Evaluation of correlated Pandora column NO 2 and in situ surface NO 2 measurements during GMAP campaign

. To validate the Geostationary Environment Monitoring Spectrometer (GEMS), the GEMS Map of Air Pollution (GMAP) campaign was conducted during 2020–2021 by integrating Pandora Asia Network, aircraft, and in situ measurements. In the present study, GMAP-2020 measurements were applied to evaluate urban air quality and explore the synergy of Pandora column (PC) NO 2 measurements and surface in situ (SI) NO 2 measurements for Seosan, South Korea, where large point source (LPS) emissions are densely clustered. Due to the difﬁculty of interpreting the effects of LPS emissions on air quality downwind of Seosan using SI monitoring networks alone, we explored the combined analysis of both PC-NO 2 and SI-NO 2 measurements. Agglomera-tive hierarchical clustering using vertical meteorological variables combined with PC-NO 2 and SI-NO 2 yielded three distinct conditions: synoptic wind-dominant (SD), mixed (MD), and local wind-dominant (LD). These re-sults suggest meteorology-dependent correlations between PC-NO 2 and SI-NO 2 . Overall, yearly daytime mean (11:00–17:00 KST) PC-NO 2 and SI-NO 2 statistical data showed good linear correlations ( R =∼ 0 . 73); however, the differences in correlations were largely attributed to meteorological conditions. SD conditions characterized by higher wind speeds and advected marine boundary layer heights suppressed ﬂuctuations in both PC-NO 2 and SI-NO 2 , driving a uniform vertical NO 2 structure with higher correlations, whereas under LD conditions, LPS plumes were decoupled from the surface or were transported from nearby cities, weakening correlations through anomalous vertical NO 2 gradients. The discrepancies suggest that using either PC-NO 2 or SI-NO 2 observations alone involves a higher possibility of uncertainty under LD conditions or prevailing transport processes. However, under MD conditions, both pollution ventilation due to high surface wind speeds and daytime photo-chemical NO 2 loss contributed to stronger correlations through a decline in both PC-NO 2 and SI-NO 2 towards noon. Thus, Pandora Asia Network observations collected over 13 Asian countries since 2021 can be utilized for detailed investigation of the vertical complexity of air quality, and the conclusions can be also applied when performing GEMS observation interpretation in combination with SI measurements.


Introduction
Rapid developments in environmental remote sensing have led to a new era of air quality observations, and recent hyperspectral data retrieval technologies have allowed for routine and accurate monitoring of air pollutants at high spatial and temporal resolution.In particular, the Geostationary Environment Monitoring Spectrometer (GEMS), which was launched on 18 February 2020, measures the total and tropospheric air pollutant columns hourly at spatial resolutions of 7 km × 8 km for gas and 3.5 km × 8 km for aerosols (Kim et al., 2020), facilitating the tracking of pollution transport from local to synoptic scales.
Recent studies have revealed the potential of satellite observations to evaluate surface air quality, particularly in regions with sparse air quality-monitoring networks.The main approach is to convert column density to surface concentrations using a shape factor of the ratio of the partial column ( z 0 ) within the lowest layer (z 0 ) to the total column ( total ) (Zhao et al., 2019), as follows: where S, C, and z 0 are the surface concentration, column density, and lowest layer thickness, respectively.Acquiring accurate profile shape information is critical for determining the relationship between the column amount and surface concentration because the shape factor is spatiotemporally variable.Considering this, numerous studies have obtained close relationships using chemical transport model simulations, aircraft in situ measurements, and satellite observations with high correlation coefficients (R) of 0.7 or more, and used them to scale up surface NO 2 to column NO 2 (Wang and Christopher, 2003;Boersma et al., 2009;Lamsal et al., 2010).This strong correlation can be explained by the generally uniform planetary boundary layer height (PBLH), and by aerosol type and abundance, which is also the case for trace gases.By contrast, the implications of weak correlations between column and surface measurements remain unclear.Engel-Cox et al. (2004) found a negative correlation of aerosol optical depth (AOD) and surface PM 2.5 in northwestern USA, and explained it based on elevated haze decoupled from the surface.Thompson et al. (2019) examined weak correlations between Pandora column (PC) measurements and surface in situ (SI) observations of NO 2 over the Yellow Sea during the Korea-US air quality (KORUS-AQ) field study, and suggested, as a possible reason, the transported non-uniform plumes originated in China and Seoul hundreds of meters above the ground from the surface layer.The estimated surface PM 2.5 concentration was weakly correlated (R = 0.4-0.49)with observed PM 2.5 concentrations in Seoul because only PBLH was added to the multi-linear regression model to correlate AOD to surface PM 2.5 (Kim et al., 2021).This effect may be related to the significant impact of long-range transport on PM 2.5 , with a contribution of up to 39 % in Seoul (Lee et al., 2021).Thus, the wide variability in the degree of correlation between PC-PM and SI-PM is closely related to vertical profile variability (Flynn et al., 2016).
It appears highly probable that several factors are responsible for the correlations between PC-NO 2 and SI-NO 2 ; therefore, it is necessary to improve our understanding of the degree of correlation through detailed measurements, including column concentration.In this study, we focused on the impact of meteorology and chemistry on correlation variability using PC, SI and aircraft measurements, and meteorological observations.Understanding vertical profile variability is also useful for evaluating the effects of various emissions on urban air quality, particularly in areas neighboring active large point source (LPS) emissions sites.Quantifying the impact of LPS emissions on downwind cities remains challenging due to the lack of three-dimensional (3D) measurements.Accurate vertical profile data are also useful for improving remote sensing retrieval algorithms because the profile shape contributes to the conversion of slant column density into vertical column density as part of the air mass factor.
In mid-2019, the Pandonia Global Network (PGN; https: //pandonia-global-network.org, last access: 22 August 2022) was launched, with support from the National Aeronautics and Space Administration (NASA) and European Space Agency (ESA), to facilitate the validation and verification of low-orbit or geostationary environmental satellites.This network is attempting to expand air quality monitoring through integration with existing long-term air quality monitoring stations.Since 2020, the National Institute of Environmental Research, Economic and Social Commission for Asia and the Pacific, and Korea Environment Corporation have been extending the Pandora Asia Network to include 13 Asian countries, with support from the Korea International Cooperation Agency.The Pandora Asia Network is expected to be widely used to study urban air quality in Asia, which is increasingly deteriorating due to rapid economic growth.
As part of the GEMS Map of Air Pollution (GMAP) campaign, a suite of Pandora instruments was deployed in Seosan, a South Korean coastal city, from November 2020 to January 2021 (GMAP-2020), and we applied GMAP-2020 measurements to explore the synergy of PC observations when evaluating air quality over Seosan.Further results from this research project are also reported in this special issue, including GEMS validation and urban air quality evaluations based on Pandora, aircraft, surface flux, and in situ surface chemical measurements conducted during GMAP-2020.

