Interactive comment on “ Atmospheric aerosol compositions over the South China Sea : Temporal variability and source apportionment

The manuscript presents results from observations of Total Suspended Particulate matter (TSP) from an approximately year long study in a marine region of the northern South China Sea (SCS). Four-day filters were collected and analyzed for major inorganic chemical ionic concentration observed in the marine boundary layer. Several source apportionment methods were then used to both differentiate between aerosol types that contributed to measured values and link them with potential sources. These included correlation between various factors, principal component analysis, positive matrix factorization, and backtrajectory analysis by means of concentration weighted trajectories for various identified sources.


Introduction
Aerosols or particulate matters potentially affect global atmospheric processes, chemistry, cloud formation, acid and nutrient deposition in sensitive ecosystems, affect human health as air pollution and have effects on climate (Chin, 2009;Davidson et al., 2005;Deng et al., 2011;S. Liu et al., 2014;Steyn and Chaumerliac, 2016;Zhang et al., 2011).Aerosols have many sources.Primary aerosols, emitted directly from natural and anthropogenic sources, include terrestrial dust from weathering, wind-blown sand and minerals, sea spray, combustion processes and biological emissions (Chin, 2009;Davidson et al., 2005;Rinaldi et al., 2010; Published by Copernicus Publications on behalf of the European Geosciences Union.Pavuluri et al., 2015).There are also secondary aerosols, which form from condensable atmospheric gases, such as SO 2 , NO x , NH 3 , and volatile organic compounds (Chin, 2009;Kolb and Worsnop, 2013;Xiao et al., 2012aXiao et al., , 2014b;;Zhang et al., 2011).Therefore, aerosols are composed of various chemical components (Xiao and Liu, 2004) and contain sulfate, nitrate, ammonium, mineral elements, black carbon and particulate organic matter (Zhang et al., 2007(Zhang et al., , 2011;;Chin, 2009;Pavuluri et al., 2015;Steyn and Chaumerliac, 2016).Because of rapid economic and industrial development in the last few decades in southern and eastern Asia, many aerosols and their precursors released by human activities have become a major environmental problem (Steyn and Chaumerliac, 2016).Aerosols from anthropogenic emissions can be transported long distances from polluted regions to remote open oceans, which is well recognized as a major pathway for the supply of anthropogenic material to ocean surface waters (Duce et al., 2008;Lawrence and Lelieveld, 2010;Jung et al., 2012).
The South China Sea (SCS) is in the tropical-subtropical rim of the northwestern Pacific Ocean and is one of the largest marginal seas in the world (Fig. 1).The SCS is adjacent to several rapidly growing Asian economies, including China, the Philippines, Malaysia, Vietnam and Indonesia (Kim et al., 2014).Consequently, the sea receives substantial amounts of aerosols from surrounding regions through longrange atmospheric transport (Fig. 1; Atwood et al., 2013;Jung et al., 2012).The SCS has a monsoon climate, with a northeast monsoon in boreal winter and spring and a southwest monsoon in boreal summer and autumn (Cui et al., 2016).Thus, aerosol optical thickness (AOT) shows spatial and seasonal variations, with higher AOT in the northern SCS during the boreal cool season and higher AOT in the southern SCS in the boreal warm season (Fig. 1).The SCS receives dust and pollution from the northeast in winter and spring, e.g., from China and Japan (Y. Liu et al., 2014;S.-H. Wang et al., 2013;Xiao et al., 2015); during summer and autumn, the SCS receives particulate and gas phase pollutants from biomass burning in the southwest of SCS, e.g., from Malaysia and Indonesia (Atwood et al., 2013).However, the observational data on aerosol chemistry over the SCS are very sparse (Xiao et al., 2015).
To get better understanding of potential sources, source contributions and spatiotemporal variations of marine aerosols over the SCS, total suspended particulates (TSPs) were continuously collected at Yongxing Island from March 2014 to February 2015.The concentrations of major inorganic water-soluble ions were determined.Furthermore, back trajectory, concentration-weighted trajectory (CWT) and positive matrix factorization (PMF) analyses were also used to identify and apportion the main sources of aerosols and their chemical composition over the SCS.

