Peroxyacetyl Nitrate and Ozone Enhancement at Taehwa Research Forest under the Influence of Seoul Metropolitan Area

Peroxyacetyl nitrate (PAN) is produced by photochemical oxidation reactions with abundant NOx and volatile organic compounds (VOCs); therefore, it is considered as a photochemical pollution indicator. In this study, PAN, O3, and their precursors were measured at three heights (5.4, 23, and 40.5 m) on a 41-m tower in Taehwa Research Forest (TRF) near the Seoul Metropolitan Area (SMA) from August 25 to September 9, 2011. The PAN was determined every 2 minutes using gas chromatography with luminol chemiluminescence detection (GC-LCD). All reactive gases were measured for 15 minutes at each height. The mean and maximum PAN concentrations were 0.3 and 3.1 ppbv, respectively. The mean and maximum O3 concentrations were 13.1 and 79.8 ppbv, respectively. The average NOx concentration was 6.57 ppbv. At the TRF, PAN and O3 concentrations were well correlated (r = 0.8) and greatly elevated when the air mass was affected by urban outflows from the SMA, which was clearly demonstrated by an increase in NO2. These high NO2 concentrations were observed along with a shift in wind direction at 17:00 (KST) and resulted in the maximum observed values of PAN and O3 in the present study. In addition, the concentration enhancement was more pronounced for PAN and at heights above the canopy. These results highlight PAN as a robust tracer indicating urban impacts at peri-urban forest sites.


INTRODUCTION
Ozone (O 3 ) is primarily recycled through photochemical reactions of NO x and volatile organic compounds (VOCs) in urban air, during which peroxyacetyl nitrate (PAN) is formed as a byproduct (Taylor, 1969;Brich et al., 1984;Jenkin and Clemitshaw, 2000;Teklemariam and Sparks, 2004;Braslavsky and Rubin, 2011;Matsumoto, 2014;Tuncay et al., 2015). As PAN is produced in the presence of NO 2 and peroxyacetyl (PA) radical which is an oxidation product of VOCs and destroyed mainly by thermal decomposition, its concentration shows greater sensitivity to precursor levels than O 3 (Nielsen et al., 1981;Tsalkani et al., 1987;Schrimpf et al., 1995;Xu et al., 2015;Han et al., 2017). Thus, PAN is considered a suitable photochemical pollution indicator and has been extensively investigated to understand the driving mechanisms of O 3 pollution in large urban areas (Finlayson-Pitts and Pitts Jr, 1986;Gaffney et al., 1989;Gaffney et al., 1999).
In this context, PAN has been measured mostly in megacities, including Los Angeles, since the 1970s (Lonneman et al., 1976;Tuazon et al., 1981;Grosjean, 1982). During the 2000s, PAN measurements with concentrations ranging from 0.8 to 30 ppbv were reported for Mexico City, Seoul, Beijing and other major Chinese cities (Table 1). On the west coast of the United States, PAN has been paid much attention in association with long-range transport of PAN-enriched Asian outflow and has served as a reliable indicator of its influence (Jaffe et al., 1999;Hudman et al., 2004;Briggs et al., 2016;Baylon et al., 2017). In these studies, PAN concentrations ranged from 0.02 to 0.2 ppbv, much lower than the concentrations observed in northeastern Asia under the influence of continental outflows (Zhang et al., 2017).
The method of measuring PAN was first introduced by Darley (Darley et al., 1963), who used gas chromatography (GC) coupled with electron capture detection (ECD). GC with luminol-based chemiluminescence (GC-LCD) detection was constructed using a packed column (Blanchard et al., 1990) and later improved by employing a short capillary column, which enabled fast and sensitive measurements with high time-resolutions and low detection limits (Gaffney et al., 1998). Recently, the thermal dissociation-chemical ionization mass spectrometry (TD-CIMS) technique has become commercially available (Slusher et al., 2004).
As a precursors of PAN, NO x is mainly emitted from vehicles in urban areas, whereas the PA radical is produced from acetaldehyde, methylglyoxal, and acetone that are directly emitted or photochemically generated by oxidation process of anthropogenic VOCs (AVCOs) and biogenic VOCs (BVOCs) (Singh and Hanst, 1981;Roelofs et al., 1997;Choi, 2003;Calfapietra et al., 2013;Fischer et al., 2014;Carslaw et al., 2016;Sobanski et al., 2017). Although AVOCs are dominant in urban areas, BVOCs are more abundant on the global scale and their emissions are estimated at 760-1150 Tg C year -1 which comprises about 90% of total VOC emissions Sindelarova et al., 2014). Therefore, there is significant PAN production in forests proximal to urban areas due to the abundant VOCs and NO x (Sobanski et al., 2017).
