A Study on Elevated Concentrations of Submicrometer Particles in an Urban Atmosphere

Mass concentrations of chemical constituents (organics, nitrate, sulfate, ammonium, chloride, and black carbon (BC)) and the number size distribution of submicrometer particles in the ambient atmosphere were continuously measured in urban Gwangju, Korea, during the Megacity Air Pollution Studies (MAPS)-Seoul campaign. Organics (9.1 μg/m3) were the most dominant species, followed by sulfate (4.7 μg/m3), nitrate (3.2 μg/m3), ammonium (2.6 μg/m3), and BC (1.3 μg/m3) in submicrometer particles (particulate matter less than 1 μm (PM1)). The potential source regions of the sulfate were located in the South and East regions of China and South and East regions of Korea, while local sources were responsible for the elevated BC concentration. Diurnal variation showed that concentrations of organics, nitrate, ammonium, chloride, and BC decreased with increasing mixing layer and wind speed (dilution effect), while sulfate and oxidized organics increased possibly due to their strong photochemical production in the afternoon. During the campaign, an elevated mass concentration of PM1 (PM1 event) and number concentration (nanoparticle formation (NPF) event) were observed (one PM1 event and nine NPF events out of 28 days). The PM1 event occurred with Western and Southwestern air masses with increasing sulfate and organics. Long-range transported aerosols and stagnant meteorological conditions favored the elevated mass concentration of submicrometer particles. Most of the NPF events took place between 10:00 and 14:00, and the particle growth rates after the initial nanoparticle formation were 7.2–11.0 nm/h. The times for increased concentration of nanoparticles and their growth were consistent with those for elevated sulfate and oxidized organics in submicrometer particles under strong photochemical activity.


Introduction
Fine particulate matter (PM) in the ambient atmosphere is of current interest due to its effects on climate change (radiation balance and cloud formation) [1] and human health [2][3][4].Due to multiple sources and formation pathways, chemical constituents in fine PM differ significantly among locations and times.Fine PM is directly emitted from various sources (primary aerosols) and/or produced by gas-to-particle conversion processes (secondary aerosols) and rapidly experiences the aging process (condensation, coagulation, oxidation, cloud processing, and so on), leading to different chemical and mixing states from the original particles.Additionally, the existence of ultrafine particles (<100 nm) in the ambient atmosphere have often been observed due to direct emissions from combustion sources and nanoparticle formation (NPF) with subsequent rapid growth [5][6][7].Due to the rapid dynamic process of fine PM, real-time measurements of their physical and chemical properties are essential to better understand their sources and formation pathways when elevated PM mass and number concentrations are observed.
Long-range transport (LTP) aerosols play an important role in the elevated mass concentration of submicrometer particles in East Asia [8][9][10].It has often been observed that the chemical constituents in PM less than 1 µm (PM 1 ) or 2.5 µm (PM 2.5 ) were dependent on the types of long-range transported air mass arriving at sites [7,8,11,12].Sulfate-dominant and organic-dominant LTP events have been found with different air mass types in Korea [8].A significant number of elevated PM 2.5 concentrations in Korea during winter and spring seasons have been explained by long-range transported aerosols [13].
Nanoparticle formation significantly contributes to an enhanced number concentration of submicrometer particles in the ambient atmosphere [12,14].NPF events have often been observed in many locations worldwide including East Asia [11,12,15].It has been reported that the low surface area of pre-existing particles and high growth of initial clusters favors NPF in the ambient atmosphere [16].
In this study, mass concentrations of organics, nitrate, sulfate, ammonium, chloride, and black carbon (BC) of submicrometer particles and the number size distribution of particles from 10 nm to 25 µm in an urban atmosphere (Gwangju, Korea) were continuously measured during the period of the Megacity Air Pollution Studies (MAPS)-Seoul 2015 campaign (18 May 2015-14 June 2015).Air mass backward trajectory analysis, residence time analysis (RTA), and potential source contribution function (PSCF) methods were applied to examine the effects of long-range transported aerosols on elevated PM 1 and their chemical constituents, as well as to infer possible geographic source regions of the long-range transported aerosols.Number size distribution data were used to examine NPF in relation to the chemical composition of submicrometer particles and meteorological parameters.

