Chemical Composition and Size-Fractionated Origins of Aerosols over a Remote Coastal Site in Southern Taiwan

In spring 2013 air samples were collected from a coastal site in the sparsely populated far south-west of Taiwan and analysed for ambient gases, inorganic salts, carboxylates, and saccharides. Concentration of ambient gases was in the order SO2 > HCl > HNO3 > NH3 > HNO2. Day-night variation in concentrations indicated that photochemical conversion of HNO2 to HNO3 occurs during the day. PM2.5 (16.16 ± 5.30 μg m) accounted for 61.1% of PM10 mass concentrations. The main inorganic salts were SO4, NH4, Na, NO3, and Cl, collectively accounting for 48.8 ± 27.4% of the PM2.5. Cldepletion during the day was higher than during the night due to the presence of reactive photochemical products. The average Cl-depletion of PM2.5 (53.1%) was markedly higher than that of PM2.5–10 (26.0%), indicating that in PM2.5, a high amount of Cl reacts with acidic gases to form HCl, which then escapes into the atmosphere. The carboxylate concentration in PM2.5 was 0.50 ± 0.24 μg m. It was found that low-molecular-weight carboxylates formed more readily in the open coastal region than in urban regions of southern Taiwan. Additionally, the daily mean ratio of Oxalate/non-seasalt SO4 (6.15 ± 2.28%) in the coastal region was higher than that in the urban regions in southern Taiwan. The most prevalent saccharide in PM2.5 was myo-inosital (333 ± 300 μg m), a type of soil fungus metabolite. Emissions of arabitol and mannitol, emitted through lichen and fungal activity, were markedly higher during the day. Only a trace amount (8.92 ± 16.92 μg m) of Levoglucosan (Levo), an indicator of biomass burning, was detected. The mean Levo/organic carbon ratio was 5.04 ± 8.72‰, suggesting that biomass burning contributed slightly to aerosols in the study area. An analysis of air mass backward trajectories showed that the products of biomass burning in Southeast Asia and southern China may be transported to the study area through long-range transport. This effect is more noticeable during the day when onshore breezes support the transport of particles sourced from the west of Taiwan.

Levoglucosan (Levo) and mannosan (Manno) are stable compounds and are the main saccharide species in PM 2.5 .Levo in particular can serve as a tracer of biomass burning products in PM 2.5 (Souza et al., 2014).Levo can be detected in rural, suburban, urban, and coastal regions, and is highly correlated with mannosan (Simoneit et al., 2004;Hsu et al., 2007b;Lee et al., 2008;Engling et al., 2009;Tsai et al., 2015).A study conducted on the remote island of Chichi-Jiam (Japan) in the north-west Pacific showed that the total concentration of levoglucosan had increased between 1990 and 2009 and, additionally, that concentration was highest during winter and spring.These observations are mainly attributable to the long-range transport of biomass burning products and terrestrial organic matter from the Asian continent in winter and spring (Chen et al., 2013).
Pingtung County is a sparsely populated region of Taiwan that extends from the Central Mountain Range to the narrow southernmost tip of the island.Sea-land breezes exert a marked effect on aerosol composition in this southernmost area.At night, offshore breezes carry aerosols produced on land to the sea.These aerosols aggregate with marine particulates and primary marine aerosols to form complex sea aerosols.Subsequently, these sea aerosols are transported to the land through onshore breezes during the day.Longrange transport may also contribute to aerosol composition in southern Taiwan, though evidence is scarce (Tsai et al., 2012;Lin et al., 2013;Tsai et al., 2015).Few studies have explored whether products from biomass burning in Southeast Asia can be transmitted to southern Taiwan through longrange transport and fresh marine air.Therefore, in the present study we investigate the properties and sources of aerosols collected at a site on the west coast of Pingtung, Taiwan.