GMAP-2020 campaign
GEMS was launched on 19 February 2020; it is the first instrument to observe air quality from a geostationary Earth orbit.GEMS provides hourly air quality data on aerosols and gases at a spatial resolution of 7 km × 8 km.It is a scanning During GMAP-2020, Pandora instruments (PA 1 -PA 4 ) were deployed near large point sources (LPS 1 -LPS 4 ) in Seosan; in situ surface air quality monitoring systems (AQM 1 -AQM 6 ) and meteorological observations (Met 1 -Met 3 ) were also used in this study (see their locations in Fig. 1).Aircraft measurements were also used to validate GEMS and diagnose LPSs located in industrial areas surrounding Seosan.We explored the synergy of Pandora observations and SI measurements, based on measurements collected during GMAP-2020, by evaluating air quality in industrial Seosan (where LPSs are densely clustered).We particularly investigated the impacts of vertical profile and sub-pixel variability for trace gases and aerosols, for further GEMS validation.All measurement sites for both GMAP-2020 campaigns are indicated in Fig. 1.

Study area
Seosan, the target area of the GMAP-2020 campaign, is a small city with a population of 174 780 in 2017; it is accessed via three expressways to the east and four national highways cross the city.It is located in midwestern South Korea, and is affected by > 300 emissions point sources including LPSs.Coal-fired power plants including Taean, Dangjin, the Hyundai Dangjin steelworks, and the Daesan petrochemistry industrial complex (LPS 1 -LPS 4 , respectively, in Fig. 1) have the highest emissions rates in South Korea.The Hyundai Dangjin steelworks (LPS 3 ) and Taean and Dangjin power plants (LPS 1 and LPS 2 ) emit 10.5, 11, and 8.8 Gg of NO x per year, respectively.Although Seosan accounts for only 1.8 % of the population of Seoul, its NO x emissions (10.2 Gg yr −1 ) account for 13.2 % of its total NO x emissions.The transportation sector of Seosan is a far greater NO x source than the industrial sector of Seoul (ratio of 99 : 1); however, within Seosan, the industrial sector is on par with the transport sector (52 : 48; https://www.air.go.kr/en-main.do,last access: 22 August 2022).
During the past decade, the annual mean NO 2 level in Seosan has been 17 ppb, which is approximately half of that https://doi.org/10.5194/acp-22-10703-2022Atmos.Chem.Phys., 22, 10703-10720, 2022 in Seoul (31.2 ppb).NO 2 exhibits strong seasonal variation, reaching a minimum in summer and maximum in winter, due to meteorological factors and greater energy use during winter (Kim and Kim, 2020).Therefore, the timing of the GMAP-2020 campaign was well suited to tracking pollution.

Pandora measurements
Pandora measures the UV and visible wavelengths (280-525 nm) of direct sunlight with a spectral resolution of 0.6 nm, to determine the vertical column density of NO 2 , O 3 , and HCHO (Hermanet al., 2009).For measurements in Dobson units (DU; 1 DU = 26.9Pmol cm −2 ), column NO 2 has a very high signal-to-noise ratio (700 : 1) and very high precision (0.01 DU) for clear skies (Herman et al., 2009).The vertical column density of NO 2 can be determined using DOAS software (Van Roozendael and Fayt, 2001).Pandora directsun measurements are advantageous in that the air mass factor is simplified, and therefore is dependent only on the geography for a known solar zenith angle.
From retrieved Pandora measurements, tropospheric and total (= tropospheric + stratospheric) vertical column densities are both available.However, it should be noted that appreciable uncertainties cannot be neglected in the tropospheric NO 2 profiles obtained from Pandora instruments, particularly for the high aerosol-loading areas such as East Asia.In this background, we used total vertical column densities in the present study, and also confirmed that they have a high correlation with the tropospheric column densities observed in our study period, with little change in stratospheric column density in space and time at the local scale.
Four Pandora instruments were installed at sites to the south of LPSs (Fig. 1) during the GMAP-2020 campaign, i.e., at Seosan Daehoji, Seosan Dongmun, Seosan city council, and Seosan super site (PA 1 -PA 4 in Fig. 1).The presence of clouds reduces vertical column density precision by decreasing the number of photons arriving at Pandora instruments within a fixed integration time.Therefore, the retrieved Pandora measurements were cloud-screened using an observed cloud cover of 0.6.Cloud cover was provided by the Korea Meteorological Administration (KMA), and the precision improvement afforded by cloud screening was verified by comparing each Pandora-derived vertical column density with the median vertical column density, with and without cloud screening within the inter-comparison period.
At PA 4 , the operating period was extended to cover almost the entire year (12 November 2020-30 October 2021) including the GMAP-2020 campaign period, and the Pandora spectra were processed into vertical column density data for trace gases using the standard NO 2 algorithm in BlickP software provided by PGN (Cede, 2019).The resultant PC-NO 2 data were obtained from the PGN website (https:// pandonia-global-network.org, last access: 22 August 2022) for the 1-year period from 12 November 2020 to 30 Octo-ber 2021, and were also used as PC-NO 2 statistics for PA 4 , in this study.

