017;Kim et al., 2017).Although the data quality of total O 3 from the Pandora spectrometers differs by instrument, the difference in total O 3 between Pandora and other instruments has been less than 2% in all comparison studies to date (Baek et al., 2017;Kim et al., 2017).
Because variations in total O 3 can primarily be attributed to changes in the stratosphere, the pattern of total O 3 density shows a weak dependence on regional or local emissions near the observation sites.In contrast, NO 2 column is affected primarily by vehicle emissions, industrial activities, and local short-range transport in the troposphere.Although some previous work analyzed the effect of the long-range transport of NO 2 (e.g., Donnelly et al., 2015), the lifetime of NO 2 near the surface is on the order of hours to days (Seinfeld and Pandis, 2012), indicating that a careful approach is required for analyses of the effects of regionalscale transport on variations in NO 2 concentrations.
Thus, the objective of this study is to investigate the effects of local air-mass transport from NO 2 sources on vertical column densities (VCD) of tropospheric NO 2 from Pandora observations.As a preliminary study for the KORUS-AQ campaign, we analyzed the regional characteristics of NO 2 measurements at six Pandora sites obtained during the MAPS-Seoul campaign while considering wind patterns and the locations of regional emission sources.Using these findings, we report the characteristics of NO 2 VCD during the MAPS-Seoul campaign from May 18 to June 14, 2015 in South Korea.

DATA AND ANALYSIS
Variations in tropospheric NO 2 VCD were analyzed at six Pandora sites.Also, the characteristics of NO 2 transport during the MAPS-Seoul campaign were investigated using the wind information from European Centre for Medium-Range Weather Forecasts (ECMWF) reanalysis data.To evaluate the characteristics of transport in the troposphere at the Pandora observation sites, we obtained tropospheric NO 2 VCD by subtracting Ozone Monitoring Instrument (OMI) stratospheric NO 2 VCD from Pandora total NO 2 VCD.

NO 2 from Pandora
The Pandora spectrometer system is configured with mounting systems for sun-tracker and sky-scanner controls, and a temperature stabilizing system for radiometric calibration (e.g., Herman et al., 2009;Tzortziou et al., 2012Tzortziou et al., , 2013)).The Pandora instruments in Korea are direct-sun (DS) spectrometer systems with a spectral range of 280-525 nm (spectral resolution: 0.42-0.52nm for Full-Width at Half Maximum), and are used to measure the absorption spectra of trace gases (Baek et al., 2017).The first two Pandora spectrometers in Korea were respectively installed at Yonsei University and Pusan National University in March 2012 to measure total column NO 2 and O 3 .These instruments participated in the DRAGON-NE Asia campaign in South Korea at that time.Since the initial installations, six Pandora observation sites came into operation over the Korean Peninsula as part of the MAPS-Seoul campaign.Table 1 summarizes the locations and altitudes of the Pandora observation sites operated during the campaign.Fig. 1 shows their locations with a ten-year (2005-2014) mean tropospheric NO 2 VCD in Dobson Units (DU; 1 DU = 2.69 × 10 16 molecules cm -2 ) from Level 3 OMI NO 2 product (OMNO2d) in May and June, and mean surface wind reanalysis data from ECMWF for the same period.
Total NO 2 VCD from Pandora was retrieved using the spectral fitting method described by Herman et al. (2009).For the analysis, we used Pandora level-3 data, which includes the normalized root-mean square error (normalized RMS), uncertainties in the NO 2 column, and the index of the filter wheel position.Threshold values for quality control that we used were 0.05 and 0.05 DU for normalized RMS and uncertainties in the NO 2 column, respectively.Pandora provides total NO 2 VCD every 80 seconds (e.g., Tzortziou et al., 2012;Yun et al., 2013).We converted the Pandora total NO 2 VCDs to hourly values by averaging over a ± 30-min window every hour.In this procedure, the resultant hourly data which have less than three points in their windows (~5% of total measurable points in an hour) were excluded.Total VCD includes both stratospheric and tropospheric NO 2 .However, tropospheric NO 2 VCD usually dominates total NO 2 VCD variations near urban and industrial areas.To accurately estimate tropospheric NO 2 VCD from Pandora observations, stratospheric NO 2 information is required.For this information, we used the OMI standard product, which is explained in detail in the following subsection.Thus, tropospheric NO 2 VCD was estimated for the MAPS-Seoul campaign at each observation site, and the number of data points at each site is listed in Table 2.Because of problems with the tracker cable on the Pandora instrument at the Yonsei/SEO site beginning May 29, the number of valid data points at this site was less than half the number of data points at the other five Pandora observation sites.
The six Pandora sites were selected to include both urban and rural areas on the Korean Peninsula.