Study site
Aerosol samples were collected from March 2014 to February 2015 on the rooftop of Xisha Deep Sea Marine Environment Observation and Research Station, South China Sea Institute of Oceanology, Chinese Academy of Sciences (SCSIO-CAS).This station is at the Yongxing Island (YXI; Fig. 1; 16.83 • N, 112.33 • E).This island has a tropical monsoon climate, with northeast monsoons in winter and spring, and southwest monsoons in summer and autumn (Xiao et al., 2015).The island is located in the high aerosol concentrations area of the northern SCS during winter and spring, and also at the periphery of the high concentrations area of the southern SCS during summer and autumn (Fig. 1).Therefore, pollutants can arrive from northeastern Asia in winter and spring, and Southeast Asia in summer and autumn (Y.Liu et al., 2014;S.-H. Wang et al., 2013;Xiao et al., 2015;Atwood et al., 2013).It is also influenced by local SCS marine sources.More information about Yongxing Island is given elsewhere (Xiao et al., 2015(Xiao et al., , 2016)).In the present study, the sampling period was divided into three seasons based on wind direction and back trajectories.The cool season was March, April, October, November, December 2014 and January, February 2015, with principal air masses from the northeast and 25.9 • C in temperature, and the warm season was June-September 2014, with air masses primarily from the southwest and 29.7 • C in temperature.The transition season was May 2014, with local and short air masses primarily from the southeast.Yongxing Island has an annual average temperature of 27.7 ± 2.7 • C, relative humidity (RH) of 80 ± 7 % and wind speeds of 3.6 ± 1.8 m s −1 .There were strong seasonal variations between March 2014 and February 2015 (Fig. 2).Accumulated annual rainfall was 1526 mm during this period, with about 30 % occurring in the cool season (Fig. 2).
In order to identify the effect of sulfate, dust and smoke on Yongxing Island, we obtain NAAPS Global Aerosol Model data (gif format, https://www.nrlmry.navy.mil/aerosol/#aerosolobservations) between March 2014 and February 2015 to synthesize 1 year of the picture (gif format; Fig. S1 in the Supplement).

Sample collection and chemical analyses
Aerosol was collected on quartz filters using a special highflow rate (1.05 ± 0.03 m 3 min −1 ) KC-1000 sampler, which was installed 1 m above the building's (about 15 m) roof surface of the station of SCSIO-CAS.The sampling time was nominally 96 h (4 days one sample).All samples were stored in a refrigerator at −20 • C until analysis in the laboratory.
In the laboratory, one-eighth filters were cut and placed in a clean 50 mL Nalgene tube with additional ultrapure water.The extract was filtered using pinhole filters, which were then  rinsed twice with 5 mL ultrapure water.The extract and rinse were put into 50 mL tubes together and stored in a refrigerator at −20 • C until chemical analyses.Aerosol water-soluble major anion concentrations (F − , Cl − , NO − 3 , SO 2− 4 , Br − ) were determined by ICS-90 ion chromatography (Dionex, California, USA), and water-soluble metal and non-metal el-emental concentrations (Al, Ca, Fe, K, Mg, Mn, Na, SiO 2 , Sr) were analyzed by an inductively coupled plasma optical emission spectrometer (ICP-OES; Vista-MPX, CA, USA); the NH + 4 concentration was determined by spectrophotometry after treatment with Nessler's reagent, as described in detail information in the Supplement Sect.S1.

Sea salt and non-sea salt contributions
In order to determine the source contributions to the ionic compositions in TSPs, the sea salt ion (SS) and the non-sea salt ion (NSS), are calculated using Na as a reference element for marine origin by the following equations (Eqs. 1 and 2; Xiao et al., 2013): SS(%) = 100(X/Na + ) sea /(X/Na + ) TSPs , (1) where X represents the ionic composition and the data of (X/Na + ) sea come from Keene et al. (1986).

Back trajectories and concentration-weighted trajectories analysis
Back trajectories and CWTs are used to determine the longdistance transport of atmospheric pollutants and regional source areas by the program of TrajStat (version 1.2.26) (Cheng et al., 2013;Wang et al., 2009;Xiao et al., 2014aXiao et al., , 2015)).For each day, 10-day (240 h) back trajectories of air masses (Pavuluri et al., 2015) arriving at Yongxing Island were computed, with the top of the model set to 1000 m above sea level (Xiao et al., 2015).In CWT model, each grid cell receives a weighted concentration obtained by averaging sample concentrations that have associated trajectories crossing the grid cell (Xiao et al., 2015).The detail methods are described in Supplement Sect.S2.