Study Area and Sampling
The aerosol collection site was located in open grassland next to the northern student dormitory at the Museum of Marine Biology (22°03′17.27′′N,120°41′57.91′′E;Fig. 1) in Checheng Township, Pingtung County.The site is a remote coastal environment (140 m from the seashore) and therefore directly exposed to the marine atmosphere.The nearest main road is 2 km away and hence the direct effects of traffic and industrial emissions could be minor.The southeast corner of the area is the Hengchun Township and Kengting National Park which are extremely lowdevelopment areas.The nearest major urban and industrial center, the city of Kaohsiung, is 80 km to the north, thus the background air quality is clean.
Samples were collected between February 24, 2013 and April 14, 2013.At this time the study site is directly affected by biomass burning products transported from Southeast Asia and by the northeast monsoon (Lin et al., 2013).Surface wind speed at an altitude of about 4.5 m above ground level at the open coastal site was > 11.3 m s -1 for 25%-30% of the time.Daytime winds were from the northwest and nighttime winds were from the east, indicating a noticeable land-sea breeze effect.
Daytime and nighttime aerosol and gas samples were collected using a Versatile air pollutant sampler (VAPS, URG-3000K).The VAPS with three filter packs and two denuders was designed to avoid sampling errors and artefacts (Matsumoto and Okita, 1998;Tsai and Perng, 1998;Hsieh et al., 2008;Tsai et al., 2015).The VAPS consists of two parallel fine particle (PM 2.5 ) collection channels (noted as L1, the right channel, and L3, the left channel) operated simultaneously at a sampling flow rate of 15.0 ± 0.2 L min -1 , and one coarse particle (PM 10 ) sampler (noted as L2, the center channel) operated at 2.0 ± 0.1 L min -1 .Among the collection devices, one channel (noted as L3) prior to PM 2.5 particles collection equipped with a Na 2 CO 3 -coated annular denuder and a CH 2 COONa-coated annular denuder was operated on a 47-mm Teflon filter (Zefluor, Pall) (1st stage filter).PM 2.5 particles were collected by the front Teflon filter, while the particle-stripped downstream passed on to a 47-mm nylon filter (Nylasorb, Pall) (2nd stage filter), which is a reactive filter for trapping HNO 3 gas that has volatilized from the front Teflon filter.A 47-mm backup NaCl-coated quartz filter (2500 QAT-UP, Pall) (3rd stage filter) for collecting HNO 3 gas probable escaped from nylon filter to assure the complete collection of volatilization of ammonium nitrate behind the Teflon filter and nylon filter.The sum of the collected and then measured nitrate on the Teflon, nylon and quartz filter is the complete PM 2.5 nitrate (Tsai and Perng, 1998;Bai et al., 2003;Hsieh et al., 2008;Tsai et al., 2015).On which coarse particles were collected.The other parallel channel (noted as L1) for PM 2.5 collection was operated on another 47-mm pre-combusted quartz filter for determining carbonaceous materials in PM 2.5 (2500 QAT-UP, Pall).The L2 collection channel was also fitted with a 47-mm pre-combusted quartz filter (2500 QAT-UP, Pall) for coarse particle collection.Because the aerosol concentrations in this remote area might be low, each sample was collected over a 3-day period (three daytime samples, 0700-1900, and three nighttime samples, 1900 to 0700).A total of 14 sampling sets were collected for each group category (daytime/nighttime, course/fine, and aerosol/ gas), for 84 samples in total.

Sample Handling
Filters were conditioned at 35 ± 1% RH for 24 h and then weighed before and after sample collection.All weight measurements were performed at 50 ± 3% RH using a Sartorius CP2P analytical balance with a sensitivity of 1 µg and a Mettler Toledo AT261 analytical balance with a sensitivity of 10 µg, and were repeated three or more times to ensure reliability.Blank filters were prepared by purging in 99.995% pure nitrogen for 30 s and then processing as for sample-containing filters.Quartz fiber filters were heated prior to use for 4 h at 950°C to reduce carbon content.
The collecting solution in the first denuder, used to collect HNO 3 , HNO 2 , SO 2 , HCl and oxalic acid gas, was 1 g Na 2 CO 3 + 1 g glycerine in a 50 mL:50 mL methanol/water.The solution in the second denuder was 1 g CH 2 COONa in 100 mL 100% methanol, and was used to collect NH 3 (Hsieh et al., 2008;Tsai et al., 2015).The inner walls of the denuder tubes were coated with denuder solution (10 mL added), the excess solution was decanted and the tube was dried for 10 min with 99.995% pure N 2 gas.Both ends of the tube were then capped with silicone plugs.The quartz filter (3rd stage filter in L3) was then treated by immersing it into a collecting solution of 1 g NaCl in 1:9 (V/V) 100 mL methanol/water in order to capture nitrate potentially missed by the Teflon filter (1st stage filter) and nylon filter (2nd stage filter) (Hsieh et al., 2008;Tsai et al., 2015).