Surface and airborne chemical measurements
Hourly average data for SI-NO 2 over a period of 1 year were obtained from Ministry of Environment AQM network stations in Seosan: Pandori, Leewon, Taean, Dongmoon, Seongyeon, and Daesan (AQM 1 -AQM 6 , respectively, in Fig. 1).The Seosan super site (PA 4 /AQM 1 ) provided hourly data for NO and NOy via an NO-DIF-NOy analyzer (42i-Y; Thermo Scientific, Waltham, MA, USA), and for PM 2.5 chemical species using an ambient ion monitor (AIM; URG 9000D, URG Corp., Chapel Hill, NC, USA).Weekly zero and span checks were conducted for NO y calibration to ensure that differences between checks remained < 3 %.Water-soluble ions in aerosol and gaseous species were measured hourly using an AIM, and ion mass balance was used to ensure data quality under the quality control procedures of the AQM network installation and operation guidelines (NIER, 2021).
Aircraft measurements were conducted during the GMAP-2020 campaign period, and nine flights were conducted on 8 d (26, 27, and 28 November and 1, 6, 8, 9, and 12 December 2020).The horizontal and vertical distributions of NO 2 and O 3 over Seosan were measured during GMAP-2020 using an NO 2 monitor (T500U; Teledyne, Thousand Oaks, CA, USA) and an O 3 analyzer (TEI49C; Thermo Scientific) onboard the Cessna Grand Caravan 208 B. These instruments had response times of < 40 and < 20 s, and detection limits of 40 ppt and 1 ppb, respectively.The flight paths included a raster mode over all of Seosan at a height of 500-700 m and a profiling mode from 500 m to 1.5 km over PA 1 and PA 4 (Fig. 2).

Meteorological measurements
Ground-based hourly observation data for meteorological variables were obtained from Seosan Automated Synoptic Observing System (ASOS) stations maintained by the KMA, and wind and temperature profile data were obtained twice daily (00:00 and 12:00 UTC) via a rawinsonde instrument at the Osan World Meteorological Organization upper air measurement station (47122) near Seosan.Due to time constraints of the sonde measurements, information on PBLH variation was obtained from Unified Model (UM) simulation results provided on the KMA website (https://afso.kma.go.kr, last access: 22 August 2022).
During the GMAP-2020 campaign, a 3D sonic anemometer (CPEC200; Campbell Scientific, Logan, UT, USA) was also installed on the rooftop at PA 4 for turbulent flux measurements at the city-atmosphere interface (Hong et al., 2019).All wind components and sonic temperatures were measured at a 10 Hz sampling rate, and ground-level sensitive heat flux was measured directly using a 30 min averag- ing period.Quality controls such as double rotation, spike removal, and outlier filtering were also applied.

Correlation analyses
We examined the synergy of PC and SI data obtained during the GMAP-2020 campaign, and combined these measurement data to evaluate air quality in Seosan, South Korea.We attempted to interpret the meteorological and photochemistry data measured during GMAP-2020, and to demonstrate that caution is required when attempting to study the vertical structure of air pollutants using either surface observations or satellite data only, particularly in industrial areas.
First, we examined the combined use of year-long PC-NO 2 and SI-NO 2 measurements, and investigated the factors modulating their correlation.Numerous studies have examined the correlations between chemical species, including aerosols (Thomspon et al., 2019;Wang et al., 2019;Kim et al., 2012Kim et al., , 2018;;Jo and Kim, 2013;Sanchez et al., 1990).Wang et al. (2019) reported that aerosols moderately correlate with NO 2 due to the frequent occurrence of lifted layers probably related to the transport of pollutants.Jo and Kim (2013) differentiated haze types using the trajectory speed and direction and different synoptic conditions.In this background, we hypothesized that their differences in PC-NO 2 and SI-NO 2 were due to meteorological conditions, and performed k-means and agglomerative hierarchical cluster analyses of various meteorological variables.Clustering is the grouping of objects that are alike and are different from the objects belonging to other clusters.As a first step, kmeans clustering was applied to find smaller clusters until each object was classified in one cluster.Subsequently, agglomerative hierarchical steps are applied to make up for the shortcomings of k-means clustering, in which once merging (or splitting) is done, it can never be undone.More details are found in Venkat Reddy et al. ( 2017).We used XLSTAT software (Addinsoft, Paris, France) for the cluster analysis with eight meteorological variables representing local and synoptic circulations in the cluster analysis: surface wind speed (Wsfc), 925 hPa temperature (T925), sea level pressure (Psfc), pressure tendency (dPsfc/dt), 850 hPa wind speed (W850) and its north-south and east-west components (NS850 and EW850), and 500 hPa geopotential height (GPH500).We subtracted 30 d moving averages from all data to account for typical seasonal variation.Monthly averages were used for PC-NO 2 analysis due to the limited availability of hourly data.
Correlations between PC-NO 2 and SI-NO 2 were analyzed in each meteorological group and the impact of photochemistry was interpreted based on case-specific features.We also investigated correlations in association with near-surface micrometeorological variables such as PBLH in each meteorological group.

