Current Issue

The field measurements were made at the Baengnyeong Island atmospheric research supersite, Korea (37.97°N, 124.63°E, 100 m above sea level), operated by the Korea National Institute of Environmental Research (NIER). The supersite is located in between west coast of the Korea, and Shandong Peninsula in eastern China (Fig. 1). The measurement site is suited for monitoring the characteristics of background aerosol level, because it is located at the top of a hill aparted from the island’s sparsely populated area. The measurements were conducted from February to November 2012 including an intensive period of September 21 to 29, 2012.

Fig. 1. 
Map of Korea showing the sampling location at Baengnyeong Island. The satellite images were obtained from google map.

Non-refractory sub-micron aerosol (PM₁) composition was measured using HR-ToF-AMS. The operation of the Aerodyne HR-ToF-AMS has been described in detail elsewhere (DeCarlo et al., 2006; Drewnick et al., 2005; Jimenez et al., 2003; Jayne et al., 2000). Ambient air was drawn through a URG cyclone (D₅₀(mass median diameter)=2.5 μm, 3 L/min). Perma Pure dryers (MD-110-24, Perma Pure LLC.) were used to control the sample humidity, and the uncertainties from the bounce-related changes in the collection efficiency and the particle transmission through the aerodynamic lens were therefore both reduced. Detailed descriptions and procedures for the calibration of the AMS at Baengnyeong Island atmospheric research supersite are given in Lee et al. (2015). In this study, a six-minute time resolution was used to determine the non-refractory fine particle composition, including nitrate, sulfate, ammonium, and organic matter concentrations.

The PTR-ToF-MS was deployed to collect the VOC mixing ratio at the site. The basic principle of this instrument has been described elsewhere in detail (Graus et al., 2010; Jordan et al., 2009). The gas sample inlet was heated to ~60°C to prevent the wall loss of VOCs. During the intensive period, the drift tube of the PTR-ToF-MS was operated at a temperature of 60°C, a drift voltage of 600 V, and a pressure of ~2.2 hPa, as controlled by software. These conditions correspond to an E/N (electric field to number density of air ratio) value of ~138 Td. Ions were pulsed every 60 μs into the time-of-flight region, and they were detected by the Multi Channel Plate (MCP) at 0.2 ns resolution. Data were stored every one minute in the compressed HDF5 format. The data processing for the determination of the mass calibration, ion fitting (e.g., peak shape, peak height, and peak width), and transmission efficiency by standard gases was carried out according to the Igor routine.

The aerosol number concentrations were measured by using a scanning mobility particle sizer (SMPS, TSI model 3034) with Nafion^® dryer. The SMPS was operated with an aerosol flow rate of 1.0 L/min through the Kr-85 neutralizer, and a sheath airflow rate of 5.0 L/min. The size range was set as 10-470 nm for the mobility diameter, with a measurement interval of 5 minutes. The general principle and operation of the instrument has been described in detail elsewhere (Kim et al., 2013; Kulmala et al., 2004).

The measurements of gaseous pollutants (O₃, NO_x, CO, and SO₂) were conducted on an hourly basis. The general details for the instrument has been described in Lee et al. (2012). The hourly relative humidity data were obtained from an auto weather station (AWS; operated by the Korea Meteorological Administration) at a distance of 0.5 km in a northwest direction.

The NOAA HYSPLIT model (GDAS1) was used to analyze the aerosol composition and the VOC by computing four back-trajectories per day for the air masses arriving at the sampling site with a receptor height of 150 m, which is the same as the height of the site. The back-trajectories were run for 72 hours.

The timeline of AMS aerosol mass spectra was analyzed using positive matrix factorization (PMF), a multivariate factor analysis tool (Paatero, 1997; Paatero and Tapper, 1994). We used the Igor Pro-based PMF evaluation tool (PET) (Ulbrich et al., 2009). PMF was performed upon the V-mode AMS data after applying the error preparations outlined in SQUIRREL and PET. The PMF is described in detail in Lee et al., 2015.