Chemical Analysis and Quality Assurance
Reagents were of analytical grade, obtained from Merck (Darmstadt, Germany) or from Sigma-Aldrich (St. Louis, MO, USA), and were used without further purification unless otherwise indicated.The sample-containing filters and unexposed blanks were stored in petri dishes placed inside an unlit refrigerator at -18°C to prevent loss of semi-volatile species, especially carboxylates, ammonium and nitrate.To analyze carboxylates, saccharides, cations and anions, the filter paper (L3) was placed in a PE bottle, 8.0 mL of Milli-Q deionized water (specific resistivity > 18.2 MΩcm at 25°C, Millipore Direct 8/16 System) was added, and the contents were shaken (TS-500 Shaker, Yihder, Taiwan) in an unlit refrigerator at 4°C for 90 min.The liquid was then filtered through a 0.2 µm ester acetate filter and the filtrate was characterized using IC, following a slightly modified version of the method of Hsieh et al. (2007Hsieh et al. ( , 2008)).The IC (DX-600, Dionex) was equipped with a gradient pump (Model GP50), an ASRS-Ultra anion selfregenerating suppressor, a conductivity detector (CD25), a Spectrasystem automated sampler (AS1000) with 2 mL vials, and a Teflon injection valve using a 1000 µL sample loop, in combination with an AS11 analytical column (250 mm × 4 mm I.D.) with an AG11 guard column (50 mm × 4 mm I.D.) and an anion trap column (ATC-3) and a 5-100 (gradient) mM NaOH and 100% methanol eluent.Flow rate was maintained at 2.0 mL min -1 during the carboxylate analyses.This method allowed for the analysis of formate, acetate, malonate, succinate, maleate, fumarate, tartarate, malate and glutarate (Tsai et al., 2015).
Anhydrosugars and sugar alcohols were determined using a Dionex ICS-2500 IC high-performance anion exchange chromatogram equipped with pulsed amperometric detection (HPAEC-PAD), a GP50 gradient pump coupled to a Teflon injection valve with 400-mL sample loop, a CarboPac MA1 guard column (50 mm × 4 mm I.D.) and anion-exchange analytical column (250 mm × 4 mm I.D.), a Dionex ED50 electrochemical detector with a gold working electrode and a pH electrode as reference, and using 200-600 (gradient) mM NaOH eluent at a flow rate of 0.4 mL min -1 , following a slightly modified version of the method of Engling et al. (2006), Caseiro et al. (2008) and Tsai et al. (2010).NaOH solutions were protected from exposure to atmospheric CO 2 to prevent dissolution of carbonate into the NaOH hindering separation.Also, to prevent "poisoning" of the electrode surface, it was continuously cleaned via different potential being applied followed by a regeneration step (Tsai et al., 2015).
Two one-eighths of the quartz filter for PM 2.5 collection (L1) were cut off and stored in a freezer to prevent organic carbon (OC) loss.One was analyzed using a Heraeus CHN-O-Rapid Elemental Analyzer (EA) to determine total carbon (TC) content.Acetanilide was used as the standard to prepare the calibration curve.Analysis consisted of heating for 1.5 min with an oxidation tube temperature of 950°C and a reduction tube temperature of 600°C.A total carbon detector was used to detect the CO 2 content of particles carried by helium after heat treatment.The output signal value was converted for input into an integrator and weight percentage of carbon to the sample was determined, allowing the measurement of TC and EC content.The other one eighth filter was heated at 340°C for 90 min to expel OC (Cadle and Groblicki, 1982;Tsai and Chen, 2006a;Tsai et al., 2015) and the same process used to obtain TC was applied to obtain EC.
The method detection limit (MDL) for individual species was defined as a signal-to-noise ratio from multiple injections of a lowest-concentration standard of 3:1.MDLs of four chemical compound groups measured using IC systems and carbon determination are listed in Table 1.MDLs for cationic species have been recorded between 0.007 µg m -3 (NH 4 + ) and 0.015 µg m -3 (Na + ), for anionic species between 0.009 µg m -3 (F -) and 0.014 µg m -3 (SO 4 2-), for carboxylates between 2.44 ng m -3 (oxalate) and 2.91 ng m -3 (acetate), for sugar alcohols and sugars between for 0.075 ng m -3 (erythritol) and 1.973 ng m -3 (trehalose), for anhydrosugars between 0.091 ng m -3 (mannosan) and 0.276 ng m -3 (galactosan), and for carbon, below 0.092 µg m -3 at a nominal volume of 32.4 m 3 for each typical 36 h threedaytime or three-nighttime PM 2.5 and gaseous sampling period.Average recoveries of inorganic ions ranged from 95.7% for NH 4 + to 103.6% for Ca 2+ .Average recoveries of Cl -, NO 3 -, SO 4 2-, K + and Mg 2+ were near 100% with < 2% relative standard deviations (RSDs).About 95.7-99.8% of carboxylic acid was recovered with < 3.8% RSD.About 97.9-105.4% of sugar alcohols and sugars, and anhydrosugars were recovered with < 4.5% RSDs.Determination of all species in the study achieved a high degree of accuracy and reproducibility.Additionally, because pre-treatment did not involve organic solvent, possible analytic losses and contamination in extraction steps were minimized.
Excess sulfate, specifically non-sea-salt sulfate (NSS-SO 4 2-), in aerosols was calculated by subtracting the amount of SO 4 2-in sea salt from that of total SO 4 2-measured in the ambient environment.The amount of marine sulfate was estimated from the SO 4 2-/Na + mass ratio for bulk seawater.Sea-salt sulfate (SS-SO 4

2-
) in µg m -3 was estimated by /Na + mass ratio in Taiwan west-coast seawater (Cheng and Tsai, 2000).The concentration of aerosol NSS-SO 4 2-in µg m -3 was calculated as NSS-SO 4 . Aerosol Na + in the atmosphere was assumed to be derived only from sea salt aerosols.