Since BVOCs were first pointed out as the precursors of "blue haze" in mountainous regions (Went, 1960), researchers have demonstrated their role in tropospheric photochemistry that isoprene, the most abundant species of BVOC, has been intimately involved in southern U.S. ozone formation (Chameides et al., 1988). It is quite recent in East Asia that BVOCs have been investigated in the view of its role in atmospheric chemistry and have been measured in Beijing (Ran et al., 2011), Kyoto (Bao et al., 2010), and Shanghai (Geng et al., 2011). In the vicinity of California, secondary organic aerosols become highly enhanced when urban plumes are mixed with biogenic emissions (Shilling et al., 2013). Consequently, this study emphasizes the importance of measurements in forests, which are major sources of BVOCs.
In South Korea, forest areas consist of coniferous trees (42%) and deciduous trees (26%) comprises 65% of the country (KFS, 2013). The percentage of total forest area is even larger in major cities with 72% on average, including Seoul Metropolitan Area (SMA). In these major cities of Korea, O 3 concentration has increased since 2005 (Ghim, 2012) and a PM 2.5 mass concentration was implemented in 2015 as a national environmental standard. Previous studies have shown that O 3 formation shows more sensitivity to VOCs than NO x in SMA . Thus, with the growing demand for improved air quality, the impact of BVOCs on O 3 , PAN, and secondary aerosols formation needs to be thoroughly understood (Ran et al., 2011). Within this context, the National Institute of Environmental Research (NIER) initiated research to investigate the effect of BVOCs on air quality and established a measurement station with a tall tower at Taehwa Research Forest (TRF). In 2016, the TRF served as a ground platform for the  Kim, 2014;Kim et al., 2015a;Kim et al., 2015b). In this study, the temporal and vertical distributions of PAN were thoroughly examined to understand how the urban emissions of NO x affect PAN and O 3 concentrations when it is mixed with biogenic precursors in a peri-urban forest.

EXPERIMENT
The measurements of PAN and other trace gases, including O 3 , NO x , CO, and SO 2 , were conducted at TRF from August 25 to September 9, 2011. The TRF constitutes approximately 800 ha of afforested area that was established by education and research of the College of Agriculture and Life Sciences at Seoul National University. The TRF is located ~25 km southeast of Seoul and its satellite cities lie on the west (Fig. 1). On the east side of the forest, the Central Highway extends to the north and south. The observation facilities including a 41-m walkup tower and laboratory are located 165 m above the sea level (37. 18°N, 127.19°E) in a planted Korean pine (Pinus koraiensis Siebold et Zucc.) coniferous forest (299 ha). The surrounding region (476 ha) is a natural deciduous forest such as Murray (Quercus serrata Murray) and Blume (Quercus aliena Blume). The tower is equipped with sampling inlets at heights of 5.4, 23, and 40.5 m and has meteorological sensors at 5.4 and 40.5 m. The canopy height of TRF is approximately 21 m above the ground surface.
The air was pulled down to the manifold through PFA tubing (OD of 6.35 mm, ID of 3.96 mm) at 10 L min -1 at the three heights of the tower. At each height, air was sampled for 15 minutes, producing hourly vertical measurement profile. From the manifold, air was distributed for PAN, O 3 , NO x , CO, and SO 2 measurements. PAN was measured every 2 minutes using a fast GC-LCD system (Marley et al., 2004;Lee et al., 2008). The sampled air was introduced to the 6-port valve (Cheminert C22, Valco Instruments, TX, USA) at 100 mL min -1 and injected into the capillary column (DB-1, J&W Scientific, CA, USA) through a 2 cm 3 sample loop by nitrogen gas at 30 mL min -1 . The air flow was controlled by a mass flow controller (Alicat Scientific Inc., AZ, USA). Luminol solution constantly flowed into the reaction cell at 0.6 mL min -1 , where the chemiluminescence signal was detected by a photon counter (HC135-01, Hamamatsu, NJ, USA). The PAN standard solution was synthesized by peracetic acid with n-tridecane at temperatures below 20°C (Gaffney et al., 1998). The detailed calibration procedure can be found elsewhere (Han et al., 2017). The nominal PAN detection limit was 0.1 ppbv (Lee et al., 2008).