Fig. 2a depicts the timelines of the concentrations measured between February and November, 2012, and Fig. 2b presents a time series of the concentrations that were observed during the intensive study period of from September 21th to 29th, 2012. Organic matter and sulfate were generally the most abundant aerosol components, and they exhibited maximum concentrations of 30 μg/m³ and 29 μg/m³, respectively (Fig. 2a). The nitrate concentrations peaked at 25 μg/m³, but they were typically much lower than the sulfate and organic matter concentrations. The period from February to June and October featured the higher concentrations of the organics, nitrate, and sulfate, with the lower concentrations observed from July through September due to the relatively clean air masses that generally reach the site during those months of the year (Lee et al., 2015). Another interesting observation is the variation of the nitrate concentration at the site according to the different months. Nitrate reached a low concentration when the temperature reached a maximum in the summer time period of July to October, supporting the theory that temperature has an impact on the gas-particle partitioning of nitric acid and NH₄NO₃ dissociation (Lee et al., 2008a, b; Yu et al., 2006).

Fig. 2. 
Timelines of PM₁ organic, ammonium, sulfate, and nitrate concentrations measured using HR-ToF-AMS during: (a) February to November, 2012 and (b) the intensive study period of September 21-29, 2012.

Mass peaks with signals above the noise were identified, and include major primary ions (e.g. (H₂O)H⁺, (H₂O)₂H⁺, and water clusters), impurities such as O₂⁺and NO₂⁺, and ammonium ions. In this paper, we focus on the 17 selected ion species shown in Fig. 3. The detectable biogenic VOCs (BVOCs) from the site were isoprene (m/z 69.071), 2-methyl-3-buten-2-ol (MBO, m/z 87.077), dimethyl sulfide (DMS, m/z 63.071), and monoterpenes (m/z 81.070, 95.086, 137.131), with the averaged mixing ratios of 0.24 ppbv, 0.25 ppbv, 0.20 ppbv, and 0.38 ppbv during the intensive study period, respectively. Toluene (m/z 135.116, 0.45 ppb) and benzene (m/z 79.054, 0.32 ppb) comprised the majority of the anthropogenic VOCs (AVOCs) with the averaged mixing ratios of 0.45 ppb and 0.32 ppb during the intensive study period of September 23 to 29, respectively. For the oxygenated VOCs (OVOCs), acetone (m/z 59.048), acetaldehyde (m/z 45.033), acetic acid (m/z 61.027), formic acid (m/z 46.997), and methyl vinyl ketone and methacrolein (MVK+MACR, m/z 71.048) were measured.

Fig. 3. 
Averaged mixing ratios (labels in each bar) for 17 selected ion species during the intensive study period of September 23-29, 2012.

MVK+MACR are well known major secondary products from the atmospheric oxidation of isoprene. The mixing ratio of MVK+MACR was measured between 0.01 and 1.81 ppbv at the site. Acetone is ubiquitous in the atmosphere and has a variety of sources including terrestrial vegetation, anthropogenic emissions, and photochemical productions. Acetone was the most abundant VOC with an average of approximately 4 ppbv, but its range is from 0.96 to 17.53 ppbv, which is very similar to the ranges found in previous studies (Park et al., 2013; Yuan et al., 2013). Acetaldehyde is emitted by live leaves, and its sources are similar to acetone. The acetaldehyde was measured with an average of 2.67 ppbv with mixing ratios between 1.09 and 12.48 ppbv. Acetic acid was among the top three measured species with mixing ratios ranging from 0.01 to 28.34 ppbv, and an average of 1.67 ppbv. OVOC may be a result of the oxidation processes of BVOC and AVOC in the atmosphere. Acetone, acetaldehyde, acetic acid, and formic acid were measured for approximately 74% of the total measured VOCs, suggesting strong oxidation processes at the site (Fig. 3).

The total BVOCs shown in the lower panel of Fig. 4 include isoprene, MBO, DMS, and monoterpenes. MBO is emitted from certain pine trees (Holzinger et al., 2005). Isoprene, MBO, and monoterpenes were measured for approximately 7.4% of the total measured VOCs (Fig. 3). The measured BVOCs showed that emissions were at a maximum during the daytime around 14:00 local time (Fig. 4). The highest concentrations of BVOCs were observed at the beginning of the intensive study period with a maximum of ~2 ppbv, and with concentrations decreasing to 0.5 ppbv at the end of the study (Fig. 4).