Correlation analysis results for PC-NO 2 and SI-NO 2
The yearly PC-NO 2 statistics at four Pandora sites (PA 1 -PA 4 ) are summarized in Table 1.The total averaged PC-NO 2 over all sites was 0.45 DU during GMAP-2020, which is well above the typical values (0.1-0.2 DU) for Anmyeondo (the location is shown in Fig. 1), a representative background site (Herman et al., 2018).Although site PA 3 is located in a rural area, it nevertheless exhibited the highest PC-NO 2 amounts, suggesting that plumes were frequently transported from nearby point sources and/or urban areas.
Scatter diagrams of hourly PC-NO 2 and SI-NO 2 measurements from Pandora sites PA 1 -PA 3 (GMAP-2020)   are shown in Fig. 3a.These hourly data had a fair 1 : 1 linear relationship (R = 0.41), implying the overall uniformity of NO 2 profiles, whereas the linear relationship with PC-NO 2 weakened as SI-NO 2 levels increased (Fig. 3a).It appears that the SI-NO 2 has a distinct diurnal change despite the same PC-NO 2 , and higher variable surface NO 2 levels may result from the relatively weaker linear relationship between PC-NO 2 and SI-NO 2 .To explore these anticorrelation cases further, we selected the lower and upper bounds of the tendencies; these are plotted in Fig. 3b, which shows that PC-NO 2 was positively correlated with SI-NO 2 on 24 February 2021 (R = 0.88), while a negative correlation occurred on 21 April 2021 (R = −0.88),indicating a wide range of case-specific correlations.The negative correlation on 21 April (Fig. 3b) implied that the nonhomogeneous NO 2 distributions vertically were partially due to the photochemical process.For example, the decrease in PC-NO 2 despite an increase in SI-NO 2 might have occurred because NO 2 is removed by photochemical loss; it can occur more severely in the upper atmosphere with high OH concentrations.Another possible reason is the occurrence of lifted layers related to pollutant transport, yielding sharp changes in vertical concentration from the surface to the upper layer.The casespecific discussion follows.

Impacts of meteorological conditions on correlations between PC-NO 2 and SI-NO 2
Our k-means cluster analysis distinguished three groups with the lowest within-group variance and largest among-group variance.Among the total of 141 cases, 47, 66, and 28 were classified into groups 1-3, respectively.Thus, group 2 had the largest proportion of cases (47 %) and group 3 had the smallest (20 %).The combination of meteorological components in group 1 indicated the end of a high-pressure system (Psfc > 0, dPsfc/dt < 0), with southerly winds (NS850 > 0) bringing warmer air (T925 > 0) to the region, leading to stable atmospheric stratification and weak surface winds (Fig. 4).This group 1 meteorological mode appeared to result in very weak NO 2 ventilation, which produced the highest PC-NO 2 and SI-NO 2 values.Group 3 showed the opposite trend, with strong northerly winds bringing colder air into the region, leading to an unstable atmosphere and stronger surface winds, and ultimately decreasing PC-NO 2 and SI-NO 2 to their lowest levels.SI-NO 2 was approximately twice as high in group 1 than group 3, whereas PC-NO 2 showed no significant difference (Fig. 4a).We hypothesized that PBLH might also differ significantly under these micrometeorological conditions; therefore, we further explored daily maximum PBLH simulated by the Global Forecast System (GFS) and Lagrangian back- ward trajectories obtained from Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) and GFS system for the 141 cases.The mean simulated PBLH in Seosan, our study area, was 942.1 ± 405.3 m for 2020, which was similar to the annual mean daily maximum PBLH (1013.6 m) in Osan (Lee et al., 2013).However, the simulated PBLH differed significantly among the three groups (767.0 ± 304.8, 923.2 ± 335.3, and 1280.6 ± 501.2 m for groups 1-3, respectively).The PBLH for group 3 was 1.7-fold higher than that for group 1 (Fig. 4k).We also detected significant differences among the three groups in synoptic components of the lower troposphere including W850, and in local meteorological parameters such as the sea breeze index (SBI) suggested by Biggs and Graves (1962), which is defined as, SBI = U 2 C p T , where U is Wsfc (Fig. 4c), C p is specific heat, and T is the temperature difference between T925 and the sea surface temperature.Thus, the SBI represents the ratio between inertial (ρU 2 /2) and buoyance forces (ρgc p T ), where ρ is air density and g is gravity, and its value provides an indication of the likelihood of local circulation events such as sea breezes; at higher SBIs (i.e., SBI > 3), sea breezes cannot overcome the prevailing wind, whereas lower SBIs (i.e., 0 < SBI < 3) can indicate strong sea breezes.
In the example shown in Fig. 4l, the SBI for groups 1-3 was 0.1 ± 4.5, 0.1 ± 9.2, and −0.2 ± 12.5, respectively.Most SBIs in group 1 ranged from 0 to 3, indicating that group 1 corresponded to the dominant local circulation (LD), whereas the SBIs in group 3 had the lowest frequencies comparing 1 and 3, which corresponded to a dominant synopticscale circulation (SD).Group 2 can be considered a mixture of local and synoptic-scale circulation (MD).These results indicate that Seosan may experience frequent LD conditions (with sun on 1/3 of the days of the year), with infrequent SD conditions (1/5 of all days).https://doi.org/10.5194/acp-22-10703-2022Atmos.Chem.Phys., 22, 10703-10720, 2022 obtained previously in a study of surface and OMI-NO 2 measurements downwind of strong point sources in Israeli cities (Boersma et al., 2009).The intercept (0.17 DU) was within the range of previous Anmyeondo Pandora measurements, suggesting that intercepts of 0.15-0.2DU may represent the local background PC-NO 2 amount (including the stratospheric NO 2 ), rather than the influence of local anthropogenic NO 2 emissions.We classified daily averaged PC-NO 2 and SI-NO 2 data according to the three meteorological conditions (LD, MD, and SD) and detected a weak correlation under LD conditions (Fig. 5b); the lowest coefficient of determination for the LD condition (R 2 = 0.34) was approximately half of those for the MD (0.359) and SD (0.64) conditions, suggesting that NO 2 vertical profiles were more complex under LD conditions, with anomalous layers.

