Regional characteristics of fine aerosol mass increase elucidated from long-term observations and KORUS-AQ campaign at a Northeast Asian background site

Urban areas in Northeast Asia have experienced heavy air pollution over the past few decades. Severe and persistent haze with high levels of PM_2.5 (particulate matter with aerodynamic diameter smaller than 2.5 mm) has been of major public concern due to its potential adverse effects on human health, such as cardiovascular and respiratory disease and lung cancer, which leads to increased risk of mortality (Atkinson et al., 2015; Lelieveld et al., 2015; Apte et al., 2018; Burnett et al., 2018). In Seoul, the capital of South Korea, the annual PM_2.5 concentration has remained as high as 20–30 μg m⁻³ over the past 10 years (Yeo et al., 2019), easily exceeding the national (15 μg m⁻³) and World Health Organization (WHO) (10 μg m⁻³) annual air quality standard. High-concentration PM_2.5 events tend to occur frequently in the cold months from November to March (Lim et al., 2012; Lim et al., 2014), when the daily average PM_2.5 concentrations often exceed the national air quality standard (35 μg m⁻³). Consistently, in China, 56% of regional pollution events occurred during the cold months in 2014–2017 (Li et al., 2020). The chemical composition of the cold-month high PM_2.5 is dominated by secondary inorganic aerosols (SIAs), namely SO₄⁻², NO₃⁻, and NH₄⁺ (e.g., Huang et al., 2014; Liu et al., 2018). In comparison, organic carbonaceous aerosol accounts for a large proportion in fine aerosol mass during warm months (e.g., Kim et al., 2018). Meanwhile, black carbon (BC) that is directly emitted to the atmosphere from combustions of biomass and fossil fuel, a strong climate forcer (Bond et al., 2013; Boucher et al., 2013), has been recently known that its small, insoluble, and inert nature facilitate its long-range transport across China during haze pollution events (Wang et al., 2016; Zheng et al., 2019). Co-emitted with various gaseous precursors (e.g., NO_x and SO₂) and particulates, BC undergoes physical evolution in size and mixing state upon transportation through the formation of coatings by non-BC materials after being emitted (Lim et al., 2018; Zanatta et al., 2018; Liu et al., 2019). The occurrence of high-concentration PM_2.5 events has become a regional characteristic of the Northeast Asia (Liu et al., 2018; Fan et al., 2020).

Long-term regulatory actions to control emissions have been implemented in South Korea as well as China. However, severe PM_2.5 events have been frequently encountered over the last few years in Seoul (Park et al., 2021). Current knowledge is insufficient to explain the complex processes that generate unexpectedly high PM_2.5 masses (e.g., An et al., 2019). Most of all, aerosol formation mechanisms and the interplay of physicochemical transformation mechanisms with meteorological conditions that are vital for the effective mitigation of haze pollution, have yet to be fully identified.

Theoretically, new particle formation (NPF) and subsequent growth occur via the gas-to-particle conversion of gaseous precursors, such as SO₂, NO_x, VOCs, and NH₃. This process, known as NPF, is primarily dependent on 2 parameters: (1) source strength of condensable gases and (2) pre-existing particles that act as a condensation sink (CS) (Kerminen et al., 2018). It was also found that a minimum in CS correlates with a maximum frequency of nucleation on a regional scale of Europe (Elser et al., 2016). However, in polluted cities of China with high levels of both precursors and CS, NPF occurs frequently, showing distinct features from pristine environments (Kulmala et al., 2017; Chu et al., 2019). For example, nanoparticles undergo rapid and continuous growth in size from smaller than 10 nm to larger than 100 nm in 1 day (Wiedensohler et al., 2009) or from smaller than 50 nm to larger than 200 nm in a couple of days (Guo et al., 2014), even during heavy pollution episodes. In comparison, the entire processes from nanoparticle formation to increase in fine aerosol mass have rarely been observed in ambient atmosphere including Seoul or background sites in South Korea (e.g., Kim et al., 2013) and may not be explained in the same manner throughout the Asian continent. Therefore, the conditions and mechanisms responsible for the rapid increase in PM_2.5 mass are still poorly understood.

Recently, meteorological conditions are thought to be closely related to persistent PM_2.5 haze pollution over China on a regional scale (Wang et al., 2014a; Wang et al., 2014b; Zhang et al., 2018; Wang et al., 2019). Strong correlations of daily PM_2.5 concentrations with several meteorological variables (local or climate variables) have been found and unambiguous spatial patterns indicate the importance of synoptic meteorology affecting haze developments (Cai et al., 2017; Leung et al., 2018). Among these, the dynamic evolution of planetary boundary layer (PBL) has drawn much attention due to its direct effects on PM_2.5 concentrations such as stable atmospheric stratification accumulating local emissions, positive feedback between boundary layer stability and pollution intensity, and downward transport of pollutants from the free atmosphere through PBL processes (e.g., Zhong et al., 2018; Li et al., 2020; Quan et al., 2020). For NPF, the combined effect of boundary layer dynamics and atmospheric chemistry on aerosol composition was emphasized (Song et al., 2010; Hao et al., 2018). As is commonly observed during several haze events, high relative humidity (RH) is believed to facilitate the hygroscopic growth of SIAs, thereby increasing PM_2.5 concentrations (Zhang et al., 2015; Wang et al., 2019). Recent studies have highlighted the role of aerosol liquid water content (ALWC) in the rapid formation of SIAs in Northeast Asia (Liu et al., 2017; Wu et al., 2018). ALWC is known to depend on RH and, mass and chemical composition of aerosols (Pilinis et al., 1989). At high RH, ALWC will increase uptake coefficients of N₂O₅, thereby enhancing the formation of particulate nitrate (Bertram and Thornton, 2009; Bertram et al., 2009). Recently, an isotope-based field study that shows wintertime NO₃⁻ concentration formed via the N₂O₅ pathway increased with ALWC in Seoul (Lim et al., 2022). It has already been suggested that ALWC promotes the aqueous conversion of SO₂ to SO₄⁻². However, a comprehensive and systematic understanding of particle formation/transformation and haze developments is hindered by the complexity of local to large-scale meteorology, local emissions versus transboundary transport of pollutants, and particle (trans)formation mechanisms. Therefore, long-term observations in the background region are essential to better understand the phenomena on a regional scale.