Fig. 4. 
Timelines of the intensive study period of 9/23-29, 2012 of gaseous pollutants (CO, NO_x, O₃, and SO₂ upper panel); O/C, f₄₄ : f₄₃ ratios for PTR-ToF-MS and relative humidity (middle panel); and major anthropogenic VOCs (toluene and benzene and total biogenic VOCs) and organics (bottom panel) (Note: (a) aqSOA formation, (b) gasSOA formation, and (c) nighttime aqSOA).

The highest concentrations of toluene were observed from September 25 to 27 when concentrations of BVOCs were low. In general, benzene was measured at low concentrations, with a maximum of 0.89 ppbv during the entire intensive study period. The AVOCs (toluene and benzene) were measured for approximately 5.4% of the total measured VOCs. Fig. 4 demonstrates an interesting relationship between toluene and benzene. Benzene is predominantly derived more from vehicles, and it is well-correlated with other aromatic VOCs such as toluene when the main source of aromatic VOCs is vehicle exhausts. The benzene concentration trends were similar to those of toluene on September 23 to 24, indicating that toluene and benzene may be derived from the same source, as follows: local motor vehicle exhausts. Periods of relatively high toluene concentrations were, however, observed on September 25 and 27 along with the lowest concentrations of the benzene. This is associated with the transport of polluted air masses to the site, as shown Fig. 5. The air mass trajectories in Fig. 5 show that air masses on September 27 spent time over industrial areas in Korea before arriving at the site, while air masses passed over China until 12:00 LST on September 25, when the high concentration of toluene was observed.

Fig. 5. 
Back trajectories of air masses arriving at Baengnyeong Island for (a) 9/25 and (b) 9/27, 2012.

It is well accepted that SOA forms through the partitioning of semivolatile products of the gas-phase photochemical reactions of VOCs and atmospheric oxidants (Pankow, 1994a, b). This process has been studied extensively in smog chambers. According to partitioning theory, the major factors are considered to be the volatility of gas-phase oxidation products, the organic mass in the pre-existing particles, and temperature. Precursors must be large (>C₇) to produce high SOA yields, because they must produce products with low enough volatility to partition substantially into the organic matter in the particle phase. Because precursors are large and only a few oxidation steps are needed to form condensable products, the O/C ratio of “smog-chamber” SOA is modest (0.3 to 0.4; Aiken et al., 2008). Higher SOA yields typically occur in lower NO experiments (higher VOC/NO), because the products from high NO are smaller and more volatile. Partitioning theory is a fundamental tool in the modeling of SOA (Donahue et al., 2006; Odum et al., 1996; Pankow, 1994a, b).

While clearly important, it fails to consider that, in many locations, the liquid water in aerosols, clouds, and fog may be a more accessible medium than organic matter for the partitioning of gas-phase oxidation products (Turpin et al., 2000). Lim et al. (2013) explained that glyocal (C₂), methyglyoxal (C₃), and acetic acid have great potential to form SOA via aqueous chemistry in clouds, fogs, and wet aerosols. For aqueous SOA (aqSOA) formation to be substantial, the precursor compounds must be highly water soluble. This suggests that aqSOA will have high polarity and O/C ratios (O/C)=1-2, 1.5, and 2 for glyoxal, methyglyoxal, and oxalate, respectively, and O/C=~1 at aerosol relevant concentrations (Perri et al., 2009; Altieri et al., 2008).