Aerosols over SCS compared with global marine aerosols
The annual average TSP concentrations at Yongxing Island is also lower than the annual average value over the northern Yellow Sea (123.2 µg m −3 ), another Chinese marginal sea (L.Wang et al., 2013).However, the mean TSP concentrations at Yongxing Island are not lower than those at other remote islands or other seas, such as the Indian Ocean (21.1 µg m −3 ), Pacific Ocean (36.7 µg m −3 ), Mediterranean Sea (46.9 µg m −3 ), South Atlantic (39.1 µg m −3 ) and three islands of Okinawa (22.6 µg m −3 in summer and 44.5 µg m −3 in spring) (Arakaki et al., 2014;Balasubramanian et al., 2013;Zhang et al., 2010).The aforementioned annual average TSP and ionic concentrations are comparable to those reported in many remote oceans (Fig. 5), e.g., Hedo, which is at the junction of the East China Sea and northwestern Pacific.The marine ions (Na + and Cl − ) accounted for 53 % of total major ions (Fig. 4).Na + and Cl − concentrations at Yongxing Island were higher than most reported values among all locations and remote sites (Fig. 5), such as the Indian Ocean (3.0 and 4.4 µg m −3 , respectively), Arabian Sea (1.9 and 2.2 µg m −3 , respectively), Oki (3.3 and 4.3 µg m −3 , respectively) and Rishiri islands (1.2 and 2.7 µg m −3 , respectively) in the Sea of Japan, Amsterdam Island (1.4 and 2.3 µg m −3 , respectively) in the Southern Ocean, Bermuda (3.3 and 4.8 µg m −3 , respectively) in the Atlantic Ocean and Hawaii (0.5 and 0.4 µg m −3 , respectively) in the Pacific (Kumar et al., 2008;Okuda et al., 2006;Claeys et al., 2010;Moody et al., 2014;Carrillo et al., 2002).However, Na + and Cl − concentrations were lower than samples from cruises, such as over the North Atlantic 2 (7.0 and 9.6 µg m −3 , respectively) and Pacific (5.8 and 9.0 µg m −3 , respectively) (Zhang et al., 2010).The Ca 2+ concentration was the highest among all locations and contributes 10 % of major ions (Fig. 4), followed by the Mediterranean Sea, South Atlantic and North Atlantic 2 (Fig. 5).The relatively high Ca 2+ concentration may be because of Asian terrestrial dust transported to Yongxing Island.NSS Ca 2+ accounted for 93 % of total Ca 2+ , ranging from 0.14 to 9.31 µg m −3 with an annual average of 1.99 µg m −3 .Large contributions of NSS Ca 2+ were also found in the Mediterranean Sea, South Atlantic and North Atlantic 2, being at 88.4, 90.3 and 90.0 %, respectively (Zhang et al., 2010).The relatively high NSS-Ca 2+ concentrations in those oceans were potentially from the crust or dust from some deserts (Zhang et al., 2010).Comparing Yongxing Island with all locations, average Mg 2+ concentrations were higher at Yongxing Island than most reported values among all locations (Fig. 5).K + was also the highest among all locations (Fig. 5).were major in the form of secondary inorganic aerosols.They accounted for only 34.0 % of total inorganic ionic concentrations, giving them an intermediate position among all locations (Fig. 5).The average SO 2− 4 concentration at Yongxing was the highest among all locations.As shown in Fig. 4, the mean contribution of SO 2− 4 to major inorganic ionic components was ∼ 25 % at Yongxing.The NSS-SO 2− 4 concentration was 3.66 µg m −3 , with a contribution of 66.1 % to total SO 2− 4 .Similar to SO 2− 4 , the average concentration of NO − 3 in this study was the highest among all locations.It accounted for 9 % of major ions at Yongxing Island.This indicates that a large number of anthropogenic sources affected the concentrations of SO 2− 4 and NO − 3 .It was surprising that NH + 4 had relatively low concentrations over most oceans, except for the South Atlantic and Mediterranean Sea (Fig. 5).The average NH + 4 concentration was 0.07 ± 0.07 µg m −3 in aerosol at Yongxing Island (Fig. 3), representing < 1 % of total major ions (Fig. 4).Further, low concentrations of NH + 4 were also observed in rainwater on the island (Xiao et al., 2016).