Diurnal variation in column-surface NO 2 under LD, MD, and SD conditions
Diurnal patterns of PC-NO 2 , SI-NO 2 , and O 3 under SD, MD, and LD conditions are shown in Fig. 6.Under LD conditions, PC-NO 2 increased from morning to afternoon (Fig. 6a), whereas under SD conditions, it had a weak morning peak and subsequent decrease until late afternoon (Fig. 6c).Under MD conditions, PC-NO 2 had one large peak in the morning and a shoulder peak in the late afternoon (Fig. 6b).However, SI-NO 2 showed nearly identical diurnal patterns among the three meteorological conditions, with an early morning peak followed by a second peak in the late afternoon (Fig. 6df).Diurnal patterns of O 3 were strongly associated with O 3 -NO 2 photochemical reactions under both LD and MD conditions (Fig. 6g and h), whereas no particular photochemical effects were detected under SD conditions (Fig. 6i).
A simple linear regression was applied to daytime average (11:00-17:00 LST) measurements of both PC-NO 2 and SI-NO 2 under the three meteorological conditions, and yielded correlation coefficients (R) of 0.51 and 0.41 for SD and MD conditions, respectively; however, LD conditions produced a significantly lower R (0.27).Thus, under SD conditions, strong synoptic winds suppressed PC-NO 2 and SI-NO 2 diurnal fluctuations, rendering them similar to each other.Strong winds also inhibited local effects of O 3 formation on the diurnal variation in PC-NO 2 , and the smaller impact of chemical conversion from local NO 2 to O 3 lowered R values during the day.Under MD conditions, both PC-NO 2 and SI-NO 2 exhibited distinctive peaks in the morning with a degree of time lag; both subsequently declined toward noon, and showed higher R values than those obtained under SD conditions.By contrast, under MD conditions, correlations were enhanced due to a minimum around 15:00 KST for both PC-NO 2 and SI-NO 2 , despite time lags in both peaks in the morning and afternoon.Previous studies of the Megacity Air Pollution Seoul (MAPS-Seoul) and KORUS-AQ campaigns reported a typical pattern of continuously increasing PC-NO 2 over the Seoul metropolitan area (Chong et al., 2018;Herman et al., 2018).However, in the current campaign, we found similar results only under LD conditions.The diurnal patterns reported in previous studies were mainly caused by the dominance of NO 2 emissions sources over NO 2 losses (Chong et al., 2018;Herman et al., 2018) among several processes associated with NO 2 photochemical loss, including transport and deposition, which were also investigated in specific cases in the current study.
In this study, we extended the correlation analysis and investigated the correlation between hourly PC-NO 2 and SI-NO 2 data.The results show a lower correlation in the morning and a higher correlation in the afternoon (Fig. S1 in the Supplement).The respective median correlation coefficients for the LD, MD, and SD meteorological condi-tions were −0.71, 0.18, and 0.22 in the morning (09:00-12:00 LST), and 0.84, 0.77, and 0.79 in the afternoon (12:00-14:00 LST).These values may reflect PBL development.SI-NO 2 decreases in the morning due to the rapid growth of the PBL, while PC-NO 2 increases due to the accumulation of NO 2 in the atmosphere, deriving a lower correlation.However, there is very little change in the PBL in the afternoon, and PC-NO 2 and SI-NO 2 show similar changes, yielding a positive correlation between them during the GMAP-2020 campaign.