Backward Trajectory Analyses and Ancillary Parameters
Using correlation analysis, derived from StatSoft STATISTICA Edition 7.0, we found the relationships between aerosol components.Furthermore, typical backward trajectories for air parcels that affected Hengchun's remote coastal atmosphere were conducted using the Hybrid Single-Particle Lagrangian Integrated Trajectory model (HYSPLIT) (Version 4.9) (Draxler and Hess, 1998;Draxler, 2003;Draxler and Rolph, 2003), based on wind fields generated by the medium-range forecast model (MRF) and developed by the National Oceanic and Atmospheric Administration (NOAA), USA.We computed backward trajectories at intervals of 12 h for the five to eight days prior to sampling at three different altitudes above the study site (0 m, 1000 m, and 2000 m).Backward air mass trajectory pathways were compared against locations of burning sites over a 10-day period as detected by the MODIS Fire Locating and Modeling of Burning Emissions System (FLAMBE, Reid et al., 2009Reid et al., , 2012)).The distribution of fires in East Asia region was focused during the period of February and April 2013.Each marked red dot indicates a location where MODIS detected at least one fire during the intensive biomass burning season.

Differences in Gaseous Pollutants
Fig. 2 illustrates changes in background concentrations of ambient gases at the study site.The SO 2 concentration was the highest of the acidic gases, and was higher during daytime (0.80 ± 0.56 µg m -3 ).HCl exhibited the secondhighest concentration, and was similarly higher during the daytime (0.76 ± 0.35 µg m -3 ).These acidic gases were present in higher concentrations during the day because of nonlocal industrial and waste incineration emissions during the daytime.HCl was in part generated through reactions between sea salts and acidic gases (Kerminen et al., 1997(Kerminen et al., , 1998;;Ohta and Okita, 1990;Kerminen et al., 2000).SO 2 concentrations were 6.69-9.98 µg m -3 in the Tainan urban area (Tsai et al., 2015), 15.5 ± 8.9 µg m -3 in New York, Manhattan (Bari et al., 2003), and 5.87 ± 4.96 µg m -3 in a German forest region (Plessow et al., 2005).These levels are far greater than in southernmost Taiwan.Levels at other remote sites unaffected by industrial or wood-burning emissions such as Ali Mountain (0.14 µg m -3 ) (Weng, 2006) and the northeast Atlantic Ocean (0.26~1.32 µg m -3 ) (Quinn et al., 2000) confirm that SO 2 concentrations are markedly lower in remote areas compared with areas with high anthropogenic activity.While HNO 2 concentrations at night (0.34 ± 0.61 µg m -3 ) were markedly higher than during the day (0.03 ± 0.06 µg m -3 ), HNO 3 concentrations were substantially higher during the day (0.67 ± 0.37 µg m -3 versus 0.45 ± 0.14 µg m -3 ).This implies that in the presence of sunlight, the NO x emitted from non-traffic sources in the area undergoes photochemical reaction, causing the conversion of intermediates (HNO 2 ) to HNO 3 (Sjödin and Ferm, 1985;Lin and Cheng, 2007), thereby leading low HNO 2 and high HNO 3 daytime concentrations.