Global marine aerosol chemical patterns
Globally, sea salt ions (Na + and Cl − ) were the most important components in marine atmospheric aerosol, with higher concentration of Cl − than Na + , except over the Mediterranean and northern seas (Fig. 5; Zhang et al., 2010;Ebert et al., 2000).In the marine atmosphere, sea salt aerosol (NaCl) can react with sulfuric acid and nitric acid to release HCl, which results in a deficit of Cl − relative to Na + (Zhang et al., 2010).It is also found that a deficit of Cl − in transition season at Yongxing Island (Fig. 4 and Table 2) was most likely because air masses were primarily from remote ocean regions far from the continent (Fig. 1), where wind is weak (Fig. 2).The mole equivalent ratios of Cl − / Na + (neq L −1 ) in aerosols were slightly larger than seawater in annual, cool and warm seasons at Yongxing Island (Table 2).This suggests that Cl − enrichment had an anthropogenic or other nat-  ural origin (Duan et al., 2006;Jung et al., 2012;Xiao et al., 2013).For SO 2− 4 with SS SO 2− 4 and NSS SO 2− 4 , NSS SO 2− 4 was greatly influenced by anthropogenic sources from developed industrial areas, leading to higher concentrations of SO 2− 4 than Na + and Cl − .Examples were Bermuda, Ogasawara and the Arabian Sea (Fig. 5; Moody et al., 2014;Kumar et al., 2008), where NSS SO 2− 4 was the preferred species for acid displacement (Zhang et al., 2010).As another important ion of anthropogenic sources, NO − 3 concentrations were often good relationships with those of NSS SO 2− 4 (Zhang et al., 2010), with relatively high concentrations among major ions (Fig. 5).Relatively high concentrations of SO 2− 4 and NO − 3 were also found over the SCS (Figs. 3 and 4).NH + 4 had the lowest concentrations among the major ions in most marine atmospheric aerosols, suggesting little ammonia transport to the open ocean, such as Yongxing Island.However, there were some exceptions.For example, the South Atlantic and Mediterranean Sea had the highest NH + 4 concentrations among major ions (Fig. 5).Over most seas, the order was Ca 2+ > Mg 2+ > K + .However, we found that Mg 2+ had higher concentrations than Ca 2+ in some remote ocean areas, such as in the Pacific, Atlantic and Southern oceans (Zhang et al., 2010).This indicates that Ca 2+ of crustal origin was difficult to transport to the remote oceans, and Mg 2+ may mainly be from sea salt over the open ocean (Moody et al., 2014).

Seasonal patterns of aerosol chemical species over
SCS and adjacent areas

Seasonal characteristics at Yongxing Island
As illustrated in Figs. 4 and 6, seasonal and monthly TSP concentrations and major inorganic water-soluble ion concentrations had distinct features at Yongxing Island.Generally, concentrations of TSPs and major inorganic ions were higher in the cool season than in the warm season (Fig. 6).Seasonal variations were the same as those in other studies, such as Okinawa, and 18 urban, rural and remote sites in various regions of China (Arakaki et al., 2014;Wang et al., 2006;Xiao and Liu, 2004;Zhang et al., 2012).Average TSP concentrations were 114.7 ± 82.1, 60.4 ± 27.0 and 59.5 ± 25.6 µg m −3 in the cool, warm and transition seasons, respectively, with the highest monthly average in November 2014 and the lowest in April (39.4 µg m −3 ) and September (39.9 µg m −3 ) of the same year (Fig. 6).There were lower concentrations in the warm season than in the cool season because 70 % of rainfall at Yongxing Island happens during the warm season (Fig. 2), being the same as other studies, such as Shanghai and over the China Sea (Wang et al., 2006;Zhao et al., 2015).The positive correlation between TSP concentrations and wind speed (p < 0.01) shown in Table 3 suggests that relatively high speeds can produce many particles from both sea spray and terrigenous matter.We discovered negative correlations between TSP concentrations and temperature (p < 0.01) and relatively humidity (p < 0.01) (Table 2), indicating that warm temperatures and high relatively humidity enhance particle activation and scavenging is happening (Liu et al., 2011).