Experimental Section
The ambient aerosol measurements were conducted at an urban Gwangju site during the period of the MAPS-Seoul 2015 campaign (18 May 2015-14 June 2015).As shown in Figure 1, the sampling site is under the influence of multiple local aerosol sources such as residential/commercial areas (~0.6 km away from the site), highways (~1.5 km away from the site), a small industrial area (~1.5 km away from the site), and an agricultural area (~0.8 km away from the site) [5,17].The site is also influenced by short-range transported aerosols from a petrochemical industrial complex and steel works located ~90 km Southeast of the site and long-range transported aerosols from China, North Korea, and Mongolia in East Asia.Table 1 summarizes the measured parameters at the sampling site and nearby national monitoring sites.Chemical constituents (organics, nitrate, sulfate, ammonium, and chloride) of non-refractory submicrometer particles were measured in real time using an aerosol mass spectrometer (AMS) (Aerodyne, Billerica, MA, USA).A detailed description of the AMS is described elsewhere [13,[17][18][19].
Briefly, the ambient air was sampled into the AMS through a 100-µm diameter critical orifice with a flow rate of 0.07 liters per minute (lpm).The aerodynamic lens system focused submicrometer particles into the vaporization and ionization region.Particles were vaporized on the V-shaped tip of the vaporizer at the temperature of 600 • C, then the vapors were ionized by electron impaction (70 eV).The produced ions were detected with a quadrupole mass spectrometer.The collection efficiency of ambient particles in the AMS was assumed to be 50% [20].Calibrations of the electron multiplier gain value, ionization efficiency, and particle size of the AMS were conducted using the previous methods in Jimenez et al. (2003) [21].Ammonium nitrate particles (300 nm) were used for the ionization efficiency calibration, which was conducted every week.Organic aerosols (OAs) were further classified into hydrocarbon-like OAs (HOAs) and oxygenated OAs (OOAs) by positive matrix factorization (PMF) analysis [13,22].The HOAs are related to primary OAs produced from combustion sources (the HOAs are often correlated with combustion tracers such as CO, elemental carbon (EC), and NO x ), while the OOAs are related to secondary organic aerosols (SOAs) produced by photochemical activity [23].The number of factors was determined based on dissimilarities in the average mass spectra (m/z 43, 44, 55, and 57), Q/Q exp , and temporal variation of the factors [22,24].The concentration of BC was measured with an Aethalometer (AE51, Magee Scientific, USA) every 5 min with a flow rate of 0.2 lpm.The Aethalometer measures BC concentration by measuring the attenuation of the transmission of light through the collected samples on the filter [25].The filter was changed every day [26].The number size distribution of particles from 10 nm to 25 µm was measured using a NanoScan-scanning mobility particle sizer (SMPS) (3910, TSI, USA), a differential mobility particle sizer (DMPS) consisting of a differential mobility analyzer (DMA) (3081, TSI, USA), a condensation particle counter (CPC) (TSI 3022a, TSI, USA), and an optical particle counter (OPC) (1.108, Grimm, Germany).The scan times for the NanoScan-SMPS, DMPS, and OPC were 1 min, 5 min, and 1 min, respectively.The number size distributions from NanoScan-SMPS (10-133 nm) and DMPS (146-608 nm) were combined together.The combined size distribution was shown in this paper.Size-dependent transport losses of particles through tubes in the NanoScan-SMPS or DMPS were corrected theoretically [27].The mode diameter was estimated by fitting the number size distribution data into a Gaussian distribution.The number of modes (mono-modal or bi-modal) was selected from the best fit results.
The PM 2.5 mass concentration, gaseous components, and meteorological parameters were obtained from the nearest air quality monitoring network stations.Temperature, wind speed, wind direction, precipitation, solar radiation, and relative humidity (RH) were obtained from an automatic weather station (AWS) and the automated surface observation system (ASOS) operated by the Korea Meteorology Administration (KMA).
A 48-h air mass backward trajectory analysis was conducted with 1-h intervals for each trajectory using the NOAA Hybrid Single-Particle Lagrangian Integrated Trajectory model (HYSPLIT, version 4) with meteorological field data (GDAS1) [28].A residence time analysis (RTA) method was used to infer the dominant region where air mass passed over, affecting a receptor site.The air mass trajectory consisted of 48 endpoints.The RTA value for each grid cell (0.2 • latitude × 0.2 • longitude) was calculated using the number of endpoints passing over the grid cell (i.e., the number of endpoints is the residence time of the air mass in each grid cell).Additionally, a PSCF method was applied to find the source regions of the chemical constituents in PM 1 [29][30][31] (i.e., elevated concentrations of chemical constituents at a receptor site can be related to the air mass trajectory).The PSCF value for each grid cell was calculated using the ratio of the number of endpoints passing over the grid cell when the concentration of specific chemical constituents was within the upper 35% level to the total number of endpoints passing over the grid.A weight function was applied to place less weight on the cells with a low number of endpoints.Furthermore, aerosol optical thickness (AOT) data were obtained from MODIS sensors aboard the Aqua and Terra satellites [32].