In this study, we thoroughly examined the size-separated aerosol number distributions in relation to the mass of PM_2.5 and its chemical composition, gaseous precursors, and meteorological characteristics observed in the East Asia outflow region for multiple years. Based on the results, we aimed to elucidate general characteristics of mass increase in fine aerosol at a regional site in Northeast Asia.

A suite of aerosol instruments was deployed to measure aerosol number concentrations, PM_2.5 mass concentrations and its chemical constituents, the mixing ratios of reactive gases, and meteorological parameters at the Jeju Air Quality Monitoring Research Center (33.21°N, 126.23°E, 600 m asl; hereinafter referred to “Aewol”) in Jeju Island, South Korea (Figure S1), from January 2013 to December 2016. This station has been served as the national background air pollution monitoring station without evidently significant local pollution sources.

The number concentrations of size-separated particles at diameter (Dp) ranging from 10.4 to 469.8 nm were measured every 30 min by the scanning mobility particle sizer (SMPS, model 3034, TSI, USA), and PM_2.5 mass concentrations were measured every hour by the beta attenuation mass monitor (BAM1020, MetOne Instrument Inc., OR, USA). For PM_2.5 chemical composition, water-soluble ions and organic and elemental carbon (OC and EC) were determined. Water-soluble ions were measured every hour by a Monitor for Aerosols & Gases in Ambient Air (MARGA, Metrohm, Switzerland). OC and EC concentrations were simultaneously determined with the thermal/optical transmittance method (NIOSH 870 protocol) by a Sunset Lab OCEC Analyzer (Sunset Laboratory Inc., OR, USA). Reactive gases including SO₂, NO₂, NH₃, and O₃ were continuously monitored, and their nominal detection limits were 0.1 ppbv, 0.1 ppbv, 0.1 ppbv, and 2 ppbv, respectively (An et al., 2015). These instruments were calibrated and checked up on a weekly, monthly, or annual basis according to the information provided by the manufacturer and guidelines established by the Ministry of Environment and National Institute of Environmental Research (National Institute of Environmental Research, 2021). Details about measurement methods can be found elsewhere (e.g., Shin et al., 2016; Park et al., 2018).

Meteorological variables including temperature, RH, wind speed and direction, SO₂, and cloud properties were measured at Gosan Climate Observatory (GCO, 33.17°N, 126.10°E) in Jeju Island by Korean Meteorological Administration (KMA, https://data.kma.go.kr/cmmn/main.do). In addition, a set of cloud fraction (low- and mid-level clouds; 1–10 in fraction) and the base height of cloud were measured with a ceilometer (CL31, Vaisala), which were incorporated into the measurements at Aewol. All measurements were assimilated into 1-h averages.

The Korea–U.S. Air Quality (KORUS-AQ) campaign provided a unique opportunity to observe distinctive events from May to June 2016. Along with measurements at Aewol, the physical properties of refractory BC, including number and mass concentrations, size distributions, and mixing state (coating thickness), were determined at GCO using the Single Particle Soot Photometer (SP2, Droplet Measurement Technology, Boulder, CO, USA) (Stephens et al., 2003; Schwarz et al., 2006; Moteki and Kondo, 2010).

SP2 utilizes a laser-induced incandescence technique to quantify the mass of a single refractory BC (Petzold et al., 2013). For individual detection of incandescence and scattering signals, the SP2 is equipped with 2 photomultiplier tubes and 2 avalanche photodiodes, each yielding high- and low-gain values. The incandescence detector was calibrated using well-characterized fullerene soot particles (Alfa Aesar; #FS12S011), and the scattering detector was calibrated with a spherical polystyrene latex size standard (PSL, refractive index (RI) = 1.59; Thermo Fisher Scientific, Waltham, MA, USA) (Baumgardner et al., 2012; Laborde et al., 2012). The detection range of a single-particle mass is 0.33–128 fg, corresponding to rBC mass equivalent diameters (D_rBC = 70–514 nm), assuming a void-free material density of 1.8 g cm⁻³ (Bond et al., 2006). To correct for the rBC mass outside the lower detection limit, the rBC mass size distribution was fitted with a lognormal function, and a correction was achieved by adding 10% of the total uncorrected rBC mass on average. An rBC mass larger than the upper detection limit was treated as the largest mass detectable, that is, 128 fg.