Other trace gases, such as O 3 , NO x , CO, and SO 2 , were measured every minute by a conventional monitoring instrument (49i, 42i, 48i, and 43i, Thermo Electron Incorporation, MA, USA). For discussion, all measurements data were averaged for 2 minutes according the PAN measurement resolution. The temperature and relative humidity (RH) measurements are available from August 29 onwards at TRF.
The backward trajectories of air masses were calculated using the Hybrid Single Particle Lagrangian Integrated Trajectory (HYSPLIT) model with Global Data Assimilation System (GDAS) data from the National Oceanic and Atmospheric Administration (NOAA), in conjunction with Trajstat software (Wang, 2014).

Characteristic Distribution of PAN and Other Trace Gases
During the experiment, temperature gradually decreased from August to September (Fig. 2). During the first four days, without meteorological measurements from the TRF, the daily maximum temperatures were above 30°C in Seoul and nearby cities. Although the experiment period was characterized by low levels of primary pollutants (e.g., SO 2 , NO x , and CO) in Korea under the influence of maritime air (MOE, 2011), trace gases show considerable variations and their highest concentrations are commensurate to SMA levels.
Throughout the experiment, PAN concentration had a mean of 0.3 ppbv and reached a maximum of 3.13 ppbv ( Table 1). The mean and maximum concentrations of all measurements are listed in Table 1. The mean PAN concentration in this study is comparable to concentrations observed at forest sites (0.1-0.45 pbv), but lower than those at rural sites (0.43-0.76 ppbv) ( Table 2). In contrast, the maximum PAN concentration in this study is significantly higher than concentrations measured at other forest sites (0.53-1.47 ppbv) and comparable to those at rural sites (3.90 ppbv). O 3 concentration averaged 13.7 ppbv, which is considerably lower than measurement in forest regions (53-57 ppbv) and background site of the study region (34 ppbv). The maximum O 3 (79.8 ppbv) was also lower than those of other sites (83-161 ppbv). The CO and SO 2 levels of the TRF were lower than those of the SMA and comparable to those of background sites of the study region (MOE, 2011). The mean concentration of NO x was much lower than SMA, but the highest concentrations were higher than those typically observed in forests (Kim, 2014).
The temporal variations of PAN and O 3 are shown in Fig. 2 with the NO x and CO concentration. Apparently, the experiment period falls into two high-and two lowconcentration episodes from the variations of these trace gases: August 25-28, August 29-September 2, September 3-4, September 5-9. In the averaged diurnal variations, PAN and O 3 concentrations reached maximum at 15:00 (local time) (Fig. 3(a)). Although the maximum NO x concentration was identified at 09:00, diurnal variation was unclear throughout the day and the amplitude of averaged diurnal change was less than 3 ppbv from the 5 ppbv baseline concentration.
A polar plot is known to be useful tool to investigate source characteristics of air pollutants (Grange et al., 2016). The polar plot incorporated with wind measurements illustrates that NO x is affected by nearby sources, such as emissions from the highway in the east and the urban areas in the west (Fig. 3(b)). In comparison, the highest PAN and O 3 concentrations are associated with relatively weak northwesterly and strong northerly winds, respectively.
The vertical distributions of PAN, O 3 , and NO x concentrations were compared on a frequency histogram overlaid with the three heights (5, 23, and 40.5 m) (Fig. 3(c)). For these three species, the highest concentrations were mostly observed above canopy (i.e., 23 and 40.5 m). These results demonstrate that the TRF represents a peri-urban forest where nearby urban emissions are mixed with biogenic emissions from vegetation (Ham et al., 2016). The characteristic distributions of PAN, O 3 , and NO x are further discussed in the following sections.