Results and Discussion
Figure 2a shows the time series of meteorological data (wind speed, wind direction, RH, precipitation, solar radiation, and temperature) and concentration of gaseous species (CO, O 3 , SO 2 , and NO 2 ) during the sampling period.The average temperature and RH were 20.6 • C and 56.2%, respectively.The average mixing ratio of O 3 was 47.6 ppb, and O 3 had a distinct diurnal pattern with correlations with temperature (r = 0.78) and solar radiation (r = 0.56).The average mixing ratios of NO 2 and SO 2 were 0.018 ppm and 0.003 ppm, respectively.The elevated mixing ratio of SO 2 was observed from 7 June 2015 to 8 June 2015 and 10 June 2015 to 11 June 2015.
Temporal variations of PM 2.5 , PM 1 , and the number size distribution are shown in Figure 2b.The PM 1 was calculated as the summation of organics, nitrate, sulfate, ammonium, chloride, and BC concentrations.In addition to the measured chemical constituents (organics, nitrate, sulfate, ammonium, chloride, and BC), other chemical components from dust and sea salts may exist, although their contribution to PM 1 should be small.The average mass concentrations of PM 2.5 and PM 1 were 30.3 ± 13.5 µg/m 3 and 20.8 ± 10.9 µg/m 3 , respectively, as indicated in Table 2.The PM 1 explained approximately 68.5% of PM 2.5 with a correlation coefficient of 0.70 between PM 1 and PM 2.5 .The maximum concentrations of PM 2.5 and PM 1 were 101.0 and 77.1 µg/m 3 , respectively.The daily PM 2.5 level was higher than the WHO standard value (25 µg/m 3 ) on 17 days out of 28 days.One strong PM 1 event where daily average PM 1 was higher than 30 µg/m 3 was found during 11 June 2015 to 13 June 2015.PM 1 increased more than 2.1 times when compared to the average concentration during the whole sampling period.More on the characteristics of the PM 1 event will be discussed later.The organics were the most dominant species in PM 1 .The average mass concentrations of organics, sulfate, nitrate, ammonium, and BC were 9.1 µg/m 3 , 4.7 µg/m 3 , 3.2 µg/m 3 , 2.6 µg/m 3 , and 1.3 µg/m 3 , and their fractions in PM 1 were 43.7%, 22.5%, 15.4%, 12.6%, and 6%, respectively.The fractions were similar to those at other urban sites [33] and at the same sampling site in 2011, 2012, and 2013 [8,13,17].
As shown in Figure 2b, nine NPF events were found in 28 measurement days.The NPF event here was defined when the number of particles less than 50 nm increased significantly with subsequent growth.The growth rate varied in the range of 7.2-11.0nm/hour.The growth rate was calculated from the change of the mode diameter during particle growth time (the slope in the linear curve between the measured mode diameter and time represents the growth rate) [6,15,34].The number concentration of particles was not correlated to the PM 2.5 mass concentration.When the NPF event occurred, the mass concentration of pre-existing particles (i.e., PM 2.5 ) was low, suggesting that the low surface area of pre-existing particles favored NPF.More discussion on NPF will be provided later.To determine any relationship among the PM 1 mass concentration, chemical constituents in PM 1 , and air mass, air mass backward trajectories (48 h) were clustered using the K-means algorithm [35].As shown in Figure 3, five major clusters were found: Western (cluster 1), Northern (cluster 2), Northwestern (cluster 3), Southeastern (cluster 4), and Southwestern (cluster 5) trajectories explaining 15%, 13%, 34%, 23%, and 14% of the total air masses, respectively.With cluster one, the highest PM 1 (28.6 µg/m 3 ) was found with dominating species of both organics (41.9%) and sulfate (24.4%).Cluster one passed over an industrial region in the Shandon Peninsula and the Yangtze River Delta (YRD) in China (~664 km west of the site) where elevated SO 2 emissions and AOT are often observed [36,37].The PM 2.5 mass concentration also showed a similar pattern to that of PM 1 , as shown in Figure 3.With cluster four, which had the most stagnant air condition (the air mass speed was as low as 2.8 km/h and the local wind speed was as low as 1.0 m/s), an elevated PM 1 concentration was also observed.It is possible that the PM 1 produced from local sources was not diluted efficiently at the stagnant condition, leading to the elevated PM 1 .Additionally, the cluster four air mass passed over a petrochemical industrial complex and steel works located ~90 km Southeast of the site.Our data suggest that long-range transported aerosols and stagnant conditions played an important role in the increased