Basically, adopting the leading-edge-only approach (LEO approach; Gao et al., 2007; Laborde et al., 2012), the optical diameter of the rBC-containing particle, D_shell, was inferred from the leading-edge scattering signal using Mie theory with RI of 1.5 − 0i and 2.26 − 1.26i for the rBC coating material and rBC core, respectively, at an SP2 operating wavelength of 1064 nm. These RI values have been suggested in previous studies: 2.26 − 1.26i for bare or freshly emitted ambient rBC particles (Moteki and Kondo, 2010; Taylor et al., 2015) and 1.5 − 0i for inorganic salts and secondary organic aerosol (Toon et al., 1976; Schnaiter et al., 2003) have been used for similar LEO approaches (e.g., Schwarz et al., 2008; Laborde et al., 2013). The coating thickness of individual rBC particles was determined as (D_shell − D_rBC)/2. In this SP2, D_shell was detectable in the range of 170–500 nm. The proper range of D_rBC for LEO approach varies mainly depending on the instrumental setup of the SP2. To ensure the quality of LEO results, lower and upper bounds of D_rBC are selected by examining detectable optical diameter of small rBC core (here, ≥95%) and a saturation point of the low gain scattering detector (Taylor et al., 2015), respectively. In the present study, D_rBC for LEO approach was restricted to the rBC core with a D_rBC of 180–220 nm (Laborde et al., 2013; Zanatta et al., 2018). Coating was defined as a thickness larger than 10 nm, considering the systematic uncertainty involved in its estimation (Laborde et al., 2013). Measurements of rBC particle concentration, size distribution, and mixing state of rBC particles were averaged hourly, merged with the PM_2.5 chemical measurement data set, and used for subsequent discussion.

In addition, vertical profiles (up to 2,500 m) of size-segregated aerosol number concentrations and meteorological parameters were obtained by balloon measurements in Ansan as a satellite city of Seoul (33.17°N, 126.10°E, 70 m asl) (Figure S1). Balloon was launched at 11:26–11:45 on 20 May, and at 8:55–09:05 and 16:05–16:13 on 25 May (local time). Aerosol number concentrations were measured by lab-made instruments at Hanyang University including condensation particle counter (CPC) for particles larger than 3 nm and 2 optical particle counters (OPCs) for accumulation-mode particles (particle_{0.3–0.5 μm} and particle_{0.5–1.0 μm}) (Lee et al., 2014). The number concentration of nanoparticles (particle_{0.3 nm–0.5 μm}) was reported as difference between CPC and OPC. For CPC and OPCs, sampling rate and counting resolution was 0.125 L min⁻¹ and 1 s, respectively. The results of vertical measurements using these instruments can be found elsewhere (Querol et al., 2017; Querol et al., 2018). The vertical profiles of meteorological variables and aerosol number concentrations were analyzed along with the physical properties of refractory BC (Section 2.2). To aid in data interpretation, the ALWC was calculated in “forward mode” using the ISORROPIA-II thermodynamic model (Fountoukis and Nenes, 2007).

The PM_2.5 concentration of Aewol was much lower as a level of background site than that of Seoul (Figure 1). In Seoul, the annual mean PM_2.5 concentration ranged from 27.3 μg m⁻³ to 38.3 μg m⁻³ during 2013–2016, exceeding the national annual air quality standard (15 μg m⁻³) every year. For the same period, the annual PM_2.5 concentration was 13.9–19.7 μg m⁻³ at Aewol. The daily mean PM_2.5 concentration violated the air quality standard (35 μg m⁻³) by 33% of the entire period in Seoul. It is also worth noting that 9% of the entire period exceeded the standard at the background site, Aewol. Despite the difference in PM_2.5 level, the seasonal variation was similar at the 2 sites with high levels in the cold months, demonstrating that the enhancement of wintertime PM_2.5 is a significant regional issue over the East Asia (Liu et al., 2018; Fan et al., 2020; Li et al., 2020).

Continuous measurements of size-separated particle number concentrations, spanning 4 years from 2013 to 2016, were analyzed using an empirical orthogonal function (EOF) method, in which the loading vector was constrained to be periodic with a 24-h period (e.g., Kim et al., 2014). As a result, the first and second mode accounted for a significant fraction of 55% and 9%, respectively, while each of the other modes explained less than 4% of the total variance. The first and second modes, which captured the largest variability, contrasted sharply in the diurnal variations in particle number and size (Figure 2). The difference between the 2 modes is explicitly demonstrated when comparing the average number–size distributions of ambient particles for the top 10% days with the highest principal component (PC) scores for each mode (Figure S2a). Thereinafter, these cases were considered to be representative of the 2 modes and referred to as “EOF1” and “EOF2,” respectively. Out of the 4-year measurement period, each 143 days were tagged as EOF1 and EOF2, respectively. While EOF1 frequently occurred in May and October, the monthly frequency of EOF2 was high in March and June (Figure S2b). In EOF2, the mode of number, surface area, and volume was larger in size, compared to EOF1 (Figure S3). Using these particle number distributions, CS was calculated for the entire range of diameters from 10 nm to 470 nm (Text S1) (Kulmala and Kerminen, 2008). In the following section, particle number–size distributions and levels of CS and gaseous precursors will be compared between EOF1 and EOF2.