Stratospheric NO 2 from OMI
The OMI instrument was designed to observe total column amounts of several trace gases including NO 2 , and thus aid analyses of air quality and assessments of the climate effects of changes in chemical composition.OMI was launched in 2004 onboard the Aura satellite and measures backscattered solar light in the ultra violet (UV)visible range (270-500 nm) using three different channels (two UV and one visible; e.g., Levelt et al., 2006;Buchard et al., 2008).The nominal spatial resolution is from 13 × 24 km 2 (at nadir) to 28 × 150 km 2 (at the edge of the swath).Because the Aura is a sun-synchronous satellite with a 98° inclination angle, daily values over the Korean Peninsula from OMI were obtained from data from one or two overpasses per day in the afternoon.The nadir pass was obtained close to 13:30 local time.
The NO 2 VCD from OMI is based on the level-2 standard product of NASA's version 3 OMI NO 2 (OMNO2) dataset.For the retrieval of NO 2 , spectral radiance data from 405 to 465 nm in the visible spectrum were used to estimate slant column amounts (Bucsela et al., 2013).As part of improvements to the retrieval algorithm between versions 2 and 3, chemical transport model simulations with timedependent emissions were adopted to account for variations in the vertical profiles for air mass factor calculations (Ialongo et al., 2016;Krotkov et al., 2017).The OMNO2 provides total, tropospheric, and stratospheric NO 2 VCDs.Thus, OMI stratospheric NO 2 data can be used to estimate the tropospheric NO 2 VCD from Pandora observations.Stratospheric NO 2 is affected by chemical reactions and lightning activity (Noxon, 1979;Zhang et al., 2000;Wenig et al., 2004) and thus varies spatiotemporally.We sampled the stratospheric NO 2 VCD from OMI for each Pandora station, and thus the tropospheric NO 2 VCD from Pandora data was calculated as where VCD PAN,Trop , VCD PAN,Tot , and VCD OMI,Strat are the tropospheric and total NO 2 VCD from Pandora, and the stratospheric NO 2 VCD from OMI, respectively.For spatial co-location with Pandora, we selected the OMI pixels within 20 km from each Pandora site.While the Pandora total NO 2 VCD was calculated for each hour, OMI values occur only once daily, in the afternoon.This is one potential error source for using OMI to estimate tropospheric NO 2 VCD from the Pandora observations.In this study, we assumed the temporal and spatial distribution of stratospheric NO 2 was stable throughout the day and over the 20 km window.This approach is adapted from previous studies (e.g., Knepp et al., 2015;Kollonige et al., 2017) which estimated tropospheric NO 2 VCD by subtracting OMI stratospheric NO 2 VCD from Pandora total column NO 2 observations, assuming constant stratospheric NO 2 column on a daily basis under polluted conditions.In less polluted regions, such as NIER/BYI in this study, negative tropospheric NO 2 values can occur occasionally with this method.In addition, during the stratosphere-troposphere separation procedure for OMI NO 2 standard product, the stratospheric NO 2 VCD is smoothed interpreting features smaller than 300 km in scale in the initial estimate as tropospheric (Bucsela et al., 2013).Thus, assuming negligible spatial variability in the 20 km window, which is much smaller, is reasonable.Fig. 3 shows the total NO 2 VCD observed from Pandora and sampled stratospheric NO 2 VCD from OMI for each site during the MAPS-Seoul campaign.Tropospheric NO 2 dominates total NO 2 VCD in the SMA (at Yonsei/SEO and HUFS/Yongin), and PNU/Busan, as can be inferred from Fig. 1.Despite showing lower levels of total NO 2 compared to those sites, the amounts and variability of NO 2 in the stratosphere at NIER/BYI, KMA/AMY, and GIST/Gwangju are relatively small in comparison to total, allowing stratospheric NO 2 VCD to be assumed constant on a daily basis.However, there were days for which no data from OMI were available within 20 km from a ground-based observation site.In this case, the averaged OMI stratospheric NO 2 VCD for that particular site over the course of the campaign period was used (red dots in Fig. 3(b)).