Variations in Chemical Compositions in PM 2.5
Fig. 3 and Table 2 show the composition of PM 2.5 at the coastal study site.The PM 2.5 mass concentrations recorded (16.16 ± 5.30 µg m -3 ) was lower than along the Taichung coastal region of Taiwan (north of the study site) and the urban areas of central Taiwan (Cheng and Tsai, 2000).It was similar to PM 2.5 concentrations recorded in Ali Mountain (16.7 ± 5.5 µg m -3 ) (Cheng and Tsai, 2000) and was slight higher than those at the same location during non-Asian dust storm (non-ADS) period in the spring of 2010 (10.9 ± 4.45 µg m -3 ), but was lower than those influenced during ADS (23.1 µg m -3 ) (Tsai et al., 2012).Furthermore, the PM 2.5 mass concentrations in the Tainan urban area of southern Taiwan during the autumn ranged between 15.9 and 25.6 µg m -3 (Tsai et al., 2015).In some periods, the PM 2.5 concentrations on the Taichung coast (50.2 ± 25.5 µg m -3 ) were considerably higher than those at the study site.Concentrations in the urban areas of central Taiwan (71.2 ± 39.8 µg m -3 ), however, were 4.4-fold higher.This is mainly attributable to the contributions from more extensive anthropogenic activities in central Taiwan (Cheng and Tsai, 2000).
Of inorganic species SO 4 2-concentrations were the highest in PM 2.5 (3.40 ± 1.80 µg m -3 ), followed by NH 4 + (1.50 ± 1.19 µg m -3 ).Concentrations of these two species at the study site were markedly lower than in the Yangtze River Delta region (9.6 ± 6.1 and 4.3 ± 3.5 µg m -3 ) (Meng et al., 2014).In descending order of concentration, the remaining species were Na + (1.14 ± 0.74 µg m -3 ), NO 3 -(0.97± 0.71 µg m -3 ), and Cl -(0.88 ± 0.58 µg m -3 ).NO 3 -is usually the third most abundant species in PM 2.5 in urban and suburban areas in southern Taiwan (Tsai and Chen, 2006a, b;Hsieh et al., 2008;Tsai et al., 2015).At the study site, however, Na + was the third most abundant species, with a concentration higher than that of NO 3 -illustrating that the composition of aerosols in the coastal region of southern Taiwan differs from that typical in urban areas (Remoundaki et al., 2013).The composition of PM 2.5 observed in this study was similar to that in the same location during the spring of 2010 (Tsai et al., 2012).Sea spray carries NaNO 3 and naturally emitted NO 3 -and therefore the concentration of NO 3 -forming as secondary photochemical product is low (Hsieh et al., 2009).The concentrations of crust elements Mg 2+ and Ca 2+ were 0.09 ± 0.06 and 0.25 ± 0.17 µg m -3 , respectively.The PO 4 3-aerosol concentration was 0.12 ± 0.12 µg m -3 , indicating that the coastal study site is slightly affected by fertiliser use in the neighbouring regions during the spring.
Average PM 2.5 carboxylate concentration was 0.50 ± 0.24 µg m -3 and acetate was present in the highest concentrations (0.32 ± 0.18 µg m -3 ), followed by Ox (0.18 ± 0.08 µg m -3 ).The mass ratio of Ox/non-sea-salt sulfate (Ox/NSS-SO 4 2-), which represents the photochemical formation potential of organic acids and inorganic salts, was 6.15 ± 2.28% compared with average Ox/NSS-SO 4 2-ratios in urban and suburban areas of southern Taiwan of 3.19 ± 0.36% (Tsai et al., 2015) and 4.45 ± 1.66% (Hsieh et al., 2008), respectively.Concentrations of acetate and oxalate at the study site were higher than in the urban areas of southern Taiwan.These outcomes indicate a strong photochemical formation potential at the study site and that some organic acids in this region are converted to the final products of carboxylates through such reactions.Additionally, it should be noted that the origin of the Ox at the study site and in urban regions differs: The high concentrations of Ox and acetate at the study site were mainly attributed to natural marine emissions as well as terrestrial plant and soil microbe metabolic emissions, whereas in urban regions traffic emissions play a role (Wang et al., 2007;Guo et al., 2015).
The concentration of total saccharides in PM 2.5 at the study site was 590 ± 422 µg m -3 .myo-Inositol was the main saccharide species, as it is in urban areas of Taiwan, but it was present in substantially higher concentrations than in these urban areas, and this implies rich soil biota activity in the open grasslands and low-development regions around the study site (Simoneit et al., 2004;Caseriro et al., 2007;Tsai et al., 2015).Concentrations of the remaining saccharide, in descending order, were trehalose (242 ± 213 ng m -3 ), glucose (2.67 ± 6.42 ng m -3 ), and mannitol (0.11 ± 0.58 ng m -3 ).The large standard deviations in the concentrations of these species are mainly attributable to the rich natural ecology of the region, which features natural emissions from lichens, fungi, and soil biota (Caseriro et al., 2007).Levo concentrations (8.92 ± 16.92 µg m -3 ranging from nondetectable to 79.4 ng m -3 ) were 1.85-fold less than in the coastal urban regions of southern Taiwan (16.5 ± 12.7 ng m -3 ) (Tsai et al., 2015) and also exhibited large standard deviation, indicating that biomass burning events were infrequent in the region of the study site and that associated pollutants were rarely transported to the study area and hence that the atmosphere of the coastal region around the study site was affected only slightly by biomass burning.
The daytime concentration (764 ± 456 ng m -3 ) of total saccharides was higher than that at night (415 ± 310 ng m -3 ).Of the saccharides, myo-inositol exhibited the highest concentrations in the daytime (405 ± 350 ng m -3 ) and nighttime (261 ± 231 ng m -3 ), followed by trehalose (daytime 337 ± 236 ng m -3 and nighttime 147 ± 139 ng m -3 ).The daytime concentrations of both species were 1.6-to 2.3-fold higher than those at night.Mannitol (0.22 ± 0.82 ng m -3 ) was only observable in PM 2.5 at night.The daytime Levo concentrations (12.7 ± 21.2 ng m -3 ) were 2.44-fold higher than those observed at night (5.20 ± 10.8 ng m -3 ), indicating that the long-range transport of biomass burning products to the study site mainly occurs during the day.
The daytime concentrations of total saccharides (127.6 ± 34.6 ng m -3 ) in PM 2.5-10 were higher than those observed at night (70.2 ± 21.2 ng m -3 ).The daytime concentrations of the main saccharide species were also higher than at night, but the daytime erythritol concentrations (0.77 ± 2.55 ng m -3 ) Fig. 6.Average PM 2.5-10 component concentrations during the daytime and nightime at the coastal study site in spring 2013.
were lower than at night (1.94 ± 4.45 ng m -3 ), although the standard deviations were large.The main source of erythritol might be from soil biota metabolites (Caseriro et al., 2007).No detected Levo was observed in the daytime or nighttime coarse saccharides.
Figs. 2, 5 and 6 summarized lower daytime NH 3 and higher daytime NH 4 + aerosol indicated NH 3 gas was converted to NH 4 + particles through photochemical reactions during the day and more accumulated NH 3 at night without photochemical mechasism was emitted from local biogenic sources in the coastal region.Meanwhile, higher daytime SO 2 and SO 4 2-aerosol showed SO 2 arrived at the study site from nonlocal anthropogenic activities and, simultaneously, was converted to SO 4 2-aerosol through photochemical reactions during the day.