In EOF1, the variation of the number concentration corresponds to what used to be observed during the NPF and growth event. The nucleation- and Aitken-mode particles dominated number concentrations with the maximum number concentration greater than 10⁴ cm⁻³ at approximately 10–30 nm around noon (Figure 2c). They grew in size with a mean GR_10–50 of 2.1 ± 1.6 nm h⁻¹ (growth rate [GR] of particles at 10–50 nm in diameter; Table 1 and Text S1), which was close to the lower bound of the GR observed across China (1–15 nm h⁻¹; Chu et al., 2019, and references therein) and within the range of the global GR (1–10 nm h⁻¹; Nieminen et al., 2018, and references therein). In this case, the mean mixing ratios of gaseous precursors, such as SO₂ and NO_x, were higher by 45% and 14%, respectively, relative to the mean of the entire period (Table 1). In contrast, CS_{100–470 nm} (1.3 ± 0.9 × 10⁻² s⁻¹) (Table 1 and Text S1) was relatively lower than the mean. Thus, EOF1 was characterized by the dominant nucleation- and Aitken-mode particles and relatively high gaseous precursors and low CS, which clearly represents an NPF or at least a nanoparticle burst.

Contrary to EOF1, EOF2 was dominated by the accumulation-mode particles throughout the day (Figure 2d). As the number concentration of accumulation-mode particles increased in the afternoon through the evening, the maximum concentration reached 5,000 cm⁻³ at 100–130 nm; in addition, the mode of the volume concentration was found to be at larger sizes over 300 nm (Figure S3). In this case, CS_{100–470 nm} (1.8 ± 1.0 × 10⁻² s⁻¹) was much higher but levels of gaseous precursors were comparable to EOF1 (Table 1). The large surface areas of pre-existing particles would strongly suppress NPF, as newly formed clusters are scavenged by pre-existing particles before reaching observable sizes of 3-nm diameters (Kulmala and Kerminen, 2008; Kulmala et al., 2017). The most evident difference between the 2 modes was the higher mass and ionic constituents of PM_2.5 in EOF2 compared to EOF1 (Table 1).

The 2 EOF modes are supposed to represent physically independent phenomena. If EOF1 characterizes the pattern of increase in number concentration by nucleation- and Aitken-mode particles, it is clear that EOF2 diagnoses the case where the number concentration of accumulation-mode particles increased noticeably in the absence of such small particles. However, EOF2 was rarely recognized as an episode because it does not exhibit as prominent evolution in particle number as EOF1.

The chemical properties of the 2 modes were clearly distinguished in the particle phase, including size mode, CS level, and PM_2.5 chemical compositions, but not in the gaseous precursor. Since EOF1 is much better understood as an intrinsic phenomenon for particle formation, the meteorological conditions were further delved into to elucidate EOF2 compared to EOF1.

In accordance with what has been reported previously (Dall’Osto et al., 2018; Chu et al., 2019), aerosol number–size distributions like EOF1 was frequently observed on sunny and clear days in the study region, where it was situated under a high-pressure system after frontal passage during spring and fall (Figures S2b and S4a). Figure 3 compares the monthly characteristics of the 2 modes, including the frequency of occurrence, levels of meteorological variables, and PM_2.5 mass concentration. The latter two are given as the difference from the monthly mean. EOF1 was most evident with much lower cloud coverage (down to −60%) and much higher cloud base height (CBH, up to +800 m) than the monthly average (20–60% and 916–1,289 m, respectively) (Figure 3), leading to more solar radiation (SR) incoming to the surface (Table 1). In this period, cold and dry northerly winds prevailed (Figures 3 and S5) and precursor gases and PM_2.5 concentrations were low compared to the mean over the entire period due to a limited contact with urban sources. Of precursor gases, SO₂ mixing ratio was slightly increased in EOF1 (0–111%; 0.3–2 ppbv; Figure S6) probably from upwind sources in the continent (Wu et al., 2007; Liang et al., 2016; Tian et al., 2019). Even in March, September, and October, when both frequency and amplitudes of EOF1 was strongest (Figures S2 and 3), SO₂ showed only a marginal increase (0–6%; 0.3–1.0 ppbv). Given that the particle formation has been observed to strongly depend on the abundance of sulfuric acid (H₂SO₄) (Weber et al., 1996; Sihto et al., 2006; Hirsikko et al., 2011), our result of long-term measurements suggests that readily condensable gas concentration was somewhat insensitive to local precursor levels. In other words, the oxidation of SO₂ to H₂SO₄ is likely to be decoupled from nucleation or promoted by unidentified processes. As observed ubiquitously at midday on a regional scale, it is not surprising that NPF and subsequent growth events are sensitive to synoptic meteorological conditions and daily evolution of boundary layers, especially in East Asia (Takegawa et al., 2020).

The mean PM_2.5 concentration of EOF2 was 26.5 ± 21.3 μg m⁻³, being higher by 40% than EOF1 (19.1± 15.0 μg m⁻³). The PM_2.5 mass was dominated by SIA, with a mean contribution of 41% to the mass (20% SO₄²⁻, 10% NO₃⁻, 10% NH₄⁺) (Table 1). The difference in PM_2.5 mass between the 2 modes contrasts more sharply when compared, as deviations from the monthly mean (Figure 3g). PM_2.5 concentrations were 14–88% higher in EOF2 than in EOF1, and the difference was 55–88% (35–39 μg m⁻³ in monthly concentration for EOF2) in cold months. Of the days exceeding the daily Korean air quality standard (35 μg m⁻³) during the 4-year period, 29% (10%) belong to EOF2 (EOF1). Note that in the previous section, EOF2 was characterized by an increased number of accumulation mode particles. Therefore, it is reasonable to argue that EOF2 represents a distinct episode of elevated PM_2.5 mass particularly driven by high SIA mass fractions.