PM 1 and PM 2.5 concentrations during the sampling period.The PSCF results for sulfate, organics, and BC in PM 1 are shown in Figure 4. Elevated mass concentrations of chemical constituents at a receptor site can be directly related to the air mass trajectory and source regions using the PSCF method.The potential source regions of sulfate were located in the South and East regions of China and South and East regions of Korea, suggesting that long-range transported aerosols played important roles in the elevated concentrations of sulfate at the sampling site.A lower number of source regions with a long distance were found for the elevated organics when compared to the sulfate.The potential source region for the BC was not clearly identified outside of Korea, suggesting that local sources may be responsible for the elevated BC concentration.The average diurnal variations of the chemical constituents in PM 1 are shown in Figure 5a.Organics, nitrate, ammonium, chloride, and BC showed a similar diurnal pattern with increased concentrations in the morning and evening, and low concentrations in the afternoon.This behavior was related to the diurnal variation of mixing height, which led to changes in the wind speed and temperature, as shown in Figure 5b [38].As the mixing layer increased in the afternoon, the concentrations of organics, nitrate, ammonium, chloride, and BC decreased significantly (i.e., the dilution effect).However, the OOAs decreased a little and sulfate increased, overcoming the mixing layer effect.This occurred as the strong photochemical activity in the afternoon led to the production of significant amounts of sulfate and OOAs.
During the PM 1 event (significant increase in PM 1 mass concentration) and NPF event (significant increase in the number concentration of submicrometer particles) days, the air mass retention times resulting from the RTA analysis were compared, as shown in Figure 6.During the PM 1 event (11 June 2015 and 12 June 2015), the Western and Southwestern air mass influenced the site.Both sulfate and organics increased significantly during the PM 1 event.Sulfate and organics were dominant on the first day of the PM 1 event (11 June 2015), and organics became more dominant than sulfate on the second day (12 June 2015) with the change in air mass and wind direction as shown in Figure 6a.We believe that the high concentration of PM 1 at the receptor site may be caused by long-range transported aerosols.The locations where the air mass passed over included major industrial complexes in China and Korea, possibly contributing to the elevated PM concentration.Satellite data from 12 June 2015 are shown in Figure 6b.The AOT was high in China (east region) and Korea (west region including the sampling site) when the PM 1 event occurred.The AOT was determined from the extinction of light in the vertical column.Both ground and satellite data suggested that long-range transported aerosols played an important role in the elevated PM 1 .The number of particles in the larger size range (650 nm-20 µm) increased significantly during the PM 1 event, as shown in Table 2. Significantly different air masses were found during the NPF events.NPF events occurred when the Western air mass minimally influenced the site as shown in Figure 6a.The Western air mass was related to the elevated PM 1 or PM 2.5 mass concentration, providing a high surface area of pre-existing particles, which possibly led to suppressing the NPF [39].The average surface area of particles during PM 1 event days was 1007 µm 2 /cm 3 .The surface area of particles during the NPF event was small as shown in Table S1 in the Supplementary Materials.During the NPF events, the time for the increased concentration of nanoparticles occurred from 10:00 to 14:00 with subsequent growth in the afternoon with strong solar radiation as shown in Figure 7 and Table S1 in the Supplementary Materials.Furthermore, this is consistent with the times for elevated sulfate and oxidized organics in submicrometer particles, as shown in Figure 7.Our data suggest that a low surface area of pre-existing particles and strong solar radiation favored NPF at the sampling site.Combustion can also contribute to the elevated number concentrations of particles less than 50 nm.However, typical combustion particles did not show significant growth, as we observed here (banana shape in Figure 7).Furthermore, the concentration of BC, one of the major combustion particles, did not increase significantly during NPF events.