To gain further insight into the mechanisms of formation of the SOA at the Baengnyeong Island sampling site, several cases were considered. Fig. 4 also shows the timelines of the gaseous pollutants, and the relationship of the O/C and f₄₄ : f₄₃ ratios with precursor VOCs and OAs during the intensive study period. The values of f₄₄ : f₄₃ describe ambient OA oxidation. The relative abundances of the key OA ions at m/z 44 (CO₂⁺) and m/z 43 (C₂H₃O⁺) distinguish the semi-volatile oxygenated OA (SV-OOA) from the low volatility oxygenated OA (LV-OOA). The low f₄₄ : f₄₃ values (≤~1) suggest that the bulk of the observed OA is slightly aged/oxidized, and likely resembles the previously reported SV-OOA from the SOA formation chamber studies for relatively fresh aerosol, whereas the high f₄₄ : f₄₃ values indicate that the OA is very oxidized or very similar to previously reported reference spectra of POA emitted directly from anthropogenic sources. The timeline of the f₄₄ : f₄₃ ratio shows that the observed OA was relatively oxidized from September 23 to 25, while the OAs was more likely to be fresh aerosol in SOA. The ratios of O/C and f₄₄ : f₄₃ decreased with the increasing availability of the precursor BVOCs and AVOCs. Different chemical pathways have been proposed here for the mechanisms of formation of SOA based on the O/C ratio described by Ervans et al. (2011). In this study, the aqSOA formation was investigated during both the daytime and the nighttime, as in cases (a) and (c), respectively, and the SOA that was formed in the gas phase and was then partitioned to an organic phase (gasSOA) was also examined, as in case (b) (Fig. 4). Cases (a) and (c) showed an aqSOA formation with an O/C ratio of more than 0.5 when a high concentration of VOCs (BVOCs and AVOCs) is present. On the other hand, the O/C ratio of less than 0.5 in case (b) suggests a possible mechanism of gasSOA formation. The time series of the relative humidity during the intensive study period did not, however, show a significant variation, except during the last gasSOA episode (after September 28, 12:00). The mean relative humidity of the aqSOA and gasSOA episodes were 78.7% and 73%, respectively, which were relatively high due to the geographical location. Compared with the aqSOA episode, the gaseous pollutants (especially SO₂ and O₃) and the toluene rapidly increased during the gasSOA episode. It can therefore be easily deducted that Baengnyeong Island was dominated by aqSOA due to a high relative humidity, except when the gaseous pollutants and VOCs (especially toluene) increased to cause the formation of gasSOA.

To identify the reasonable factors, the oxygenated OA (OOA) components in the triangle plot are transformed into the Van Krevelen diagram, which can indicate bulk changes in oxygenation in ambient datasets, and it may also be used to explore the reaction mechanisms in isolated airmasses (Fig. 6a). The data points in the apex of the triangle area (higher O : C and lower H : C) indicates highly oxidized particles on the van Krevelen-triangle diagram (Ng et al., 2011). The SV-OOA located in 1.3-1.6 for H : C and 0.1-0.6 for O : C, whereas the LV-OOA is more oxidized, with lower H : C than SV-OOA at 1.2-1.5 (Aiken et al., 2008, 2007). Therefore, we concluded that there is no POA from the intensive period of this study, and it would be reasonable to divide into the two factors of SV-OOA and LV-OOA.

Fig. 6. 
(a) Van Krevelen -triangle diagram for the OA components from the organics with HR-AMS data. The mass spectra of the two OA factors ((b) SV-OOA and (c) LV-OOA) were determined on the basis of PMF analysis of the HR-AMS data from September 2012.

The results from the PMF analysis of AMS observations consequently yielded the two factors of SV-OOA (46% of OM) and LV-OOA (54% of OM). Both SV-OOA and LV-OOA were indicated by a higher signal at m/z 44 (CO₂) than m/z 43 (C₃H₇⁺ and CH₃CO₃⁺), whereas an LV-OOA that is highly oxidized has a more dominant peak at m/z 44 than SV-OOA. The LV-OOA factor has a fractional m/z 44 (f₄₄) of 17% and a fractional m/z 43 (f₄₃) of 3%, similar to the typical LV-OOA factors that are based on the aerosol volatility (f₄₄=12-15%, f₄₃=5-7%) determined in previous studies (Ng et al., 2011, 2010; Zhang et al., 2011; Hildebrandt et al., 2010). The SV-OOA factor is distinguished by a much lower fraction of m/z 44 (f₄₄=12%) than seen in LVOOA, indicating less oxidized (Ng et al., 2011). Typically, SV-OOA and LV-OOA have characteristics that are well correlated with nitrate and sulfate, respectively (Zhang et al., 2011). Indeed, the two factors were highly correlated with nitrate and sulfate, with high correlation coefficients of 0.93 and 0.91, respectively, according to the AMS (not shown).