Aircraft measurements collected during GMAP-2020
Data collected via aircraft during GMAP-2020 are summarized in Table 2.A total of nine aircraft measurements were conducted during the campaign period (12 November 2020-20 January 2021).Four of nine flights were conducted under LD conditions, and the remaining flights (except that on 27 November 2020) were conducted under MD conditions.https://doi.org/10.5194/acp-22-10703-2022Atmos.Chem.Phys., 22, 10703-10720, 2022 No aircraft measurements were consistent with SD conditions during the GMAP-2020 campaign.
We examined spiral segments from each flight over Seosan during 11:00-17:00 KST to exclude marginal effects of diurnal variation in NO 2 (Fig. 2).The overall results indicated that the vertical O 3 profiles were relatively constant in the PBL, whereas NO 2 profiles appeared to be highly dependent on meteorological conditions.We compared data collected during flights conducted under LD (one flight) and MD conditions (two flights) during the GMAP-2020 campaign, to examine differences in the vertical structures of the PA and SI observations.Aircraft measurements of vertical NO 2 and O 3 profiles for flights FL-5 (6 December) and FL-6 (8 December) under LD conditions are shown in Fig. 7, along with 24 h backward trajectories starting at different altitudes (100, 500, and 1000 m).All observed NO 2 profiles shown in Fig. 7 ap-peared to have generally exponential curves, with anomalous features at higher altitudes.For example, when vertical turbulent mixing prevailed within the PBL (O 3 profile, Fig. 7b), the data were fitted with an exponential vertical curve, and the anomalous NO 2 layer aloft was found to have a height of 1.5 km, which was higher than the estimated PBLH of 1.2 km.HYSPLIT 24 h backward trajectories starting at 12:00 KST showed that all air mass from the surface to the lower free atmosphere was transported over the Yellow Sea via the Shandong Peninsula (Fig. 7c).This finding suggests that the anomalous NO 2 layer aloft was not produced locally (i.e., from local LPS emissions), but instead traveled via long-range regional-scale transport.This transport of NO 2 across the region was also discussed and might be particularly high during the winter when the NO x lifetime is relatively longer (Stohl et al., 2002;Wenig et al., 2003;Lee et al., 2013).According to Anmyeondo lidar measure- ments for 6 December (http://kalion.kr,last access: 22 August 2022), the anomalous NO 2 layer aloft corresponded well to an aerosol layer that appeared at ∼ 1.0 km at approximately 12:00 KST, persisting until 2200 KST.However, based on a cross-comparison of our data, high surface levels of SI-NO 2 (>∼ 4 ppb; Fig. 7a) were influenced more by local LPS than by that in the atmosphere aloft due to long-range transport (Fig. 7a).
Aircraft measurements for flight FL-7 (9 December) under LD conditions are shown in Fig. 7b.The NO 2 vertical profile exhibited an exponential curve, with an anomalous peak at ∼ 600 m immediately above the top of the simulated PBL.HYSPLIT backward trajectory data starting at 12:00 KST showed that the non-surface air had a different origin from the surface air (Fig. 6d), indicating that the anomalous NO 2 plume likely traveled from coal-fired power plants in a nearby industrial city (Taean) northwest of Seosan.This finding indicates a distinct vertical structure of higher NO 2 at the surface due to strong local emissions, whereas lower NO 2 levels were observed at higher altitudes, with anomalously high NO 2 levels in some layers aloft due to medium-range transport from nearby areas.Thus, despite the limited number of aircraft measurements, the elevated anomalous NO 2 structure that was observed intermittently led to a negative correlation between PA-NO 2 and SI-NO 2 .The discrepancies imply that vertical profile distribution study should proceed cautiously when only surface measurements are obtained under LD meteorological conditions.
Aircraft measurements were conducted under MD conditions on flights FL-1 (26 November), FL-3 (28 November), and FL-8 (12 December) (Fig. 8).We applied several regression models (linear, exponential, and polynomial) to three vertical structures, and obtained two distinct NO 2 https://doi.org/10.5194/acp-22-10703-2022Atmos.Chem.Phys., 22, 10703-10720, 2022 vertical profile patterns from the surface to the PBLH: decreasing linearly for FL-1 and FL-8 (Fig. 8), and constant with altitude for FL-3 (Fig. 8b).None of the three cases showed anomalous layers above the PBLH, similar to the exponentially declining profiles obtained under LD conditions (Fig. 7).These vertical structures observed under MD conditions may have been induced by strong vertical mixing within the PBL, supplemented by prominent surface photochemical losses at the same time.The vertical O 3 profile during FL-1 showed a decoupled structure, with different patterns within and above the PBL (Fig. 8d); however, the other 2 d showed uniform distributions, with no particular anomalous features between the upper PBL and surface atmosphere (Fig. 8b, c, e, and f).The observed daily maximum sensible heat fluxes measured at Seosan (Fig. S3) were much higher for FL-3 (175.9W m −2 ) than FL-1 and FL-8 (118.9 and 102.0 W m −2 ), suggesting that vertical turbulent mixing was much more prominent during FL-3.These chemical and physical characteristics are all related to MD conditions.Thus, the higher coefficient of determination (R 2 = 0.64) obtained under MD conditions (Fig. 5b) has an important bearing on the absence of irregular or anomalous layers aloft, with little variation regardless of the shape of the curve (Figs. 7 and 8).

Analyses of column-surface relationships for specific GMAP-2020 cases
Figure 9 shows examples of PC-NO 2 and SI-NO 2 diurnal variation under LD (FL-5 and FL-7) and MD (FL-1 and FL-8) conditions, and Fig. 10 shows latitudinal mean distributions for FL-5 and FL-7, based on the aircraft measurement data shown in Figs.7 and 8. PC-NO 2 was decoupled from SI-NO 2 on 2 d, FL-5 and FL-7, which were both classified as having LD conditions (Fig. 9a and b), whereas good vertical mixing and uniform NO 2 distribution were observed on the remaining 2 d, FL-1 and FL-8, which showed MD conditions (Fig. 9c and d).According to our analysis of the aircraft measurements (Fig. 7), the poor correlations between PC-NO 2 and SI-NO 2 captured by FL-5 and FL-7 were mainly due to an NO 2 polluted layer transported aloft, as described in Sect.4.3.

LD conditions
Several cases showed poor correlations between PC-NO 2 and SI-NO 2 under LD conditions within the study period.
When we examined the results of previous studies (Thompson et al., 2019;Chong et al., 2018;Herman et al., 2018;Kim et al., 2021), we first considered the possibility that LPS emissions influenced downwind regions under LD conditions because the increase in PC-NO 2 , but not SI-NO 2 , may have required an additional source of NO 2 apart from early afternoon traffic emissions.The FL-5 data for December 6 represent an example of this, showing a poor correlation between PC-NO 2 and SI-NO 2 (R 2 = 0.06; Fig. 9a).On the same day, Anmyeondo lidar detected two elevated aerosol layers at 12:00 and 16:00-22:00 KST (http://kalion.kr, last access: 22 August 2022); the first aerosol layer may reflect a PC-NO 2 peak, as shown in Fig. 9a.The HYS-PLIT backward trajectories, starting at different altitudes from the surface to the lower troposphere, revealed that all air parcels moved eastward from China to Anmyeondo and Seosan (Fig. 1); thus, other NO 2 plumes may have begun to pass over Seosan at 16:00 KST (Fig. 7c).Longitudinal SI-NO 2 distributions (Fig. 10) exhibited 5.2 ppb at 126.1 • E, 8.1 ppb at 126.3 • E, and 7.3 ppb at 126.4 • E, averaged between 13:00 and 16:00 KST by longitude (Table S1 in the Supplement), whereas they were nearly constant at a height of 500-600 m on 6 December.Therefore, westerly winds advected cleaner air from Padori (AQM 1 ) to Seosan at the surface, but not at a height of 500-600 m, contributing to low SI-NO 2 levels in the afternoon (Fig. 9a).