Contrary to EOF1, EOF2 occurred on warm, humid, and cloudy days under low CBH (Figure 3) and prevailing southwesterly winds blowing through the Yellow Sea (Figure S4b). The 2 modes were most clearly distinguished by cloud cover and CBH. Besides chemical mechanisms potentially operative in clouds, a mid-low CBH can be indicative of mixed layer height (MLH) particularly in the afternoon (Seidel et al., 2010; Huang et al., 2018). Hourly averaged CBH exhibits a clear diurnal variation similar to the typical MLH (Figure S7) and thus, CBH was used here as a surrogate for MLH. Then, the significantly lower CBH of EOF2 than EOF1 indicates that the boundary layer of EOF2 was much shallower than that of EOF1. A larger cloud cover was consistent with reduced SR reaching the surface in EOF2 (346 ± 281 W m⁻² and 257 ± 239 W m⁻² for EOF1 and EOF2, respectively: Table 1). However, like EOF1, PM_2.5 chemical constituents were completely neutralized (equivalent ratio of ([SO₄²⁻] + [NO₃⁻])/[NH₄⁺] = 1) with significantly elevated sulfate and nitrate oxidation ratios (SOR and NOR) (Table 1). Compared to the much higher SIA of EOF2 than EOF1, the difference in carbonaceous composition between the 2 modes was not significant (Table 1). For the 2 EOF modes, the aerosol number–size distribution was found to be closely related to the chemical composition of PM_2.5, where EOF2 was clearly distinct from EOF1. Therefore, it is highly plausible that the unique evolution of accumulation-mode particles in EOF2 was strongly associated with the increase in SIA in humid and shallow boundary layer, which eventually resulted in the PM_2.5 mass increase.

In this study, the number–size distribution of submicron particles evolved dynamically depending on synoptic weather conditions, resulting in 2 distinct modes. EOF1 was revealed in daily evolution of particles, explicitly indicating NPF and growth events. The other mode, EOF2, was linked with an increase in PM_2.5. Recent studies have shown that individual haze pollution episodes often took place over several days (Yang et al., 2017; Tian et al., 2019; Shi et al., 2020). Therefore, changes in the key chemical and meteorological variables of the 2 modes that occurred in March over 4 years were examined over a course of 7 days including 3 days before (d − 3 ∼ d − 1) and after (d + 1 ∼ d + 3) EOF day (d) (Figure 4). Note that in March of the year, both EOF1 and EOF2 occurred frequently, and the PM_2.5 mass increase of EOF2 was also large (Figure 3).

In EOF2, cloud cover and CBH decreased slightly, but wind speed increased (Figure 4a–c) on day d. Interestingly, SR steadily decreased at a rate of −14.1 W m⁻² day⁻¹, reaching a minimum on day d (Figure 4d), resulting in an inverse correlation with PM_2.5 mass concentration (Figure 4e). As shown above, PM_2.5 mass concentration was significantly higher on day d with a gradual increase at a rate of 5.6 μg m⁻³ day⁻¹ and decrease after reaching a maximum. These progressive changes in PM_2.5, SR, and cloud properties highlight the role of PM_2.5 in reducing SR; downwelling SR was thought to be reduced by thick aerosol layer, which was dependent on aerosol size distributions and chemical composition. This finding is in agreement with a recently proposed mechanism that explains the intensified haze pollution caused by a positive feedback between increase in aerosol concentration, reduction in surface SR, and stability of boundary layer (Ding et al., 2016; Yang et al., 2016). In Beijing, this feedback effect was reported to be 15.42 W m⁻² km⁻¹, given as a sensitivity of daily downwelling SR to visibility, where a 1 km reduction in visibility was equivalent to a 10 μg m⁻³ increase in PM_2.5 concentration (Yang et al., 2016). In this study, the combined effects of thick aerosol layer and cloud were estimated to be approximately 24 W m⁻² per 10 μg m⁻³ for the 4 days of the EOF2 episode on average.

In Figure 4e and f, PM_2.5 mass and ALWC tend to increase simultaneously. It is noteworthy that ALWC in EOF2 continued to increase until d + 1, with the greatest increase during d − 2 and d − 1 (Figure 4f). In comparison, PM_2.5 mass increased the most between d − 1 and d. The increases of SO₄⁻² and NO₃⁻, which are the major constituents of PM_2.5, were similar to those of AWLC and PM_2.5, respectively (Figure 4g). While SO₄⁻² steadily increased by 4.05 μg m⁻³ from d − 3 to d, the NO₃⁻ increase of 3.11 μg m⁻³ occurred between d − 2 and d. After d, these 2 species decreased at similar rates. During this 7-day period, RH remained high between 63% and 67% and was not correlated with ALWC. These results, therefore, indicate that SIA played a crucial role in determining ALWC, leading to a rapid increase in PM_2.5 mass. In humid atmospheric conditions with sufficient precursor gas, the conversion of water-soluble gas to particles is promoted, increasing the ALWC (Zhou et al., 2011). The increased ALWC then shifts the partitioning of soluble gas into the particulate phase, thereby constituting a positive feedback, as reported in previous studies (Hodas et al., 2014; Xue et al., 2014; Wu et al., 2018). On d + 1, a sharp decrease in PM_2.5 with an increase in AWLC is likely to reflect scavenging of particles by hydrometeors such as mist or fog at high RH. The role of ALWC was also evident in the EOF1 event.