As shown in Figs. 4 and 6, sea salt ions Na + and Cl − were characterized by a gradual increase from the transition to cool season.Their concentrations (in µg m −3 ) in the cool, warm and transition seasons were 4.91 ± 1.82 and 3.04 ± 1.08, 2.28 ± 1.35 and 9.93 ± 6.78, and 5.25 ± 2.63 and 3.73 ± 3.63, respectively, with corresponding contributions of 52, 57 and 57 % to total major ions in those seasons.The highest Na + and Cl − concentrations in a single sample were found in November, with the lowest concentrations in May and April, respectively.The highest average monthly concentrations were in November.Positive relationships between Na + or Cl − and wind speed in Table 3 (p < 0.01, correlation coefficient R = 0.44 and p < 0.01, R = 0.43, respectively) at Yongxing Island suggest that sea salt concentrations were dependent on wind speed.This is consistent with results at Chichijima Island (Boreddy and Kawamura, 2015).There was a low negative relationship between Na + and rainfall (p < 0.05, R = 0.27) but no relationship between Cl − and rainfall (p > 0.05) in Table 3, suggesting that Na + mainly existed in coarse particles and was readily removed by rainfall.As shown in Table 3, concentrations of Na + and Cl − were also negatively influenced by temperature and relatively humidity.Although Mg 2+ is often treated as crustal-derived ions and elements in continental studies (Zhang et al., 2015), its highest monthly average concentrations were in November at Yongxing Island, the same as Na + and Cl − (Fig. 5).As shown in Fig. 5, Tables 1 and 2, similar trends and strong correlation were observed among Na + , Cl − and Mg 2+ , and the ratios of Mg 2+ to Na + in aerosols were close to that in seawater, suggesting that Mg 2+ may mainly derive from sea salt rather than continental sources.However, there were no relationships between Mg 2+ and wind speed, temperature, relatively humidity or rainfall (Table 2), in contrast to other ions, such as Na + and Cl − .These results reveal that Mg 2+ has different behaviors in the marine atmosphere at Yongxing Island.The different behaviors of Mg 2+ were also found in rainwater at Yongxing Island (Xiao et al., 2016).
As shown in Fig. 6, the highest monthly average concentrations of Ca 2+ were in February.Its monthly trends were different from those of TSPs, Na + and Cl − , and the ratios of Ca 2+ to Na + in aerosols were much higher than those in seawater (Table 2), suggesting that Ca 2+ from terrestrial dust sources may be influenced by different factors.Ca 2+ accounted for 10, 13 and 8 % of total major ions in the cool, transition and warm seasons, respectively (Fig. 4).There was no correlation between Ca 2+ and wind speed, in contrast to TSPs, Na + and Cl − (Table 3).However, there was a negative relationship between Ca 2+ and rainfall (p < 0.05; Table 3).These results suggest that Ca 2+ existed in coarse particles that can be readily removed by rainfall.Thus, a low mass concentration was observed for Ca 2+ in the rainy (warm) season (Fig. 6), with a low percentage being in the warm season in Fig. 4. K + concentrations were 0.42 ± 0.23 µg m −3 in the cool season, 0.22 ± 0.18 µg m −3 in warm season and 0.15 ± 0.07 µg m −3 in the transition season at Yongxing, with the maximum monthly average concentrations in February and the minimum in July (Fig. 6).However, the lowest NSS-K + monthly average concentration was in August.The results suggest that NSS K + is derived from Chinese biomass/biofuel burning in the cool season (Lawrence and Lelieveld, 2010).Streets et al. (2003) computed that China contributes 25 % of total biomass burning in Asia.Sites in Chinese coastal regions had higher K + and NSS-K + concentrations than those at Yongxing Island (Wang et al., 2006) were observed, further indicating that Chinese and other northeastern Asian regions' biomass/biofuel burning have a strong influence on atmospheric composition over the SCS.