Controlling Factors of PAN and O 3 Levels
In this study, PAN is strongly correlated with O 3 (Pearson's correlation coefficient, r = 0.8), and there is also a strong positive relationship between the daily maximum concentrations of PAN and O 3 (r = 0.78), usually observed in urban areas (Lee et al., 2008;Zhang et al., 2009). Daily maximum PAN and O 3 concentrations have also been reported to correlate strongly with daily maximum temperature (Wunderli and Gehrig, 1991), implying that temperature is an important factor in the formation of these two species. In this study, the daily maximum PAN and O 3 concentrations did not always proportionally increase with the daily maximum temperature (Fig. 4). This inconsistency is because PAN and O 3 concentrations showed a sudden drop with temperature between September 3 and 4, along with relative humidity (Fig. 2). The time series of all measurements also reveals that PAN, O 3 , and primary pollutants level changed every 2-5 days. As meteorological changes are dynamic in northeast Asia, the air mass pathways were examined as a potential control factor affecting atmospheric processes and precursor levels at the TRF.
The backward trajectories calculated using the HYSPLIT model illustrate that air mass changes occurred frequently during the 16-day experiment (Fig. 5). At the beginning of the experiment, enhanced NO 2 and O 3 concentrations were associated with air coming from the East Sea through southern South Korea. As stagnant conditions developed, the air mass that lingered over the SMA arrived at the TRF, leading to the highest PAN and O 3 concentrations in the end of August. As air moved northeast rapidly from the East Sea, all reactive gas concentrations dropped down to the lowest observed levels with a decrease in relative humidity from September 3 to 4. PAN and O 3 levels were elevated again between September 5 and 7, when the air quickly passed through the SMA and arrived at the TRF. According to these observations, the entire experimental period was divided into four intervals (i.e., P1, P2, P3, and P4), within which the behaviors of PAN and O 3 were examined in detail. The measurement statistics for each period are summarized in Table 3. PAN and O 3 concentrations were highly elevated in P2 and P4 when the air passed over the SMA. The maximum concentrations of both species were observed in P2 and corresponded with the high NO 2 concentrations and NO 2 /NO x ratios (Fig. 6). For each period, the correlation between O 3 and PAN is presented in Fig. 6(a). In P4, PAN was relatively more elevated than O 3 , corresponding to high CO and NO concentrations and a low NO 2 /NO x ratio. The low O 3 /PAN ratios in P2 and P4 indicate that PAN was relatively more elevated than O 3 in the presence of urban emission influence at the TRF. It is also noteworthy that the PAN/O 3 ratio was highest during P4, of which urban plumes were likely less aged compared with those of P2, based on their history and chemical signature. In contrast, O 3 concentrations were relatively higher than PAN concentrations during P1,       Table 3.   which had the highest NO 2 /NO x ratios and temperatures (measured nearby). At the TRF, the measurement results of PAN and O 3 demonstrate that while the O 3 levels are sensitive to temperature and degree of chemical aging, the PAN concentrations are sensitive to NO x levels, which agrees with the current understanding of tropospheric photochemistry. As a result, PAN concentrations were greatly elevated when the air mass was affected by emissions from the SMA, of which degree of enhancement was largely dependent on meteorological conditions. These results highlight PAN as a robust tracer indicating urban emissions at a peri-urban forest.

Urban Influence on PAN and O 3 Enhancement
The correlation between PAN and O 3 was different in each of the four periods, and the best correlation was observed in P2 ( Fig. 6(a)). In P2, PAN and O 3 time-series distributions show the afternoon maximum peak split into two peaks, and more pronounced for PAN than for O 3 (Fig. 2). This was particularly evident on August 31, when the second peak was much larger than the first and concurrent with elevated NO 2 . Thus, NO 2 concentration was added to the correlation between PAN and O 3 in Fig. 6(b). In urban air, PAN and O 3 are produced by photochemical reactions and their concentrations usually reach maximum during the afternoon when NO 2 concentrations are lowest (Lee et al., 2008). However, at Jeju Island, a background site in northeast Asia, PAN concentrations were unusually elevated concurrently with NO 2 in continental outflow plumes (Han et al., 2017). Similarly, in this study, the highest concentrations of PAN and O 3 were associated with elevated NO 2 concentrations, particularly during P2. The PAN and O 3 concentrations on August 31, when the PAN and O 3 maximums were observed, are presented with NO 2 concentrations and wind vectors (Fig. 7).