In EOF1, ALWC exhibited pronounced changes over 7 days with a minimum on day d. This type of variation was consistent with changes in RH, decreasing from 74% to 65% between d − 3 and d and increasing back to 72% on d + 3. In comparison, PM_2.5 concentrations varied slightly, but dropped to a minimum on day d, when cloud cover and SR showed a clear minimum (1.5 ± 2.5) and maximum (413 ± 85 W m⁻²), respectively. As is already known, more SR boosts not only photochemical reactions but also deep vertical mixing of the boundary layer, consequently hindering SIA formation through diluting existing particles but promoting NPF and growth.

The progressive changes in meteorological and atmospheric chemical factors of the 2 EOF modes demonstrate that their interplay created the 2 distinct phenomena over a course of several days. This tight coupling allows nanoparticle formation and growth, represented by EOF1, to take place under well-constrained conditions as a stand-alone phenomenon. Likewise, EOF2 reflects the increase in fine aerosol mass as an episodic event occurring on a regional scale. Therefore, in the study region, EOF2 is an instance in which an increased number of accumulation-mode particles resulted in an increase in fine aerosol mass, whereas EOF1 is an event in which new particles are generated by a burst and dynamic evolution of nucleation–Aiken mode particles.

In May–June 2016, the KORUS-AQ campaign was run on multiple platforms and the variety sets of measurements were made in the Korean peninsula and Northeast Asia outflow region (e.g., Crawford et al., 2021). Of several distinct episodes that occurred under distinct synoptic circulation (Peterson et al., 2019), “persistent anticyclone” and “low-level transport and haze development” episodes correspond to EOF1 and EOF2 mode, respectively, in this study. The PM_2.5 mass concentration increased in Aewol, up to 63 μg m⁻³ (Table 2), as well as Seoul during “transport/haze” period (Kim et al., 2018; Jordan et al., 2020), showing a regional phenomenon of the haze development. In Aewol, the aerosol number–size distributions were obviously different during these 2 periods (Figure 5). Hereafter, the 2 distinct periods during the campaign referred to as “persistent anticyclone” and “transport/haze” periods, respectively, which were compared for meteorological and chemical properties in subsequent sections (Table 2).

The high surface temperature resulted in a deep mixing layer on May 20 of “persistent anticyclone” period (Figure 6a). The particle number concentrations at all size ranges were low over the altitude, explained by dilution of preexisting particles by vertical mixing. In contrast, in the morning on May 25 (“transport/haze” period), the mixed layer was shallow with a temperature inversion layer at about 1,000 m (top gray box in Figure 6b). The morning profile illustrates visibly high concentrations of the particle_{3.5 nm–0.3 μm} (about 10⁴ cm⁻³) both at surface and at around the inversion (gray boxes in Figure 6h). The high concentration near the surface is attributed to fresh emissions in the morning. On the other hand, high number of particle_{3.5 nm–0.3 μm} at around the inversion is similar to the results of previous studies in which a large number of ultrafine particles were commonly found above the cloud as a result of an enhanced photochemical reaction (Clarke, 1993). In the afternoon, the thin layer laden with large amount of particle_{3.5 nm–0.3 μm} disappeared, but the number of accumulation mode particles (particle_{0.3–0.5 μm} and particle_{0.5–1.0 μm}) increased remarkably, particularly at altitudes between 1,000–1,500 m (Figure 6i). It is evident for the evolution of nucleation- and Aitken-mode particles existing in the inversion layer, as the PBL developed.

In urban areas with heavy local emissions, a strong inversion forces the local circulation to be confined to the atmospheric boundary layer under undesirable ventilation, thereby acting as a lid; thus, pollutants are accumulated within the boundary layer (e.g., Ye et al., 2016; Ning et al., 2018). In contrast to the heavily polluted sites, in the background or outflow sites, the significant elevation in aerosol concentration during the day is believed to occur primarily by the intrusion of pollutants carried from the upper atmospheric layer, for example, residual layer formed at night or free troposphere, upon the expansion of the PBL. Evidently, the concentrations of PM_2.5 and all gaseous precursors increased during the daytime when the PBL expanded (Figure S7) and the wind direction was shifted to a westerly flow (Figure S5). Recently, it was suggested that pollutants were carried from the upstream areas in the upper layer by westerly winds, remained above the nocturnal boundary layer, and entrained into the PBL during the day; this led to a high PM₁₀ concentration at the surface of Seoul (Lee et al., 2019). The PM_2.5 haze pollution was accompanied with a strong temperature inversion in downwind regions of polluted Northeast China (Shi et al., 2020). Therefore, these results indicate that the rapid increase in PM_2.5 mass in Northeast Asia could be closely coupled with the development of the PBL.