Fig. 7 shows the time series of the fraction for SV-OOA, and the O/C ratios and concentrations of SV-OOA and LV-OOA during the intensive measurement period. The thresholds of the O/C ratio (0.5) for the distinguished gasSOA and aqSOA, and the negative correlation shown by the fraction of the SV-OOA (R=~- 0.55) with the O/C ratio all need to be considered here. It should also be noted that the fraction of SV-OOA in the aqSOA was relatively low compared with that in the gasSOA, whereas the fraction of SV-OOA rapidly increased in the gasSOA, especially in the first and second gasSOA periods, when the concentrations of toluene, benzene, and total BVOC was highest during the intensive period. For SV-OOA, the highest concentrations were indicated during the first and second gasSOA periods. The fraction and concentration of SV-OOA then significantly decreased for the nighttime aqSOA. However, the third gasSOA period showed an increase in the fraction of SV-OOA. Though aqSOA was dominant at Baengnyoung Island, this phenomenon can be interpreted as the activation of gasSOA by gaseous pollutants and VOCs, resulting in an increase of SV-OOA.

Fig. 7. 
Time series of fractions of SV-OOA and O/C ratio (upper) and SV-OOA and LV-OOA (bottom) during the intensive measurement period (Note: (a) aqSOA formation, (b) gasSOA formation, and (c) nighttime aqSOA).

During the intensive period, we observed several days (September 18 and 19) of increases in number concentrations of the smallest measured particles, along with a succeeding development of the aerosol size distribution (Fig. 8a). It is similar to the new particle formation (NPF) events shown in previous studies (Kim et al., 2013; Levin et al., 2012; Boy et al., 2008; Kulmala et al., 2004). Boy et al. (2008) divided the events according to nucleation and growth. For the NPF events, nucleation mode particles (3-10 nm) were clearly observed at the event, while new particle formation events (particles smaller than 6 nm) were not observed. They explained that the nucleation started upwind with a subsequent particle growth as the aerosol population was transported the site. Even though our measurements do not extend to particles smaller than 10 nm, there was no significant extension of the particle number distribution around 10 nm, as shown in Fig. 8b. The events on September 18 and 19 are more likely to be “new particle formation”. The previous studies often observed NPF under high concentrations of gaseous VOCs. However, our measurements showed that there was no new particle formation during the times of high concentrations of gaseous VOCs on September 25 and 27 (Fig. 8a). This supports that gas to particle conversion by absorption and adsorption involves interaction of the gaseous VOCs with the pre-existing aerosol particles in polluted conditions. Because concentrations of the VOCs from the PTRToF-MS are not available, NPF on September 18 and 19 cannot be clearly explained. However, a back-trajectory analysis shows that the air masses that arrived on September 18 and 19 passed through the relatively clean environment of the northern parts of China and Korea (Fig. 9a and 9b), while the air mass of September 28 passed through relatively polluted regions (Fig. 9c), and a high population of large size particles (~30 nm) was observed (Fig. 8a). It was hypothesized that the NPF began upwind with the subsequent growth as the aerosol population was transported to the site from the ocean, and that a clean condition is very important for the NPF at this site.

Fig. 8. 
(a) Time series of aerosol number distribution by SMPS, and concentrations of organic and total VOCs (biogenic VOCs+anthropogenic VOCs) measured by HR-ToF-AMS and PTR-ToF-MS, respectively, at Baengnyeong Island, and (b) aerosol number distribution evolution during a new particle formation event.

Fig. 9. 
Back trajectories of air masses arriving at Baengnyeong Island for (a) 9/18, (b) 9/19, and (c) 9/28, 2012.