Wind Data
To analyze the transport characteristics of NO 2 at the observation sites, we used the wind speed and wind direction at the surface.Because surface wind varies significantly with surface conditions over small spatial scales, high-resolution wind information is required.For the wind analysis, U and V (west-east and south-north, respectively) wind components at 10 m altitude from ECMWF reanalysis interim data (hereafter ERA-Interim) were used in this study (Dee et al., 2011; http://apps.ecmwf.int/datasets/).To decrease local surface-condition effects on wind, the ECMWF ERA-Interim were used at the highest available horizontal resolution, 0.125° × 0.125°.
The U and V wind components from ERA-Interim were interpolated to the latitude and longitude of the respective Pandora sites.Then we converted them to the wind speed and direction at each site.The original reanalysis data have a temporal resolution of 6 hours.Thus, wind information was interpolated to a 1-hour resolution from the original 6hour resolution after the spatial interpolation, for temporal co-location with the NO 2 data.temporarily increased on May 19 and 21, but otherwise remained small and constant.Negative values were found at this site, which were generated by subtraction of OMI stratospheric NO 2 VCD larger than Pandora total NO 2 VCD.However, the number of them was less than 2% of the total data points.In a diurnal scale, variations in tropospheric NO 2 were clearly observed at the Yonsei/SEO, HUFS/Yongin, and PNU/Busan sites, which are the sites nearest to large urban areas.However, the diurnal cycle in tropospheric NO 2 VCD at GIST/Gwangju was not as strong.It was similar to what was observed at KMA/AMY and NIER/BYI, although the GIST/Gwangju site is located near an urban area (see also Fig. 9).