In more detail, the vertical profiles indicate that the particle number concentrations correlated with changes in ambient RH (Figure 6). Compared to the “persistent anticyclone” period with a deep vertical mixing and dilution, “transport/haze” period was characterized by discernable layers of the vertical structure of meteorological variables including RH, temperature, and wind direction. Such a shallow boundary layer of high RH is favorable for gas-to-particle conversion and hygroscopic growth of particles, thereby greatly affecting the concentration and size distribution of particles. We considered the deliquescence relative humidity (DRH) of inorganic salts, such as NH₄NO₃ and (NH₄)₂SO₄, as a crucial factor for the rapid particle growth, because the SIA was the predominant constituent of PM_2.5 observed during the KORUS-AQ period (Jordan et al., 2020) as well as in Aewol for 4 years. In the morning, the burst of particle_{3.5 nm–0.3 μm} above 10⁴ cm⁻³ occurred at altitudes with RH below the DRHs of both (NH₄)₂SO₄ and NH₄NO₃ (Figure 6b and h). This layered structure became uniform in the afternoon with the expansion the PBL coinciding with elevated RH around the DRH of both (NH₄)₂SO₄ and NH₄NO₃, resulting in a noticeable reduction of particle_{3.5 nm–0.3 μm} and an increase of particle_{0.3–0.5 μm} and particle_{0.5–1.0 μm} in number concentration at altitudes between 1,000 m and 1,500 m (Figure 6c and h). It is likely that the mixing through inversion layer caused hygroscopic growth of existing nucleation- and Aitken-mode particles, which led to an increase in the number of accumulation-mode particles where the water-soluble vapors readily condensed, adding mass to the particles.

Therefore, the vertical profiles of aerosol number concentrations and their evolution during “transport and haze development” period suggest that (1) the entrainment of nanoparticles from the upper layer into the shallow and humid boundary layer with locally emitted or transported precursors and condensable gases, (2) the evolution of nanoparticles to accumulation-mode particles by increase in size via hygroscopic growth and mass via gas-to-particle conversion, and (3) consequently, the contribution of accumulation-mode particles to increase in PM_2.5 mass. Given that the synoptic circulation is the main deriver determining the level of atmospheric constituents over the Korean peninsula (Peterson et al., 2019; Jordan et al., 2020), this process would have been active for PM_2.5 mass increase throughout EOF2 periods.

In addition to the evolution of nanoparticles, the role of accumulation-mode particles in increasing the aerosol mass was highlighted above. Among aerosol particles, BC is of particular interest, as bare BC particle is insoluble and small, generally peaking below 100 nm in number and thus it potentially provides a chemically inert surface on which volatile substances can condense to form a coating and further grow on. In terms of climate effect, BC has been well established as the significant warming agent (Bond et al., 2013). To assess a role of BC particle in aerosol transformation, the mass and mixing state of rBC particles were determined at GCO in Jeju and compared between the 2 periods of the KORUS-AQ campaign (“persistent anticyclone” and “transport/haze”) (Table 2; Figures 5f and 7).

Figure 7a illustrates variations of D_shell (optical diameter of rBC-containing particle in a core-shell configuration with rBC core restricted to 180–220 nm; Section 2.1) as a function of PM_2.5 concentration for the entire campaign period. D_shell increased almost linearly with PM_2.5 mass concentration, reaching 300–400 nm on average at PM_2.5 higher than 40 μg m⁻³. This reveals a close relationship between rBC coating thickness and PM_2.5 mass enhancements. In Figure 7b, volume concentration of particles_{200–470 nm} was also elevated simultaneously with PM_2.5 mass concentrations, and rBC coating was obviously thicker (48 ± 39 nm) during “transport/haze” period when the PM_2.5 mass concentration was highly elevated over the Korean peninsula (Jordan et al., 2020). These results demonstrate that rBC-containing particles with thick coatings are truly representative of accumulation-mode particles in terms of size particularly during “transport/haze” period. Together with the enhanced SIA concentrations and the thicker coating of rBC (Figure 7b and c), ALWC was as high as 10.3 ± 8.2 μg m⁻³, accounting for 30% of the dry PM_2.5 mass (Figure 7c; Table 2). Given that the RH remained high at greater than 60% during the day and up to 90% at night (Figure 7b), hygroscopic growth of NH₄NO₃ and (NH₄)₂SO₄ by uptake of water was expected to be promoted, leading to a large increase in ALWC. SOR was almost linearly increased with RH (Figure S8b and d). More importantly, sulfuric acid is known to readily condense on the rBC surface (Zhang et al., 1993; Zhang et al., 2008; Kiselev et al., 2010), which provides an active surface for heterogeneous reaction by changing its hygroscopic properties (Khalizov et al., 2009). Hence, the characteristic features observed during the “transport/haze” period suggest an effective condensation of vapors to rBC surfaces and reaction with SIA, resulting subsequent growth of rBC particles in size and mass by forming thick coating on their surfaces (Adachi et al., 2014; Ueda et al., 2016; Lim et al., 2018). The shell size of rBC with a core diameter of 180–220 nm reached approximately 300–400 nm. Given that the PM_2.5 concentration increased linearly with the volume concentration of particles_{200–470 nm} (Figure 7a), the role of internally mixed BC particles is critical to the increase in mass of fine aerosol, as they provide surfaces for gaseous precursors of SIA to condense on.

In contrast, a thinner coating (29 ± 31 nm) and a lower concentration (by 40%) of rBC particles as well as the low levels of SIA were evident under low RH (53 ± 9%) conditions during “persistent anticyclone” period. In this case, the organic carbon (OC) fraction was the greatest in PM_2.5. ALWC was very low as well, possibly due to less water uptake by the aged organic aerosol (Engelhart et al., 2011). It is likely that meteorological conditions suppressed the condensation of volatile materials on aerosol surface and thus the secondary aerosol formation on rBC surface. It is also possible that organic materials may form an organic film on the BC surface, inhibiting the vapor uptake for SIA and the increase in rBC coating (Saxena et al., 1995; Sellegri et al., 2003). Instead, they promoted the growth of nucleation- and Aitken-mode particles smaller than 60 nm under low RH conditions (Figure 7a). The mass size distributions of OC, NO₃⁻, and SO₄⁻² estimated by the aerosol mass spectrometer generally show organics enriched in nucleation-mode particles (e.g., Kang et al., 2018; Kim et al., 2018).