Aerosol Composition Ratio
Table 2 shows the amount of PM 2.5 as a fraction of the PM 10 (expressed as the ratio of PM 2.5 /PM 10 ) and the corresponding concentration ratios of daytime to nighttime aerosol species and coarse to fine aerosol particles.The PM 2.5 /PM 10 of 61.11% was higher than that observed in the same location during the spring of 2010 (PM 2.5 /PM 10 : 41%) (Tsai et al., 2012), but lower than in urban and suburban areas (Tsai and Chen, 2006a;Souza et al., 2014).Except for Na + , Cl -, Mg 2+ , arabitol and mannitol, the aerosols at the study site were present mainly in fine aerosol particles.Specifically, more than 80% of SO 4 2-, NH 4 + , oxalate, trehalose, and Levo were concentrated in fine particles, indicating that the fine particles in this area were composed mainly of secondary photochemical products.Moreover, Levo was only observable in PM 2.5 , implying that only fine particles in biomass burning products are transported to the study area.The region around the study site is therefore simultaneously affected by primary and secondary pollutants.
Furthermore, the ratios of arabitol and mannitol-the indicators of bioaerosols (Dahlman et al., 2003;Müller et al., 2005;Caseiro et al., 2007;Bauer et al., 2008;Chen et al., 2013), emitted through lichen and fungal activities in coastal regions-were only 14% and 16%, respectively.These two species were mainly observed in PM 2.5-10 , and the daytime concentrations of the species were higher than those at night, indicating that these organisms are more active during the day (Elbert et al., 2007).Daytime erythritol concentrations in PM 2.5 were also higher than at night.Erythritol is from soil biota metabolites (Caseiro et al., 2007) and is transported from fire locations (Tsai et al., 2015) during the day especially, aided by onshore winds.Fig. 7 shows the percentages of components in PM 2.5 .The PM 2.5 was mainly composed of inorganic salts, OC, and EC.Of the inorganic salts, SO 4 2-composed 21.0 ± 11.2%, followed by NH 4 + at 9.27 ± 7.34%.The sea salts, Na + and Cl -, accounted for 7.05 ± 4.60% and 5.47 ± 3.56%, respectively.OC and EC were mainly observed in PM 2.5 .The OC/EC ratio derived from percentage composition was 1.15, indicating that the carbon-containing compounds of PM 2.5 at the study site were mainly from secondary aerosols generated through photochemical reactions (Turpin et al. 1991;Tsai et al., 2015).Total carboxylates accounted for 27.7 ± 13.5% of OC.Acetate was the main carboxylate species (64.6 ± 36.2%), and Ox was the main species of dicarboxylate, accounting for 35.4 ± 16.8% of total carboxylates and 9.81% of OC.This finding is consistent with that of Kawamura and Sakaguchi (1999), who reported that approximately 15% of OAs in the Pacific Ocean were dicarboxylates, of which Ox was the main species.Total saccharides accounted for 33.0 ± 23.6% of OC. myo-Inositol was the main saccharide species, accounting for 56.5 ± 50.9% of all saccharides, followed by trehalose (41.1 ± 36.1% of all saccharides).The remaining sugar species exhibited trace concentrations and accounted for less than 1.5% of the total saccharide, with Levo accounting for 1.51 ± 2.87% of all saccharides.These findings suggest that the open coastal area of the study site was only slightly affected by trace amounts of biomass burning products.

Comparison of Aerosol Compositions in Various Regions
Table 3 lists the compositions of PM 2.5 in various regions.PM 2.5 from fisheries and agricultural regions in suburban Tainan was mainly composed of NH 4 + (Tsai and Kuo, 2005), whereas in other regions it was primarily composed of SO 4 2-.In the PM 2.5 of the study site, Na + concentrations were greater than Cl -, whereas an opposite trend was observed for the eastern coastal regions of the Mediterranean Sea (Koçak et al., 2004).The PO 4 3-concentrations in PM 2.5 at the study site were 2.4-fold higher than those in northern Taiwan (Wang, 2001), indicating that agricultural activity near the study site contributed more PM 2.5 compared with this area in northern Taiwan.In comparison to the study site, Levo concentrations were 35-fold higher (320 ± 440 ng m -3 ) in Brazil where biomass burning activity occurred near the sampling site (Urban et al., 2014).The Levo concentrations in PM 2.5 on Jeju Island were 4.3-fold higher, again indicating that the island was considerably more affected by biomass burning than the study site (Fu et al., 2012).Moreover, except for trehalose, other saccharide species on Jeju Island also exhibited higher concentrations than at the study site.In the Howland Forest, the main species in PM 2.5 was glucose, the concentration of which was 14-and 2-fold higher than observed at the study site and Jeju Island, respectively, indicating that forests release more glucose (Medeiros et al., 2006).