Figure 1 .
Figure 1.Locations of the aerosol measurement site and national monitoring stations.

Figure 2 .
Figure 2. Time series of (a) meteorological data (wind speed, wind direction, RH, precipitation, solar radiation, and temperature) and concentrations of gaseous species (CO, O 3 , NO 2 , and SO 2 ), and (b) PM 2.5 and PM 1 mass concentrations, concentrations of organics, sulfate, nitrate, ammonium, chloride, BC, and number size distribution.

Figure 3 .
Figure 3. Concentrations of PM 1 and PM 2.5 with clustered air masses based on air mass backward trajectories (48 h).

Figure 4 .
Figure 4. Potential source contribution function PSCF results for sulfate, organics, and BC.

Figure 5 .
Figure 5.Diurnal variations of (a) organics, nitrate, sulfate, ammonium, OOA, HOA, chloride, and BC in PM 1 and (b) gases and meteorological data.The solid line indicates the averaged value, and the shaded area represents the standard deviation.

Figure 6 .
Figure 6.(a) Comparison of the residence times of air masses arriving at a receptor site during all sampling periods, the PM 1 event, and NPF event, and (b) AOT data obtained from MODIS sensors aboard the Aqua and Terra satellites with air mass backward trajectory (48 h) during the PM 1 event (12 June 2015).

Table 1 .
Summary of the measured parameters and instruments during the sampling period.
a BAM-1020 (Met One Instruments, USA) 18 May 2015-14 June 2015 Chemical constituents in PM 1 Sulfate, Nitrate, Ammonium, Chloride, Organics AMS (Aerodyne, USA) 18 May 2015-14 June 2015 BC Aethalometer (AE-51, Magee Scientific, USA) 18 May 2015-7 June 2015 a National air quality monitoring network station operated by Korea Environment Corporation.b National AWS observing station operated by the Korean Meteorological Administration.c National ASOS observing station operated by the Korean Meteorological Administration.

Table 2 .
Average values of the measured parameters during the whole sampling period and PM 1 event.