RESULTS AND DISCUSSION
Table 2 and Fig. 5 present details of the hourly tropospheric NO 2 VCDs during the MAPS-Seoul campaign.The Yonsei/SEO site showed the maximum mean tropospheric NO 2 VCD (1.49 ± 0.68 DU) among the six sites during the campaign period.Even the minimum tropospheric NO 2 VCD at this site was 0.60 DU, which is larger than the maximum value of tropospheric NO 2 VCD at the NIER/BYI site (0.45 DU), representing the huge NO 2 emissions in Seoul.In addition, the HUFS/Yongin and PNU/Busan sites had similar tropospheric NO 2 VCD levels.The HUFS/Yongin site had the third largest average tropospheric NO 2 VCD (0.67 ± 0.49 DU), following PNU/Busan (0.72 ± 0.39 DU), indicating that huge local emissions in Seoul also lead to high NO 2 concentrations around the SMA.The NO 2 VCD in the troposphere at GIST/Gwangju (0.26 ± 0.14 DU) was comparable to that in KMA/AMY (0.18 ± 0.14 DU).
Although the GIST/Gwangju site is located in an urban area, levels of tropospheric NO 2 VCD resembled those at the suburban and rural sites, because the industrial region around Gwangju is small.As can be seen from the boxand-whisker plots in Fig. 5, all the sites are skewed in the positive direction.From the results so far, the statistics of tropospheric NO 2 were identified for each of the Pandora sites during the MAPS-Seoul campaign.However, even if the distribution of tropospheric NO 2 VCDs is similar among sites, major contributors to the amounts may differ depending on local emissions and transport patterns.
For better understanding of major contributors to tropospheric NO 2 VCD at each site, we need to examine the NO 2 VCD change in accordance with the wind pattern.Fig. 6 shows polar plots of tropospheric NO 2 VCD, along with wind speed and direction at the six Pandora sites.Because of the regional characteristics of wind and the short campaign period, it is likely that not all wind patterns in South Korea are included in the analysis (e.g., seasonal wind pattern).However, this analysis indicates how wind pattern affects the regional characteristics of tropospheric NO 2 during the MAPS-Seoul campaign.In most cases, the wind speed did not exceed 10 m s -1 at the observation sites, and westerly wind was dominant over easterly wind.This is consistent with the general wind pattern over the Korean peninsula in May and June (see Fig. 1).Despite the limited number of observed wind patterns, Fig. 6    high tropospheric NO 2 VCDs at each site were found for specific wind fields.However, these wind field characteristics differ among the observation sites.At two urban sites, Yonsei/SEO (Fig. 6(b)) and PNU/Busan (Fig. 6(f)), high tropospheric NO 2 VCD values were found under conditions of southerly wind with low wind speed (< 4 m s -1 ).The GIST/Gwangju (Fig. 6(e)) site also showed high VCD values under these wind conditions, but the change in NO 2 VCD was smaller and may not be significant.This finding indicates that local NO 2 emissions have huge contribution to the enhancement of tropospheric NO 2 at Yonsei/SEO and PNU/Busan during the campaign period.
Specific wind fields at the NIER/BYI (Fig. 6(a)) and KMA/AMY (Fig. 6(d)) sites were also associated with high NO 2 VCD: westerly wind with speeds of 4-6 m s -1 for NIER/BYI, and easterly wind with speeds of 2-4 m s -1 for KMA/AMY.These two observation sites are in rural areas with weak local emissions.Therefore, high VCD values can be attributed to transport.However, the NO 2 transport patterns differ between the two sites.As seen in Fig. 2, NIER/BYI borders the Yellow sea on the west, while the regions to the east of the KMA/AMY site are land surfaces.Therefore, high concentrations of NO 2 at the NIER/BYI site are thought to be due to trans-boundary transport (e.g., Lee et al., 2014), while those at the KMA/AMY site are attributed to the transport of domestic (Korean) emissions near the observation sites.Large domestic sources of NO 2 around the KMA/AMY site include power plants and industrial activity.The KMA/AMY site is close to many power plants that provide power to the SMA, as well as several chemical plants.For this reason, easterly wind patterns change the characteristics of the KMA/AMY site from a classical rural site to those resembling an industrial site.
At the HUFS/Yongin site, high tropospheric NO 2 VCD values were found under northwesterly wind conditions with wind speeds of 2-6 m s -1 , as shown in Fig. 6(c).The HUFS/Yongin site is located on the southeast side of Seoul, making it downwind of Seoul during westerly wind conditions.Under these conditions, the air quality of the surrounding area is dominated by emissions from Seoul.Fig. 8 shows the correlation of tropospheric NO 2 VCDs between the HUFS/Yongin and Yonsei/SEO sites with consideration of the time difference.Because the distance between the Yonsei/SEO and HUFS/Yongin sites is about 40 km, the transport of NO 2 from Yonsei/SEO to HUFS/Yongin can be assessed using a time lag analysis between data from the two observation sites.For this assessment, tropospheric NO 2 VCD data were selected only for northwesterly wind conditions (a wind direction of 270°-360°) at HUFS/Yongin, based on the relative location of the HUFS/Yongin site from the Yonsei/SEO site.A 3-hour  time lag was used on the hourly HUFS/Yongin data for the lag correlation analysis shown in Fig. 8.In Fig. 8(a), no correlation was observed before adjusting for the time lag.Therefore, the tropospheric NO 2 VCDs observed at the same time at the two sites were not significantly related.After adjusting for the 3-hour time lag, however, a significant positive correlation was found with a correlation coefficient (r) of 0.44 (Fig. 8(b)).The time lag analysis indicated that tropospheric NO 2 in Seoul affects that at the HUFS/Yongin site within a few hours.As shown in Fig. 6(c), high tropospheric NO 2 VCDs at the HUFS/Yongin site were found when the wind was northwesterly at 2-6 m s -1 .Based on the distance between the two observation sites (~40 km) and the wind speed when tropospheric NO 2 concentrations were high at the HUFS/Yongin site, an air mass in Seoul would take 2-5 hours to flow to the HUFS/Yongin site.
Fig. 9 shows the average and standard deviation of hourly tropospheric NO 2 VCD during the MAPS-Seoul campaign.Because the Pandora instrument uses solar radiation, the temporal observation range is from 7:00 to 18:00, local time.In Fig. 9(a), a diurnal pattern of tropospheric NO 2 VCD is clearly seen for Yonsei/SEO.At Yonsei/SEO, NO 2 concentrations were high from 9:00 to 14:00.After high concentrations during the day, NO 2 VCD decreased in the late afternoon.At the HUFS/Yongin site, however, the diurnal pattern of tropospheric NO 2 VCD was different from that in Yonsei/SEO.The tropospheric NO 2 VCD in the afternoon was higher than in the morning.Because of the lower transportation density around the HUFS/Yongin site, no effects on NO 2 VCD due to transportation during the morning rush hour were observed.The significant increase in NO 2 VCD during the afternoon may have been caused by NO 2 advection from Seoul, because the temporal change in tropospheric NO 2 VCD at HUFS/Yongin was highest between 12:00 and 13:00, 3 hours after VCD peaked in Yonsei/SEO.At PNU/Busan, NO 2 was high around noon, but was small compared with that observed at Yonsei/SEO, despite being a populous region.For the three regions in Fig. 9(a), the diurnal variations in tropospheric NO 2 VCD can be readily explained.However, an explanation of the diurnal variations in the remaining three regions (Fig. 9(b)) is more challenging because these three sites are strongly affected by wind field changes related to emission sources in the vicinity (KMA/AMY and GIST/Gwangju) and transboundary transport (NIER/BYI).At NIER/BYI, GIST/Gwangju, and KMA/AMY, we detected small diurnal variability in tropospheric NO 2 VCD during the MAPS-Seoul campaign.During the MAPS-Seoul campaign period, large differences in tropospheric NO 2 VCD were found between urban and rural sites.Two urban sites, Yonsei/SEO and PNU/Busan, had large maximum tropospheric NO 2 VCD values of 4.70 DU and 2.04 DU, respectively.However, tropospheric NO 2 VCD at the NIER/BYI site was always below 0.45 DU because of its large distance from emission sources.At the sites within the SMA (Yonsei/SEO and HUFS/Yongin), mean values of tropospheric NO 2 VCD were significantly larger than those at the other sites except for PNU/Busan.This indicates that regional emissions near Seoul significantly contribute to high NO 2 concentrations in the SMA.The tropospheric NO 2 VCD at GIST/Gwangju was comparable to that at KMA/AMY, even though the GIST/Gwangju site is located near an urban area.
To understand major contributors to tropospheric NO 2 VCD at each of the observation sites, the wind field (wind speed and direction) from reanalysis data at each site was analyzed in comparison with the change in NO 2 VCD.Using this method, high tropospheric NO 2 VCDs at each site were found to be related with specific wind speeds and directions.At Yonsei/SEO and PNU/Busan, high tropospheric NO 2 VCD values were directly attributed to large local emissions.At the rural sites, the increase of NO 2 VCDs result primarily from trans-boundary transport (NIER/BYI) or transport from domestic emissions near the observation site (KMA/AMY).Because the location of HUFS/Yongin site is usually downwind of Seoul, high tropospheric NO 2 VCDs were found under conditions in which the wind blows directly from Seoul (i.e., strong northwesterly wind).In addition, tropospheric NO 2 VCD between the HUFS/Yongin and Yonsei/SEO sites was significantly correlated when a 3-hour time lag was included.This time lag is similar to the estimated transport time between the two sites, based on the distance between the two sites and the wind speed at which high tropospheric NO 2 VCD occurred.Therefore, it seems clear that the high tropospheric NO 2 VCD at HUFS/Yongin site can be directly attributed to emissions in Seoul.
Diurnal patterns in tropospheric NO 2 VCD are evident in Yonsei/SEO and HUFS/Yongin from daytime Pandora measurements, but no significant patterns were observed for NIER/BYI, KMA/AMY, and GIST/Gwangju.Although the Yonsei/SEO and HUFS/Yongin sites show diurnal variations in tropospheric NO 2 VCD, the nature of the variations differs between the two sites.At Yonsei/SEO, high NO 2 VCDs were found from 9:00 to 14:00.The HUFS/Yongin site did not show peaks in the morning, but high NO 2 VCDs were observed throughout the afternoon because of NO 2 transport from Seoul.
Tropospheric NO 2 VCD was analyzed at each site using Pandora observations.Throughout the campaign period, the SMA had higher NO 2 VCD than other areas in general.The lowest NO 2 value was observed at NIER/BYI.The relationship between NO 2 VCD and wind field was analyzed to investigate the effects of regional emissions and transport on the observation sites.The HUFS/Yongin site was significantly affected by the transport of emissions from Seoul.Although several notable results were obtained during this work, the analysis was limited to the period of the MAPS-Seoul campaign.Therefore, further analyses of variations in tropospheric NO 2 using data that span several years are required in the future.