Seasonal patterns over SCS and adjacent areas
The spatial variability in seasonal patterns of the major inorganic ionic components at Yongxing Island and adjacent sites of the Acid Deposition Monitoring Network in East Asia (EANET) in 2011 is portrayed in Fig. 7.In general, total major inorganic ionic concentrations tended to be higher in cool seasons and lower in warm seasons to the north of Phnom Penh, including Phnom Penh, Hoa Binh, Hanoi, Hongwen, Hedo, Ogasawara and Yongxing Island, being consistent with previous studies (Boreddy and Kawamura, 2015;Wang et al., 2006;Xiao and Liu, 2004).There was no substantial seasonal variation at other sites of EANET, and there was no strong seasonal variation of rainfall there either.These results suggest that rainfall, wind patterns and anthropogenic activities influence the ionic seasonal variations (Lawrence and Lelieveld, 2010;Wang et al., 2006;Xiao et al., 2013;Xiao and Liu, 2004).Additionally, total major ionic concentrations were higher in the north than in the south, indicating more anthropogenic pollutants in the north, such as SO 2− 4 and NO − 3 (Lawrence and Lelieveld, 2010).As it is well known, the most densely populated regions in the north, in- cluding Hanoi, northeastern China and the Pearl River Delta of China, Korea and Japan release large amounts of pollutants (Lawrence and Lelieveld, 2010), which then transport to the SCS in the cool seasons (Fig. 1).
The total ionic concentrations were higher at the three islands than sites to the south of Phnom Penh.As shown in Fig. 7, relatively high concentrations of Na + and Cl − were found at those islands, suggesting that ions from sea salt had large contributions to total major ions, i.e., 52.8, 62.5 and 55.6 % at Yongxing, Ogasawara and Hedo, respectively.This represents high mass concentrations of sea salt in the marine atmospheric aerosol.The highest concentrations of both Na + and Cl − appeared in November at Yongxing and Hedo islands, which were influenced by a strong northeast monsoon.The highest concentrations of both Na + and Cl − were in September at Ogasawara Island, which were influenced by a strong southeast monsoon from the Pacific.The relationship between Na + , Cl − and wind speed at Yongxing (p < 0.01) is shown in Table 3.Other sites in Fig. 7 were also influenced by wind speed and winds directly from the ocean.However, the highest Na + and Cl − concentrations at some sites did not appear in the same month, e.g., at Hongwen, the highest concentrations of Na + were in April and the highest of Cl − were in January.Excess Cl − in January has been observed by anthropogenic sources in China (Duan et al., 2006).
The highest concentrations of Mg 2+ were in the same months as Na + at most sites, indicating that Mg 2+ may be from sea salt with Na + for almost all stations.The exceptions were at Hoa Binh and Tanah Rata with the maximum Mg 2+ concentrations being in December and July, respectively.This suggests that Mg 2+ originates from the crust rather than oceans (Xiao and Liu, 2004), or from both crust and ocean at these sites.Hoa Binh was influenced by the northeast monsoon, which carries strongly weathering crustal matters from the Chinese Yunnan-Guizhou Plateau karst (Hien et al., 2004;Xiao et al., 2013), and there was a strong relationship between Mg 2+ and Ca 2+ (R = 0.7, p < 0.05).Ca 2+ had its highest concentrations in July at Phnom Penh, Tanah Rata, Petaling Jaya, Serpong and Danum Valley, all of which are located in the south of the SCS.In these regions, relatively little rainfall (rainfall data from EANET) and strong sunlight were observed in that month, leading to strong weathering that generated Ca 2+ .However, the highest Ca 2+ concentrations were found at other sites in the cool season, during which there was much dust from northeastern Asia (Fig. S1; Boreddy and Kawamura, 2015;Y. Liu et al., 2014;Wang et al., 2011).This result is consistent with earlier studies (Boreddy and Kawamura, 2015;Cheng et al., 2000; Y. Liu  , 2014;Zhao et al., 2015).The Ca 2+ data also proved that Asian dust can affect northern SCS, but it is difficult for Asian dust to be transported to southern SCS (Fig. S1).
Figure 8 shows fire spot data from MODIS global fire mapping around the SCS during March 2014 to February 2015.Additionally, smoke surface concentrations every day in that period and region are shown in Fig. S1 in the Supplement.The fire spot and smoke data give information on seasonal variations of biomass burning around the SCS.This activity was strong from January to April in the west of the SCS, including Vietnam, Thailand and Laos, and between July and October in the south of the SCS, including Malaysia and Indonesia (Figs. 8 and S1).These data are consistent with other studies showing substantial monthly CO emissions from biomass burning during February-April and August-October in Southeast Asia, and February-May in southern China and Taiwan (Streets et al., 2003).K + is commonly used as a tracer of biomass and biofuel burning (Deng et al., 2011).As shown in Fig. 7, we found that the maximum K + was in the aforementioned months at most sites, suggesting that Asian biomass burning influenced the SCS region.
Fossil fuel combustion, industrial processes, biofuel burning, and agricultural and waste handling often generate large quantities of SO 2 , NO x and NH 3 in Asia (Lawrence and Lelieveld, 2010;Liu et al., 2013;Xiao et al., 2012aXiao et al., , 2014bXiao et al., , 2015)), although natural emissions of SO 2 and NH 3 in-clude biomass burning, marine and soil biological processes (Streets et al., 2003;Altieri et al., 2014;Boreddy and Kawamura, 2015;Xiao et al., 2012a) and NO x from those processes and lightning (Price et al., 1997;Xiao et al., 2015).In general, the three marine sites (Yongxing, Hedo, and Ogasawara islands) had smaller proportions of SO 2− 4 , NO − 3 and NH + 4 than inland sites, with the three ions accounting for ∼ 35 % at the three marine sites and up to 65 % at the other sites.This indicates that anthropogenic contributions are smaller over remote open oceans than at continental sites.Figure 6 shows that the highest SO 2− 4 , NO − 3 and NH + 4 concentrations were found during the cool season in the north of Phnom Penh, including Phnom Penh, Hoa Binh, Hanoi, Hongwen, Hedo, Ogasawara and Yongxing, being consistent with total inorganic major ions.This indicates that the pollutants from northeastern Asia have a great impact on the northwestern Pacific. Figure S1 confirms these findings that pollutants from nature and anthropogenic activities effect the northwestern Pacific.We also found that most sites in the south of Phnom Penh had maximum SO 2− 4 and NO − 3 concentrations in the same months as the highest K + concentrations occurred, suggesting that biomass and biofuel burning are important sources for SO 2− 4 and NO − 3 in those regions.Lawrence and Lelieveld (2010) found that such burning was important in the emissions of SO 2− 4 and NO − 3 in southern Asia, whereas fossil fuel combustion and industrial processes tended to be dominant in northern Asia (Xiao et al., 2015).However, maximum NH + 4 concentrations at some sites (e.g., Petaling Jaya, Serpong, Danum Valley) were inconsistent with SO 2− 4 and NO − 3 .Moreover, there was no relationship between SO 2− 4 or NO − 3 and NH + 4 at these sites in the south of Phnom Penh, including Phnom Penh and Yongxing Island (both p > 0.05).The results are inconsistent with previous studies (Boreddy and Kawamura, 2015;Hsu et al., 2007;Wang et al., 2006;Xiao et al., 2013;Xiao and Liu, 2004).In the marine atmosphere, SO 2− 4 and NO − 3 are predominant in coarse particles (Boreddy and Kawamura, 2015;Xiao et al., 2015), whereas NH + 4 is often predominant in fine particles and may exist in the form of (NH 4 ) 2 SO 4 in their accumulation mode (Ooki et al., 2007;Ottley and Harrison, 1992).