In Fig. 7, high PAN and O 3 concentrations are distinguished by high NO 2 concentrations, and are observed later than 15:00 after the first peak around 13:00 that is associated with low NO 2 . PAN and O 3 gradually increased with a decrease in NO 2 until 13:00, and it agrees with the results expected from photochemical reactions in urban atmosphere. After the first peak at 13:00, PAN concentration decreased whereas O 3 concentration remained high. At approximately 15:00, PAN and O 3 concentrations rapidly increased with an increase in NO 2 , leading to maximum observed values at 17:00. During this time interval, PAN concentration increased significantly more than O 3 , despite no visible differences in PAN and O 3 concentrations among the three heights at the first peak. In contrast, the late enhancement of these two species was more evident at 23 and 40.5 m (above canopy) than at 5.4 m (below canopy) ( Fig. 3(c)). Furthermore, the rapid increases in PAN and O 3 were concurrent with the wind direction change (Fig. 7).
In the late afternoon, the northerly/northwesterly winds were consistent. These results demonstrate that the late maximum PAN and O 3 concentrations were affected by urban outflows. Although VOCs measurements were not available for this experiment, they were measured at the same site between May and August of 2013 (Kim et al., 2015a). During period, isoprene concentration reached a maximum at 17:00, concurrent with the observed PAN and O 3 maximum in this study. Late afternoon isoprene emissions have also been reported in previous work (Apel et al., 2002;Bryan et al., 2012). Although the tower is surrounded by planted pine trees, the TRF is dominated by deciduous trees that are concentrated to the northwest of the tower. Thus, northwesterly winds carry not only urban emissions from the SMA but also isoprene emitted from nearby deciduous trees that was likely to move down to the tower. Isoprene is known as a major VOC that produces PAN on a global scale (Fischer et al., 2014). These facts well agree with the source apportioned for PAN in Fig. 3(b).
As the lifetime of PAN is less than one hour at ~30°C, the observed late afternoon peak in PAN concentration was unlikely because of transport from the SMA. On the other hand, the solar zenith angle was 67° at 17:00, suggesting that in-situ photochemical reactions could not fully account for the elevation of PAN concentrations in the late afternoon. Taking these results into consideration, BVOCs may have played a role in the rapid increase of PAN and O 3 concentrations to the maximum in the late afternoon. With a shift in wind direction, the TRF was under the influence of urban emissions from SMA. On the way to being transported to the tower, the plume passed over the forest region, thereby being mixed with biogenic emissions. Therefore, it is highly likely that PAN was produced from nearby forest upon the influx of urban emissions and moved to the sampling tower. As PAN is sensitive to precursor levels, the PAN increase was instantaneous with increases in NO 2 . Although the similar increase of O 3 concentration raised to the maximum at 17:00, the enhancement of the second peak relative to the first peak was much greater for PAN (3.1 vs. 1.5 ppbv) than O 3 (79.6 vs. 56.1 ppbv) (Fig. 7).

CONCLUSIONS
The mean and maximum concentrations of PAN and O 3 were 0.3 ppbv and 3.1 ppbv, and 13.1 ppbv and 79.8 ppbv, respectively. Overall, the two species were strongly correlated (r = 0.8), but their maximum daily concentrations did not always proportionally increase with the temperature. In this study, the PAN and O 3 levels varied considerably with meteorological conditions and were largely dependent on the degree of urban influence. Accordingly, NO x and CO concentrations ranged from summer background to major Korean city levels with means of 6.6 ppbv and 254 ppbv, respectively.
As the TRF is a peri-urban forest, the PAN and O 3 concentrations there were significantly elevated under the influence of the SMA, resulting in the maximum concentrations of the two species during the experiment. The maximum concentrations were observed in late August, when the diurnal variations of PAN and O 3 showed clear peaks at 13:00 and 15:00. The later peak was associated with urban outflows from the SMA, characterized by a rapid increase in NO 2 concentration and a change in wind direction. Particularly, the enhancement of PAN was more evident than that of O 3 and occurred more obviously above than below the canopy. Another episode with lower temperatures and higher wind speeds occurred in September, during which the PAN/O 3 ratio reached its maximum in the experiment, with less aged outflow from the SMA than the other periods. When PAN and O 3 concentrations remained low, the O 3 /PAN ratio was higher, with a higher NO 2 /NO x ratio at higher temperatures. The measurements at the TRF demonstrate that O 3 concentrations are more sensitive to temperature and air mass aging, whereas PAN concentrations are more sensitive to NO x . Thus, the results of this study highlight PAN as a robust indicator of the urban impact on peri-urban forests.