MD and SD conditions
We obtained higher PC-SI correlation coefficients under MD and SD conditions than LD conditions (Figs. 5b,and 9c and d).Under MD and SD conditions, diurnal variation in PC-NO 2 and SI-NO 2 showed simultaneous declines from early morning until noon (Fig. 6).Notably, PC-NO 2 showed a continuously decreasing trend, particularly during the morning hours, in the period of approximately 09:00-12:00 KST under both MD and SD conditions (Fig. 6b and c).These diurnal patterns of decreasing PC-NO 2 in the study area were opposite to those reported in previous studies (Chong et al., 2018;Herman et al., 2018) that observed increasing PC-NO 2 in large urban areas during the daytime, caused by higher NO 2 emissions even during photochemical NO 2 losses to form O 3 .We hypothesized that decreasing PC-NO 2 can occur due to photochemical loss and surface wind transport, which both intensify with increasing solar radiation in the morning.Photochemically, NO 2 is converted into photochemical oxidants such as PAN, HNO 3 , and nitrate under sunlight, thereby disrupting the NO x -VOC-O 3 cycle.Concurrently, Wsfc intensified due to thermal turbulence transport of NO 2 emissions away from Seosan during the day.Thus, PC-NO 2 decreases under MD conditions as a result of ventilation effects caused by stronger wind speeds.There are two possible mechanisms for this: sea breeze penetration (because the study area is adjacent to the northern coast of the Taean Peninsula; Fig. 1), and vigorous turbulent mixing (which leads to vertical mixing of surface NO 2 during PBL growth; Sun et al., 2013).We investigated these factors in detail for specific cases.
Figure 11 shows the diurnal variation in selected meteorological and chemical variables measured under MD (25 November) and SD conditions (14 December).Under MD conditions (Fig. 11a-c), declines in PC-NO 2 and SI-NO 2 were observed toward noon.In particular, decreasing PC-NO 2 was accompanied by increased Wsfc (Fig. 11b); therefore, we examined GMAP-2020 campaign measurements of sea breeze penetration.Figure S2a shows diurnal variation in observed air temperatures at site Met 1 and measured sea surface temperatures at nearby site Met 2 (37.14 • N, 126.01 • E), located 55 km from PA 4 .The thermal meteorological observations were used to calculate SBI (+0.37), which was greater than +3 (the threshold for sea breeze occurrence; Brigges and Graves, 1962).Sea breeze disturbances with a sharp decrease (increase) in temperature (humidity) were observed at site Met 3 (Fig. S3b), which is located on the northern coastline of Taean Peninsula (Fig. 1).However, sea breezes did not progress inland at the Met 1 Seosan Meteorological Automated Surface Observing System (ASOS) site, which is closer to the Pandora sites; sea breezes did not correlate with NO 2 ventilation to offset its high emission.
We further detected a strong positive correlation between wind speed and sensible heat flux (Fig. 11b).We speculated that thermal and momentum turbulences caused by a vertical temperature gradient and surface friction entrained surface turbulence, thus increasing momentum in the free atmosphere downward to the surface due to strong turbulent mixing within the PBL, in turn leading to a uniform vertical NO 2 profile with a positive correlation between PC-NO 2 and SI-NO 2 .Figure S3 shows a comparison of daily maximum sensible heat and momentum fluxes under LD, MD, and SD conditions during the GMAP-2020 campaign.SD conditions showed the highest mean heat flux, followed by MD and LD, indicating that downward momentum transport led by both heat and momentum fluxes plays a greater role in Wsfc enhancement under MD than LD conditions within the PBL.
Photolytic NO 2 loss was detected as temporal variation in NO 2 , NO − 3 , and CO at PA 4 .Because no NO 2 analyzer was installed at PA 4 , NO * 2 (= NO y -NO) was used instead of NO 2 under the assumption that NO z is negligible in winter.Figure 11c shows the diurnal variation in NO 2 , O 3 , and NO − 3 under MD conditions, normalized by CO to reduce the effect of PBL evolution.The results showed that NO 2 /CO decreased after the morning peak; however, NO − 3 /CO and O 3 /CO increased toward midday, indicating that photolytic activity also contributed considerably to the concurrent decline of SI-NO 2 and PC-NO 2 (Fig. 11a).In turn, this indicated that photochemistry can contribute to higher correlation coefficients under MD conditions.
Under SD conditions (Fig. 11d-f), PA-NO 2 and SI-NO 2 exhibited weak diurnal variability compared to LD and MD conditions.SD conditions on 14 December produced significantly stronger winds (i.e., wind speed > 6 m s −1 at 13:00 KST), with generally higher PBLHs (Fig. 11e).Meteorological features, such as strong wind at both 850 hPa (18.0 m s −1 ) and 10 m height (4.26 m s −1 ), suppressed both PC-NO 2 and SI-NO 2 (7.3 ppb and 0.31 DU, respectively) to below the average, producing a strong correlation (R = 0.9 at AQM 5 ) and nearly flattening their temporal curves during the day (Fig. 11d).Thus, under SD conditions, wind speed and turbulent fluxes such as sensible heat flux had larger values, and NO 2 and NO − 3 decreased or increased at the same time during the day (Fig. 11f), indicating that the transport effect was much greater than that of local photochemical loss over the study area.
In conclusion, in this case-specific study, we assessed the correlations between PC-NO 2 and SI-NO 2 , and explored their mechanisms by investigating the impact of meteorological and photochemical conditions.A weak correlation between PC-NO 2 and SI-NO 2 occurred when anomalously high concentrations remained, with ragged fragments of NO 2 plumes in the upper or middle layers.We also found that a negative correlation occurred intermittently under LD con- ditions, with generally lower PBLH.In particular, elevated pollutant levels due to regional-scale transport or decoupled NO 2 plumes advected within the PBL may have also caused the weak correlation between PC-NO 2 vs. SI-NO 2 .These phenomena were detected only from the PA-SI coupled measurements in this study.Thus, when either PC or SI observations are applied alone for understanding the vertical structure of air pollutants, undetected bias can occur under LD conditions, particularly where transport processes prevail, and these results can be also applicable to GEMS observations analysis.

