Impacts of atmospheric vertical structures on transboundary aerosol transport from China to South Korea

To forecast haze pollution episodes caused by high concentrations of long-range transported pollutants emitted in the areas upstream of South Korea, it is crucial to study and identify their behaviour. We analysed the three-dimensional air quality structure in Seoul using ground observation data and aerosol lidar measurements to identify vertical aerosol intrusion into the Korean Peninsula during the spring of 2016. The intrusions were particularly affected by the development of the atmospheric boundary layer (ABL) in the leeward regions. The nocturnal pollutant intrusion into the Korean peninsula via the Yellow Sea was examined using measured data. The pollutants first reached the area above the nocturnal boundary layer (548 ± 180 m) and approached ground level on the following day due to convective mixing depending on the convective ABL growth (1182 ± 540 m) in daytime. These intrusion mechanisms were mostly attributed to extremely high concentrations (i.e. >100 μg m−3) of fine particulate matter in the leeward regions, accounting for four of the total of six cases for which the warnings and alerts were issued in Seoul Metropolitan Area over a year-long period (2016). The horizontal and vertical pathways of the long-range transported pollutants and the atmospheric vertical structure were identified as key factors affecting the surface air quality concentration in the leeward regions.

air pollutants generated in China into the Korean Peninsula. We employed intensive observation data, including in-situ observations at mountain and surface sites, and aerosol lidar measurements as we focused on meteorological conditions related to vertical mixing status over the ocean and into inland areas. The impact of the intrusion of long-range transported pollution was contrasted against the non-transported air pollutant by direct comparison between high-altitude mountains and surface sites. We also discussed the role of the development of both the marine boundary layer (MBL) and ABL in the intrusion of long-range transported air pollution and enhancement of surface pollutant concentrations.

Results
High pM 10 concentrations at Mt. Gwan-ak and Seoul sites. South Korea lies on the leeward side of China and is thus heavily affected by the latter's long-range transported pollutants, especially in spring and winter when westerly and north-westerly winds predominate. Seoul, the capital of South Korea, is particularly exposed to a variety of interlaced air pollution sources, namely local emission sources from its dense population as well as transboundary air pollution sources. Several measurements are hence being recorded to examine the behaviour of such pollutants in Seoul. The present study compares the concentration levels of particulate matter less than 10 μm in diameter (PM 10 ) measured at different altitudes to visualize the pollutant behaviour in Seoul while accounting for the effects of quantified long-range transported pollutants and local emissions. In situ measurements of PM 10 , at various locations such as Mt. Gwan-ak at 629 m above sea level, and Song-wol at near the surface in the urban city centre were conducted, owing to the lack of previous in situ measurement records regarding detailed vertical profiles of PM 10 concentrations over the Korean Peninsula and the Yellow Sea ( Fig. 1). According to the time series comparison of PM 10 concentration levels measured at Mt. Gwan-ak in the spring of 2016 against other sites in the city centre including Song-wol, PM 10 concentration levels at Mt. Gwan-ak were often higher than those at other metropolitan surface sites ( Fig. 2(a)). Although the measurement location of Mt. Gwan-ak was in the vicinity of the city's centre, Mt. Gwan-ak, the mountain site, is surrounded by mountains, and thus lacks anthropogenic local emission sources. For this reason, the ground-level PM 10 concentration is typically higher at the urban (Seoul) centre (58.2 ± 33.0 μg m −3 ) than at Mt. Gwan-ak (50.6 ± 41.0 μg m −3 ) by approximately 7-8 μg m −3 ( Fig. 2(b)). Here the values in parentheses are the mean ± one standard deviation estimated during the study period (March-May in 2016). Despite the lack of local emissions, residential, commercial, and industrial pollutants emitted in Seoul, as well as those advected from China and other regions, yield higher PM 10 concentrations at Mt. Gwan-ak. In Fig. 2(b,c), the measured PM 10 concentration at Mt. Gwan-ak is sometimes higher than the ground-level concentration at the urban centre by more than 50.0 μg m −3 mostly between 21:00 and 06:00 Local Sidereal Time (LST). During the high episode period on 23 April, 2016, daily average of PM 10 levels at Mt. Gwan-ak were recorded to be 262.0 μg m −3 , showing a significant contrast against 181.0 μg m −3 recorded for the urban centre. The days on which the difference in the concentration exceeding 50 μg m −3 between the urban centre and Mt. Gwan-ak was noted, and were selected to determine the cause behind the high nocturnal concentrations at the mountain site.