Fig. 1 .
Fig. 1.Locations of the Pandora sites on the Korean Peninsula with a ten-year (2005-2014) mean tropospheric NO 2 VCD from OMI in May and June, and mean surface wind reanalysis data from ECMWF for the same period.
Fig. 2 shows the topography (in color) of the regions around the observation sites.The Baengnyeongdo site operated by National Institute of Environmental Research (NIER) (NIER/BYI; Fig. 2(a)) and the Anmyeondo site operated by Korean Meteorological Agency (KMA) (KMA/AMY; Fig. 2(d)) are located in coastal areas far from industrial and urban regions.The NIER/BYI site is on an island located in northwest South Korea and is the closest site to China.KMA/AMY is also located in the western coastal region of Korea.The NIER/BYI and KMA/AMY sites are relatively free from the urban pollution and therefore were used as background sites in this study.The Busan (operated by Pusan National University) (PNU/Busan; Fig. 2(f)) site and the Gwangju (operated by Gwangju Institute of Science and Technology; GIST) (GIST/Gwangju; Fig. 2(e)) site are located on the northern part of the urban areas in Busan and Gwangju, respectively.These two cities are respectively the 2 nd and 6 th most populous cities in South Korea.The Seoul (operated by Yonsei University) (Yonsei/SEO; Fig. 2(b)) site and the Yongin (operated by Hankuk University of Foreign Studies; HUFS) (HUFS/Yongin; Fig. 2(c)) site are located in the Seoul metropolitan area (SMA); however, these two sites have different characteristics.The Yonsei/SEO site is at the center of the Seoul urban area, while the HUFS/Yongin site is in a suburban area 40 km southeast and downwind of Seoul.Therefore, the HUFS/Yongin site is expected to be affected by transport from the urban area of Seoul.