Source identification, apportionment and region
Based on the PMF 5.0 model, five potential sources of atmospheric chemical components at Yongxing Island were identified: sea salt (two species), crust, secondary inorganic aerosol (SIA) and oceanic emission.Table 4 summarizes source apportionment of the relative contributions of each identified source to major ions.Figures 9 and 10 show the modeled source profiles and the time series of modeled concentrations for each identified main source.
Further, CWTs were plotted for TSPs and major ions (Fig. 11) to explore likely regional sources and transport pathways for the island.Air masses at Yongxing Island had obvious unique and seasonal variations, from northeast of the island in the cool season, southwest in the warm season and southeast in the transition season (Fig. 11).This reveals that aerosol or chemical compositions at the island originated from different regions in different seasons (Figs. 1,8,11 and S1).The air masses with high TSP concentrations were from China coastal regions bordering the Yellow and East China seas and northern South China (Fig. 11).This is consistent with the seasonal variations of TSP concentrations in Fig. 6.The average AOT over the northwestern Pacific (Fig. 1) confirmed this result.A relatively large average AOT was found over the northern SCS and East China Sea in the cool season, and Karimata Strait in the warm season (Fig. 1).But there was a relatively low average AOT over the entire SCS in the transition season (Fig. 1).
The first source, sea salt, generally has strong marine elements, such as Na + (Cl − ) and Mg 2+ , in which Na + exists in super-micron size aerosols, whereas Mg 2+ exists in sub-micron size aerosols (Moody et al., 2014).They contributed 77.4 % (Na + ), 93.9 % (Cl − ) and 70.4 % (Mg 2+ ) from sea salt at Yongxing Island (Table 4).Although the CWT for Mg 2+ was larger in the cool season and lower in the warm season, air masses with relatively high concentrations of Mg 2+ originated offshore of China (Fig. 10).This further indicates that Mg 2+ was mainly from sea salt.Other significant sources were crust, SIA and oceanic emission, with contributions < 10 % for Na + and Cl − , and < 16 % for Mg 2+ .According to data of rainwater at Yongxing Island, a part of Na + and Cl − could be from crust and produced by burning (Xiao et al., 2016).Coal combustion and biomass  burning also produce Na + and Cl − (Liu et al., 2000;Tiwari et al., 2013;Zhang et al., 2015).Zhang et al. (2015) found that coal combustion was the most likely dominant source of Cl − in Beijing.The mole equivalent Cl − / Na + ratios were larger in the cool season than in the transition and warm seasons at Yongxing, indicating that the crust, fossil combustion, and biofuel and biomass burning affected Na + and Cl − concentrations over the northwestern Pacific (Figs. 8 and S1).As shown in Table 3, there were strong relationships between Na + (Cl − ) and SO 2− 4 (Ca 2+ , K + ), further proving that crust, fossil combustion and biomass burning can generate Na + and Cl − .Moreover, NaCl can react with acids such as H 2 SO 4 , HNO 3 and H 2 C 2 O 4 , altering Na + and Cl − concentrations in the marine atmosphere and producing secondary chlorine-containing salt (Boreddy and Kawamura, 2015).Sea salt provided K + , Ca 2+ and SO 2− 4 , constituting 53.2, 33.9 and 24.0 %, respectively.The ratios of K + / Na + , Ca 2+ / Na + and SO 2− 4 / Na + in Table 2 also indicate that part of them are from sea salt.The results are consistent with other studies (Boreddy and Kawamura, 2015).
The second source, crust, has substantial crustal elements Ca 2+ and Mn (Fig. 9), which are tracers of crust (Suzuki and Tsunogai, 1988;Xiao et al., 2013;Xiao and Liu, 2004;Norris et al., 2014).However, Ca 2+ from the crust, only had a contribution of 9.8 % (Table 4).The result indicates that it is difficult for Ca 2+ directly derived from crust to transport to open ocean.But CWTs for Ca 2+ were larger in the cool season and lower in the warm season, indicating that dust from northeastern Asia influenced aerosol chemistry in the remote marine areas (Figs. 1,11 and S1).The result suggests that when it reacts with H 2 SO 4 and HNO 3 to generate secondary inorganic aerosol, e.g., CaSO 4 and Ca(NO 3 ) 2 , it may transport to a longer distance.
The third source is relevant to secondary inorganic aerosol, which are typically characterized by remarkable SO 2− 4 and NO − 3 .They contributed 57.5 and 69.5 % from SIA at Yongxing Island (Table 4) and good relationship between them was observed (Table 2).Fossil fuel (especially coal) combustion releases large amounts of SO 2 and NO x (Xiao et al., 2012b(Xiao et al., , 2014(Xiao et al., , 2015)).Lawrence and Lelieveld (2010) attributed 61 % of total NO x and 77 % of total SO 2 emissions from fossil fuel combustion in southern Asia, and 76 % of total NO x and 75 % of total SO 2 in northern Asia, resulting in the transport of substantial secondary inorganic aerosols containing SO 2− 4 and NO − 3 to the SCS (Figs. 1, 11 and S1).As with TSPs and Ca 2+ , air masses from China, coastal regions had high SO 2− 4 and NO − 3 concentrations in the cool season (Fig. 11), owing to rapid economic development and great coal demand in the country, especially in its coastal re-