The In an effort to identify the cause of the vertical PM 10 concentration gradient, the backscattering coefficient signal of aerosol lidar, which identifies the vertical structure of atmospheric particles, was employed for the selected days. Subsequently, the atmospheric vertical structures and corresponding pollutant distributions were examined accordingly. In Seoul metropolitan area during the study period, a total of 7 days for which the warning was reported were the following: March/6, April/8-10, April/23, and May/7; March/6 was not selected in the current study because of the missing lidar measurements. www.nature.com/scientificreports www.nature.com/scientificreports/ Vertical structures of pM 10 over Seoul Metropolitan Area. The comparison of PM 10 concentration distributions in Seoul city centre and Mt. Gwan-ak as well as aerosol lidar backscattering coefficient distributions measured at Seoul National University, located near Mt. Gwan-ak, shows higher concentration levels at Mt. Gwan-ak from 18:00 to 06:00 LST of next day. The aerosol lidar measurements also display a high concentration plume at 500 m and above (Fig. 3). In particular, a layer with high backscattering coefficient was observed at 500 m and above in the presence of Asian dust between 21:00 LST of 22 April and 09:00 LST of 23 April. The vertical mixing of pollutants emitted at low altitudes was thereafter invigorated by the development of the ABL after sunrise as the ground surface heated up, causing the backscattering coefficient to increase from low to high altitudes. At noon, the sufficiently developed ABL mixed with the PM 10 plume located approximately 1 km above the ground, resulting in a high PM 10 concentration level from high to low altitudes according to both PM 10 concentration measurements and aerosol lidar data. This phenomenon became more pronounced in the presence of Asian dust on 6 and 7 May, when the Asian dust plume observed at 1-2 km above the ground gradually intruded In addition to Asian dust, the upper aerosol layer was observed from the night of 15 March until the morning of 16 March. It began to mix with the lower layer, starting at approximately noontime. Although the difference between the upper and lower layers was not apparent on the night of 5 April, the backscattering coefficient was relatively low at night near the ground, whereas aerosols were found to be distributed across several layers at 500 m and above (Fig. 3). Furthermore, the aerosol distribution signals were observed at 500 m and above both on the night of 7 April and at dawn on 8 April, and relatively low values were observed near the ground at these times. A layer with a relatively high aerosol distribution signal was also identified as an isolated layer at 500 m above ground level on the night of 25 May, although the PM 10 concentration was not very high, as indicated in Fig. 3.
The vertical aerosol concentration difference decreased when aerosols that were present as isolated layers during nighttime at 500 m and above began to mix with air masses in the lower layer as the vertical mixture induced by the development of the lower atmospheric boundary from 09:00 LST met the upper aerosol layer.