Fig. 2 .
Fig. 2. Local characteristics of the Pandora sites at (a) NIER/BYI, (b) Yonsei/SEO, (c) HUFS/Yongin, (d) KMA/AMY, (e) GIST/Gwangju, and (f) PNU/Busan.The radius of each red circle is 0.125°, representing the resolution of the wind field we employed, and the center of it indicates the location of each Pandora (red dot).The location of Pandora at Yonsei/SEO is marked in green on (c) (Maps created using Google Earth Imagery).

Fig. 4
Fig. 4 shows a time series of hourly-based tropospheric NO 2 VCD from the Pandora observations.During the MAPS-Seoul campaign period, tropospheric NO 2 VCDs at NIER/BYI and KMA/AMY were remarkably low, ranging from -0.07 to 0.45 DU and from 0.03 to 0.95 DU, respectively.However, large NO 2 VCDs were observed, up to 4.70 DU at Yonsei/SEO and 2.04 DU at PNU/Busan.While the difference between the maximum and minimum during the whole campaign period was less than 1 DU at NIER/BYI and KMA/AMY, a variation of 3.96 DU was observed in a single day on May 28 at Yonsei/SEO.Tropospheric NO 2 VCD was always below 0.45 DU at the NIER/BYI site because of the large distance from emission sources.At the NIER/BYI site, the tropospheric NO 2 VCD
clearly shows a change in tropospheric NO 2 VCD as the wind field changes.At all the Pandora observation sites, remarkably
Fig. 6(c) indicates that transport of pollutants from Seoul affected tropospheric NO 2 measurements at the HUFS/Yongin site during the MAPS-Seoul campaign.To further assess transport from Seoul to the HUFS/Yongin site, Fig. 7 shows a time series of tropospheric NO 2 VCD along with wind field data at the HUFS/Yongin site during the MAPS-Seoul campaign.Enhanced tropospheric NO 2 VCDs and high wind speed were highly connected under westerly wind conditions.Otherwise, low tropospheric NO 2 values were observed under conditions of weak wind speed or southwesterly winds.Considering the direction of Seoul from the HUFS/Yongin site (Fig. 2(c)), those wind directions causing high tropospheric NO 2 VCD at HUFS/Yongin (~285°-345°) indicate that the HUFS/Yongin site was highly affected by NO 2 transport from Seoul.

Fig. 7 .
Fig. 7. Time series of tropospheric NO 2 VCD, wind speed, and wind direction at the HUFS/Yongin site.

Fig. 8 .
Fig. 8. Scatter plots of tropospheric NO 2 VCDs between the Yonsei/SEO and HUFS/Yongin sites (a) without consideration of time lag, and (b) with a time lag of 3 hours under northwesterly wind conditions at the HUFS/Yongin site.

Table 1 .
Location and altitude of each Pandora site during the MAPS-Seoul campaign.

Table 2 .
The statistics of tropospheric NO 2 VCD at each Pandora site during the MAPS-Seoul campaign.(Unit: DU).