Aerosol Cl-Depletion
Particulate chloride depletion, i.e., chlorine loss or Cldeficit, resultsd when NaCl in sea salt aerosol reacts with H 2 SO 4 , HNO 3 , or organic acids to form Na 2 SO 4 , NaNO 3 or sodium-bound organic acids and gaseous HCl (Clegg and Brimblecombe, 1985;Ohta and Okita, 1990;Kerminen et al., 1998;Zhuang et al., 1999;Kerminen et al., 2000;Quinn et al., 2000).After sea salt particles are generated from the sea surface, they react with ambient acidic gases (e.g., H 2 SO 4 , HNO 3 , and organic acids) to form HCl gas, resulting in Cldepletion.The Cl-depletion percentage (%Cl dep ) can be calculated as follows (Kerminen et al., 1998;Quinn et al., 2000): where ([Cl -]/[Na + ]) original is the seawater Cl -to Na + mass ratio of a consistent 1.8, which would be the expected ratio if no depletion occurred (Cheng et al., 2000;Quinn et al., 2000), and [Cl -] and [Na + ] are the measured Cl -and Na + mass concentrations for all the real samples.Fig. 10 shows the daily Cl-depletions for PM 2.5 and PM 2.5-10 .The mean Cl-depletion for PM 2.5 (53.1%) was higher than for PM 2.5-10 (26.0%), indicating that Cl -in PM 2.5 is likely to react with acidic gases and other precursors to form secondary products (Sellegri et al., 2001;Hsu et al., 2007a).The Cl -/Na + equivalent ratios of PM 2.5 and PM 2.5-10 were 0.55 and 0.94.Cl-depletion and Cl -/Na + equivalent ratios were inversely correlated.A Cl-depletion of 85.5% in PM 2.5 and and 40.8% in PM 2.5-10 in the Gaomei Wetlands (Chen, 2003) accords with the results of our study, verifying that the Cl -/Na + equivalent ratio of PM 2.5 was greater than for PM 2.5-10 .However, the Cl-depletion of PM 2.5 in the Gaomei Wetlands was markedly higher than that at our study site, implying that the Gaomei Wetlands are more strongly affected by photochemical products released from neighbouring traffic and industrial emissions.
Fig. 11 shows the daytime and nighttime Cl-depletion ratios of PM 2.5 and PM 2.5-10 at the study site.The daytime and nighttime Cl-depletion ratios of PM 2.5 were 56.3% and 49.8%, respectively, and the daytime and night-time Cl - /Na + equivalent ratios were 0.51 and 0.58, respectively.Cldepletion was greater during the day mainly because daytime onshore breezes carry sea salts that react with acidic gases inland, inducing Cl-depletion.Additionally, the photochemical environment during the day facilitates photochemical reactions, thus causing additional Cldepletion.The daytime and nighttime Cl-depletion ratios of PM 2.5-10 were 29.7% and 22.4%, respectively, less than in PM 2.5 .Furthermore, low PM 2.5-10 Cl-depletion resulted in daytime and nighttime Cl -/Na + equivalent concentration ratios of 0.92 and 0.97, respectively, approximating the theoretical Cl -/Na + equivalent concentration ratio (1.00).

Ratios of Specific Chemical Compounds in PM 2.5
Table 4 lists aerosol indicator values at the study site and various other regions.The Ox/NSS-SO 4 2-ratio, which represents photochemical formation intensities of organic acids and inorganic salts, was 6.15 ± 2.28% at the study site.This compares with 4.45 ± 1.66% and 3.19 ± 0.36% in suburban areas of southern Taiwan (Hsieh et al., 2008;Tsai et al., 2015).The higher ratio at the study site indicates a strong potential for the formation of Ox (the final product of dicarboxylates) in the study area, despite no noticeable pollution source being present.The daytime ratio was slightly higher, indicating that the potential for formation of organic acid stronger during the day.