In this regard, rBC particles and their coating thickness are proposed as an agent for generating accumulation-mode particles through SIA formation and as an operational parameter to understand the chemical mechanisms responsible for fine aerosol pollution driven by SIA, respectably. This, in turn, implies that reducing primary emissions of gaseous precursors and small particulates such as BC will be one of the effective ways to combat high PM_2.5 mass concentrations in Northeast Asia, especially during the cold months.

In Jeju, a background site in Northeast Asia, a 4-year (2013–2016) data set of nanoparticle distributions, PM_2.5 chemical composition, gaseous precursors, and meteorological variables were analyzed together with the vertical profiles of aerosol number concentrations and rBC properties that were measured during the KORUS-AQ study in 2016.

An EOF analysis of the size-separated number distribution distinguished 2 modes of aerosol diurnal evolution: the burst in nucleation- and Aitken-mode (EOF1) and the dominant number loadings in accumulation-mode (EOF2) particles. Our 4-year measurements at a regional background site highlights that EOF2 is representative of an episodic event diagnosing PM_2.5 mass increase that occurs over a course of several days, constrained by persistent clouds and shallow boundary layer, but not coupled with the NPF and growth event. Moreover, our results of SIA-driven PM_2.5 increases support previously proposed hypothesis about dynamic interactions between aerosol loading and meteorological parameters, that is, enhanced particle loading and reduced SR in the boundary layer, and SIA formation and ALWC enhancement.

Analyses of measurements obtained during the KORUS-AQ provided insight into the processes involved in the increase of PM_2.5 mass in this region; nanoparticles were entrained into the shallow boundary layer upon vertical mixing and gases condensed onto these particles at high RH exceeding the DRH of major inorganic salts, leading to an increase in PM_2.5 mass. RH played a critical role in this process and ALWC reached 30% of the dry PM_2.5 mass. Moreover, the first measurement set of rBC properties in Jeju demonstrates convincing evidence that BC particles provide surfaces for precursor gases to condense and form coating on, facilitating an increase in aerosol size and mass, particularly in the accumulation-mode. As a result, the coating thickness of rBC is hypothesized to serve as a proxy indicating the occurrence of SIA-driven high PM_2.5 mass in the study region. Consequently, our result supports strategies for reducing primary emissions of gaseous precursor for secondary inorganic species and rBC particles that provide the co-benefits of mitigating fine aerosol pollution and climate change. Detailed mechanism of heterogeneous reaction involving BC particles needs to be addressed in the future research.

The meteorological data compiled by Korean Meteorological Administration are available online at https://data.kma.go.kr/resources/html/en/aowdp.html (Last access: July 3, 2022). The long-term measurements made in Jeju by National Institute for Environmental Research (Jeju Aewol 2013-2016), along with vertical measurements of aerosol size distributions (Vertical Particle# Ansan May 2016 KORUS) and black carbon measurements (rBC Mixing state Jeju May 2016 KORUS) are available at KUPMO3 (https://sites.google.com/korea.ac.kr/kumpo3/home; Last access: July 31, 2022).

The supplemental files for this article can be found as follows:

Text S1. Figures S1–S8. PDF

This research was supported by the National Research Foundation of Korea (NRF), funded by the Ministry of Science and ICT (MSIT) (NRF2020M3G1A111499813). S. Lim was supported by the National Research Foundation of Korea (NRF) from the Ministry of Science and ICT (2018R1D1A1B07050849 and 2021R1C1C2011543). M. Lee thanks to the support by the National Research Foundation of Korea (NRF) funded by the Ministry of Science and ICT (NRF2020R1A2C301459213). S.-W. Kim was supported by the Basic Science Research Program through the National Research Foundation of Korea (NRF), funded by the Ministry of Education (2017R1D1A1B06032548). Funding to K.-S. Kang was provided by the National Institute of Environmental Research (NIER-RP2017-166).

The authors declare no competing financial interest.

Contributed to conception and design: SL, ML, XS.

Contributed to acquisition of data: SL, ML, PL, SK, KA, JG, MZ, KK.

Contributed to analysis and interpretation of data: SL, ML, KA, XS, JG, MZ, KK.

Drafted and/or revised the article: SL, ML, PL, SK.

Approved the submitted version for publication: SL, ML, PL.

How to cite this article: Lim, S, Lee, M, Laj, P, Kim, S-W, Ahn, K-H, Gil, J, Shang, X, Zanatta, M, Kang, K-S. 2022. Regional characteristics of fine aerosol mass increase elucidated from long-term observations and KORUS-AQ campaign at a Northeast Asian background site. Elementa: Science of the Anthropocene 10(1). DOI: https://doi.org/10.1525/elementa.2022.00020

Domain Editor-in-Chief: Detlev Helmig, Boulder AIR LLC, Boulder, CO, USA

Associate Editor: Md Firoz Khan, Department of Environmental Science and Management, North South University, Dhaka, Bangladesh

Knowledge Domain: Atmospheric Science

Part of an Elementa Special Feature: Korea-United States Air Quality (KORUS-AQ)