In the case of aerosol lidar measurements, however, identification of the horizontal movement of high concentration aerosol layer is rather challenging as the data yield the vertical aerosol distribution over time at fixed measurement locations. Notably, highly concentrated aerosol distributions initially arose from the surface or from lower altitudes, and then gradually mixed in higher altitudes as a result of the growing ABL. It is thus difficult to distinguish, solely based on aerosol lidar and PM 10 concentrations measured at two locations, whether the phenomenon noted above is driven by such highly concentrated aerosol distributions prevailing above the nocturnal boundary layer (NBL) as its altitude decreases overnight, or by the long-range pollutants intruding into Seoul via streams at high altitudes. The horizontal aerosol distribution derived from the chemical transport model assimilated with observations of chemical compositions was evaluated to identify the path of the high concentration PM 10 plume. The spatial distribution of total AOD in the 550 nm wavelength range of the Copernicus Atmosphere Monitoring Service (CAMS) reanalysis data provided by European Centre for Medium-Range Weather Forecasts (ECMWF) is shown in Fig. 4. According to the figure, high AOD values (>0.5) were observed from China all the way to the west of the Korean Peninsula over the Yellow Sea when high nocturnal concentrations were measured at Mt. Gwan-ak on each selected day. It is also noteworthy that in some cases in this study, two or three

Influence of the nocturnal boundary layer/convective boundary layer. The horizontal and vertical
behaviours of aerosols measured and predicted for Seoul, Korea, verify the intrusion of pollutants generated in China via the upper layer at an altitude of 500 m and above instead of near the ground surface. The pollutants remain as a decoupled layer above the ABL, especially when they remain overnight, until the lower vertical www.nature.com/scientificreports www.nature.com/scientificreports/ mixing gradually occurs after sunrise and the mixing layer grows to transport the pollutants. High pollutants in the upstream area (i.e., emitted from cities and industrial plants) travel along the upper layer because they intrude through the varying land-sea-land surface conditions as the Yellow Sea is located between China and the Korean peninsula. These conditions are responsible for the long-range transport of the pollutants. Moreover, the vertical mixing induced in the emission region in the upstream area carries the local pollutants to higher altitudes. They then flow toward the Yellow Sea with the westerly and north-westerly winds. Most pollutants within the MBL are deposited at this time while the incoming pollutants above the MBL are diverted to leeward regions. Subsequently, the pollutants are believed to intrude into the Korean peninsula at high altitudes rather than near the ground surface as they are transported over the westerly MBL. When the pollutants that flow along the upper layer over the Yellow Sea reach Korea at nighttime, they remain above the lower NBL and do not mix with the air mass near the ground. The concentration levels at the city centre in Seoul are lower than those at Mt. Gwan-ak, whereas Seoul shows gradually higher levels when the long-range transported pollutants from China, which intrude into Seoul in the daytime gradually reach the lower layer after mixing at the top of the sufficiently developed ABL. The horizontal/vertical behaviour as well as the long-range transport and intrusion of pollutants generated in China are illustrated in Fig. 5.
During winter and spring, the air quality in the leeward regions in the Korean peninsula often degrades due to pollutants emitted in China, leading to severe public health concerns. As a result, the provision of prognostic air quality information to the public, by accurate identification of the intrusion period and mechanism of long-range transported pollutants, is imperative. This study suggests that long-range transported pollutants may have a greater impact on surface air quality in daytime than nighttime, and it aims to facilitate more accurate air quality forecasts for the leeward areas. In particular, many studies have demonstrated that when the altitude of the ABL is low, vertical mixing does not occur, and thus, the observed surface concentrations are high mainly due to the weaker dilution process [24][25][26][27][28] . In this study, we highlighted the inverse correlation between the boundary layer mixing process and ground-level particulate matter concentrations, suggesting that the development of the ABL can play a positive role by increasing the surface concentration by introducing inflowing pollutants from the upper atmosphere when the ABL height increases. Figure 6 shows the correlation between the estimated ABL height and the concentration differences between the upper and lower altitudes. Here, ABL heights were estimated by observing the vertical profiles of attenuated backscatter [29][30][31] ; these showed a seasonal average of 1182 (±540) m and 548 (±180) m during day and night, respectively, during spring 2016. As denoted in Fig. 6, the low NBL with the higher surface concentrations, interferes with the mixing of pollutant concentrations above and at the surface level, thus maintaining higher concentrations at the surface and lower ones at higher altitudes (PM 10 at surface > PM 10 at Mountain). This implies www.nature.com/scientificreports www.nature.com/scientificreports/ little influence of intrusion from above when meteorological conditions are stagnant. However, in the case of long-range transport, a significant difference in concentrations between the upper and surface levels with higher upper and lower surface PM 10 concentration patterns was confirmed, indicating that the upper layer pollutants do not affect the lower layer at night. In both cases, it was confirmed that the concentration difference between the upper and lower layers decreased as the altitude of the ABL increased during daytime.