Figure 1 .
Figure 1.Distribution of seasonal average aerosol optical thickness (AOT) at 550 nm (T550) over northwestern Pacific in cool, warm and transition seasons during sampling period.Monthly AOT products (from Moderate Resolution Imaging Spectrometer, MODIS) with 4 km resolution were downloaded from GlobColour (http://hermes.acri.fr/).The GlobColour project began in 2005 as an ESA Data User Element project to provide a continuous dataset of merged level 3 ocean color products.

Figure 2 .
Figure 2. 3 h temperature, relative humidity, wind speed and precipitation at Yongxing Island during sampling period (March 2014 to February 2015).

Figure 3 .
Figure 3. Annual average mass concentrations of aerosol chemical species at Yongxing Island.Box boundary indicates 25th and 75th percentile.Lines within the box show the mean.Whiskers above and below the box indicate 90th and 10th percentiles.

Figure 4 .
Figure 4. Mean contributions of each major ionic component to total ionic mass concentration of (a) Yongxing Island (YXI) annual, (b) YXI cool season, (c) YXI warm season and (d) YXI transition season.

Figure 5 .
Figure 5. Comparisons of major ions in aerosol at Yongxing Island with global ocean.Data of Oki, Ogasawara and Hedo are from EANET (www.eanet.asia);data of Rishiri Island are from Okuda et al. (2006); data of Hawaii are from Carrillo et al. (2002); data of Bermuda are from Moody et al. (2014); data of Amsterdam Island are from Claeys et al. (2010); data of the Arabian Sea and Indian Ocean are from Kumar et al. (2008); data of Helgoland are from Ebert et al. (2000); and data of the Mediterranean Sea, North Atlantic 1 and 2, Pacific and South Atlantic are from Zhang et al. (2010).Pentagrams represent sampling sites on islands while others represent cruises.N.A. indicates no data.

Figure 7 .
Figure 7.Comparison of aerosol chemical species between Yongxing Island and around the South China Sea (data from EANET).

Figure 9 .
Figure 9. Profiles of five sources identified from the PMF 5.0 model, including sea salt (two species), crust, secondary inorganic aerosol and oceanic emission.

Figure 10 .
Figure10.Time series contributions from each identified sources, including sea salt (two species), crust, secondary inorganic aerosol, and oceanic emission.

Table 1 .
Annual average, minimum and maximum mass concentrations (µg m −3 ) of TSPs and aerosol chemical species at Yongxing Island.

Table 2 .
Mole equivalent ratios for major ionic species in aerosols at Yongxing Island (annual, cool, transition and warm seasons), together with seawater ratios for comparison.

Table 3 .
Correlation coefficients among major ions in aerosol and meteorological parameters.

Table 4 .
Relative contributions (%) for different major ions from potential five sources of TSPs at Yongxing Island over the year, based on PMF 5.0 model.