Conclusion
We explored the potential applicability of combined PC-NO 2 and SI-NO 2 measurements collected at Seosan during the GMAP-2020 campaign.We characterized the correlation between PC-NO 2 and SI-NO 2 under various conditions to understand the complex air quality of Seosan, which appears to be vulnerable to LPS emissions from surrounding areas.We hypothesize that correlations between PC-NO 2 and SI-NO 2 are closely related to NO 2 vertical profiles, which also depend on meteorological conditions.We performed statistical analyses of a year-long PC-NO 2 dataset (12 November 2020-30 October 2021) combined with meteorological data, in situ ground data, and airborne chemical data measured during the GMAP-2020 campaign in the same period.
https://doi.org/10.5194/acp-22-10703-2022Atmos.Chem.Phys., 22, 10703-10720, 2022 Our results show that hourly PC-NO 2 and SI-NO 2 over the 1-year period exhibited a linear relationship with a fair correlation (R = 0.41), and daily mean PC-NO 2 vs. SI-NO 2 exhibited a good linear correlation (R = 0.73), supporting the overall uniformity of NO 2 profiles in the PBL over Seosan despite the continuous impact of LPS emissions.
The impact of meteorological conditions on the relationship between PC-NO 2 and SI-NO 2 was investigated through agglomerative hierarchical clustering, which indicated three meteorological conditions: LD, MD, and SD.Under LD conditions, southerly winds advect warm air under the upper ridge, forming stable and short PBLs and weak surface winds.By contrast, under SD conditions, cold northerly winds induce unstable and high PBLs with strong surface winds.The correlations between daily mean PC-NO 2 and SI-NO 2 levels, and their variation during 11:00-17:00 KST, weakened under LD conditions, suggesting that the shape of the NO 2 profile typically deviates from a uniform profile under SD and MD conditions.Aircraft measurements under LD conditions demonstrated NO 2 plumes aloft, with anomalous vertical structures and different horizontal (latitudinal) gradients of surface NO 2 at higher altitudes, such as 600 m over Seosan.
Thus, the relationship between PC-NO 2 and SI-NO 2 depends on the presence of NO 2 plumes aloft under LD conditions, which provide a favorable environment for LPS plumes decoupled from the surface at Seosan.Our findings suggest that the correlation between PC-NO 2 and SI-NO 2 may serve as an indicator of the degree of complexity of urban air quality.This correlation can be optimally applied for air quality evaluation and vertical analysis by combining the Pandora Asia Network with AQM networks, and the results can be also applied to environmental GEMS observation analysis in combination with SI observations.More detailed studies on urban air pollution evaluation will be undertaken based on PC, DOAS, aircraft, SI air quality, and surface turbulence observation data, as well as modeling studies of data collected during the GMAP-2021 campaign.

Figure 2 .
Figure 2. Flight tracks for two Cessna Grand Caravan 208 B aircraft over Pandora sites (a) PA 4 and (b) PA 1 during the GMAP-2020 campaign.Colored circles indicate airborne NO 2 concentration observations.Stacked circles indicate spiral flights conducted over two sites.

Figure 3 .
Figure 3. (a) Pandora column (PC) NO 2 measurements as a function of surface in situ (SI) NO 2 observations at Pandora sites PA 1 -PA 3 during the GMAP-2020 campaign and PA 4 during a 1-year period.A 1 : 1 linear regression model was used to evaluate the relationship between PC and SI measurements (black line).(b) Sample scatterplots of PC-NO 2 and SI-NO 2 for 24 February (red) and 21 April 2021 (blue).

Figure 7 .
Figure 7. Box and whisker plots of the vertical NO 2 and O 3 profiles measured by GMAP aircraft superposed with in situ AQMS 1 measurements during flights (a, b) FL-5 (6 December) and (d, e) FL-7 (6 and 9 December).Blue dashed lines are linear regression lines fitted to NO 2 and O 3 profiles within the planetary boundary layer (PBL).Black arrows indicate the simulated PBL height (PBLH) obtained from the Korea Meteorological Administration (KMA).HYSPLIT 24 h backward trajectories in Seosan are shown at altitudes of 100, 500, and 1000 m, starting at 16:00 KST on 26 November and 12:00 KST on 12 December.

Figure 8 .
Figure 8. Box and whisker plots of vertical profiles obtained from GMAP aircraft superposed with in situ AQMS measurements for (1) NO 2 and (2) O 3 for flights (a) FL-1 (26 November), (b) FL-3 (28 November), and (c) FL-8 (12 December).Blue dashed lines are linear regression lines fitted to NO 2 and O 3 in the PBL.Black arrows indicate PBLH simulated by the Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) Global Forecast System (GFS).

Figure 11 .
Figure 11.Examples of the diurnal variation on 25 November (a, c, e) and 14 December (b, d, f).(a, b) Column NO 2 at sites PA 1 -PA 4 , and surface NO 2 at air quality monitoring sites AQM 4 and AQM 6 .(c, d) Sensible heat fluxes and surface wind speed at PA 4 .(e, f) Diurnal variation in NO 2 , NO − 2 , and O 3 normalized by CO. Figure 1 shows the locations of the measurement sites.