Discussion
PM 10 warnings/alerts were issued in Seoul six times in total in 2016, and the following four such cases (67% of the total) are discussed here. They are characterised with high pollutant concentrations at Gwan-ak Mountain, which gradually affected the ground level concentrations. Thus, more than 67% of the deterioration in air quality over the Korean Peninsula was attributed to the inflow of the long-range transboundary of moving pollutant plumes. Detailed analyses of long-term 3-dimensional simulations/observations of other cases would be needed for more robust information on such intrusions, including the arrival heights and intrusion times under various meteorological conditions. This information can be used for the identification and prediction of starting and ending times of highly concentrated plumes.
In addition, understanding the vertical behaviours of these pollutant plumes can assist future studies on chemical reactions occurring in the upper and lower atmospheric layers 32,33 of the Korean peninsula. Intensive high-resolution vertical measurements of atmospheric chemical compositions during nighttime as well as daytime are required to analyse the physicochemical behaviours of the pollutants corresponding to daylight/nocturnal chemical reactions.

Methods observation data.
A ground-based surface PM 10 monitoring system operated by the Korea Meteorological Administration (KMA) provided PM 10 concentrations measured by an aerodynamic particle sizer (APS) at 5-min intervals over dozens of stations nationwide. In this study, the comparisons with PM 10 concentrations at different sites, namely Mt. Gwan-ak (37.44525°N, 126.96401°E, 622.38 m) and a ground-level site located within the urban centre (Song-wol, 37.57141°N, 126.96579°E, 85.8 m; Fig. 1), were conducted to estimate the vertical inflow structures and intrusions of aerosol plumes over the leeward areas.
In addition to PM 10 concentration measurements, real-time aerosol lidar observations at Seoul National University (37.460°N, 126.950°E, 300 m, near Mt. Gwan-ak) were used to analyse detailed vertical profiles of aerosol and atmospheric vertical structures related to evolution of the ABL. The aerosol lidar observations provided information on the vertical behaviours of the aerosols using the vertical profiles of range-correlated backscatter signals and depolarization ratios measured with 532 nm and 1064 nm wavelength lasers. However, the aerosol lidar observations were lacking in terms of missing values for extremely high aerosol pollution incidents. Detailed information on the aerosol lidar instrument appears at http://www.kalion.kr/intro/kalionRidar.
We also used ABL height data estimated from the vertical profiles of attenuated backscatter obtained by a ceilometer (Model CL51, Vaisala, Finland) at an urban meteorological observation system station located in the eastern part of the city of Seoul (37.59°N, 127.08°E, 45 m) [29][30][31] . The ceilometer provided the vertical distributions of two-way attenuated backscatter by means of a 910 nm wavelength laser as well as three levels of cloud base height. The attenuated backscatter with a vertical resolution of 10 m up to 15 km, and a temporal resolution of 1 min depended on the aerosol concentration sensitive to the 910 nm wavelength and the vertical aerosol profile represented the boundary layer structure. Thus, the boundary layer features, such as mixing layer and residual layer, could be visually determined from the time-height cross sections of the attenuated backscatter.
Model data. The Copernicus Atmosphere Monitoring Service (CAMS) reanalysis is a global reanalysis data set of atmospheric composition produced by the European Centre for Medium-Range Weather Forecasts (ECMWF). It consists of 3-dimensional time-consistent atmospheric composition fields 34 . Several satellite The grey open, black closed, and red closed circles indicate all the data (over the entire period), the measured/estimated values during nighttime (20:00-08:00 LST), and the measured/ estimated values for the long-range transport cases selected in this study, respectively.