CONCLUSIONS
In this study, we conducted a 7-week observation of the atmospheric environment in a remote region on the southwest coast of Pingtung, Taiwan, to investigate the properties and potential sources of marine aerosols.Among the ambient gases observed, SO 2 exhibited the highest concentrations.The daytime HCl concentrations were higher than those observed at night.A portion of HCl was from Cl-depletion, which occurs when sea salts react with acidic gases.The daytime HNO 2 concentrations were markedly lower than those observed at night, whereas the daytime HNO 3 concentrations were higher.This is attributable to photochemical reactions during the day, converting HNO 2 to HNO 3 .Additionally, photochemical reactions also reduced daytime NH 3 concentrations while increasing daytime NH 4 + concentrations.NH 4 + concentrations at the study site were, however, still lower than those in urban and suburban areas of southern Taiwan.
Daytime PM mass concentrations were higher than those observed at night.The mean mass ratio of PM 2.5 /PM 10 was 61.1%.Among the inorganic salts in PM 2.5 , SO 4 2-exhibited the highest concentration.NSS-SO 4 2-accounted for 92.2 ± 49.8% of PM 2.5 SO 4 2-, indicating that PM 2.5 SO 4 2-is mainly generated through secondary photochemical reactions.The sea salts (Na + and Cl -), crustal matter (Mg 2+ ), and bioaerosols (arabitol and mannitol) emitted by lichen and fungal activities were observed mainly in coarse particles.PM 2.5 at the remote coastal study site contained higher Na + concentrations compared with NO 3 -and this contrasted to urban areas of southern Taiwan, where NO 3 -concentrations were higher than NH 4 + .The lower Cl -concentration in PM 2.5 compared with the aforementioned species was attributed to Cldepletion.The daytime and nighttime Cl-depletion ratios of PM 2.5 were 56.3% and 49.8%, respectively.PM 2.5-10 exhibited lower daytime (29.7%) and nighttime (22.4%)Cldepletion ratios than PM 2.5 .The higher daytime Cl-depletion indicated that the daytime photochemical environment enabled additional Cl -reactions, resulting in increased daytime Cl-depletion.
Total carboxylates accounted for 27.7 ± 13.5% of OC in PM 2.5 .Acetate was the main carboxylate species and accounted for 64.4 ± 36.2% of total carboxylates.Total saccharides accounted for 33.0 ± 23.6% of OC in PM 2.5 .The main saccharide species in PM 2.5 and PM 2.5-10 were myo-inositol and trehalose, both of which had higher concentrations during the day than at night.Additionally, the daily combined concentration of myo-inositol and trehalose accounted for 97.6% and 90.8% of PM 2.5 and PM 2.5-10 , respectively, indicating that the plant ecology distributed throughout the grassland and coastal regions around the study site generates diverse soil biota emission products.Daytime emissions of myo-inositol and trehalose were higher than those at night.The mean daily mass ratio of Ox/NSS-SO 4 2-, which reflects the potential for organic acids and inorganic salts to form via photochemical reactions, was higher than in urban areas of southern Taiwan.
Levo, a biomass burning product, was detectable only in PM 2.5 and at a concentration of 8.92 ± 16.92 ng m -3 .Levo accounted for 1.51 ± 2.87% of all saccharides and the ratio of Levo/OC was 5.04 ± 8.72‰.Airflow back trajectories revealed that during spring, the airflow in this region is from the concentrated biomass burning areas of Southeast Asia and southern China.During this period, the PM 2.5 Levo concentration at the study site can reach 79.4 ng m -3 (Levo/OC ratio = 27.5‰), and the daytime Levo concentrations are 2.44-fold higher than those observed at night.This suggests that biomass burning products are mainly transported long-range to Hengchun during the daytime.However, airflows arriving at the study site do not continuously originate from fire locations.When unpolluted airflows arrived at the study site, no Levo was observed in PM 2.5 .Overall, the atmospheric environment above the remote coastal region of southwest Pingtung is affected only slightly by biomass burning.However, this region is influenced by a strong land-sea breeze effect and exhibits a high concentration of photochemical products, particularly low-molecular-weight organic acids with high formation potentials.

Fig. 1 .
Fig. 1.Geographical location of sampling site, shown in a red circle, and wind roses in the Hengchun Peninsula, southernmost Taiwan.

Fig. 2 .
Fig. 2. Background concentrations of gaseous pollutants at the coastal study site in spring 2013.

Fig. 3 .
Fig. 3. Concentrations of PM 2.5 inorganic salts, carboxylates, and saccharides at the coastal study site in spring 2013.

Fig. 5 .
Fig. 5. Average PM 2.5 component concentrations during the daytime and nighttime at the coastal study site in spring 2013.

Fig. 7 .
Fig. 7. Percentages of various chemical species detected in daily PM 2.5 at the coastal study site in spring 2013.

Fig. 12 .
Fig. 12. Relevant maps of fires in East/Southeast Asia region during February and April, 2013 and typical air mass backward trajectory pathways to Hengchun Peninsula.(a) Air mass across the fire regions in SE Asia; (b) Air mass across the fire regions in SE/East Asia; (c) Air mass surrounding the China Sea.The red circles are the areas of biomass burning sources.

Table 1 .
Method detection limits (MDLs) and extraction recovery ratios of four chemical compound groups measured using IC systems.
+ in µg m -3 by a factor of 0.231, where the coefficient of 0.231 is the typical SO 4 2-

Table 2 .
Concentrations of inorganic ions and organic species in size-fractionated aerosols and relevant ratios in Hengchun Peninsula, Taiwan.

Table 3 .
Comparison of concentrations of inorganic ions and saccharides in aerosol in the present study with those reported in the literature.

Table 4 .
Average and associated standard deviation of mass ratios of specific species in PM 2.5 during different period intervals in southern Taiwan.OC in this case means water-soluble organic carbon (WSOC). a