Effects of Seasonality and Transport Route on Chemical Characteristics of PM 2 . 5 and PM 2 . 5-10 in the East Asian Pacific Rim Region

This study investigated seasonal variation and transport routes of PM2.5 and PM2.5-10 associated metallic elements in the western coastal area of southern Taiwan. Particle sampling was conducted from March 2009 to February 2010. Sixteen metallic elements in PM2.5 and PM2.5-10 samples were determined by ICP-AES and ICP-MS. Multiple approaches, backward trajectory analysis, enrichment factors (EFc), and principle component analysis (PCA), were used to identify the potential sources of the metallic elements. Analysis of the temporal distribution revealed seasonal peaks for most of the trace elements in PM2.5 and PM2.5-10 during winter season and the major elements in PM2.5-10 during the autumn season. The EFc confirmed that the main contributors of Cu, As, Zn, Pb, Cd, and Se were anthropogenic sources. PCA suggested traffic emissions, coal, and heavy oil combustion from both local and neighboring areas, as the major anthropogenic contributors at the sampling site. Backward trajectory analysis, demonstrated different chemical characteristics between the northeast (winter originating in China) and southwest monsoon (summer, from the Southeast Asia). Even in the same season, route-dependent effects of long-range transport in metallic concentrations and total excess cancer risk (ECR) of health-related metals were observed.


INTRODUCTION
Ambient particulate matter (PM) is a complicated mixture of the chemical components directly released from natural or anthropogenic origins.It is emitted as primary or arises as secondary pollutants and is widely represented as suspended particles.One size class, PM 2.5 , is especially harmful to human health in terms of respiratory and cardiopulmonary diseases, morbidity, and mortality (He et al., 2010), and epidemiological studies affirm these associations (Pope et al., 2002;Englert, 2004;Laden et al., 2006;Cao et al., 2012).The particularly sensitive population includes those who have asthma symptoms (Yeatts et al., 2007).In addition, Agency for Toxic Substances and Disease Registry (ATSDR) has considered the metallic elements like As, Cd, Co, Cr, Ni, and Pb present in fine particles as human or animal carcinogens (http://www.atsdr.cdc.gov/toxicprofiles).
Daily deaths due to PM 2.5 were found correlated with specific metallic elements in this PM (Ostro et al., 2008;Valdés et al., 2012).Seasonal effects are also relevant to the epidemiology of PM 2.5 (Becker et al., 2005;Chen et al., 2013), which prompted the current work.The emission of PM is associated with a range of sources, such as coaloil combustion, industrial activities, traffic, soil dust, and sea salt (Almeida et al., 2005;Lin et al., 2008;Mamane et al., 2008;Lee and Hieu, 2011;Lee et al., 2016).
Taiwan, our study location, is located off the southeast coast of China and is often affected by the transboundary movement from the Asian continent during the northeast monsoon in winter season, with a resultant deterioration of the local air quality (Chen et al., 2015;Hsu et al., 2016;Lai et al., 2016).The prevailing winds are the southwest monsoon from the ocean in summer (June-August) and the northeast monsoon from China in the winter (December-February).The winter monsoon brings cold air to Taiwan, but it also carries air pollutants and dust (Hsu et al., 2005;Chi et al., 2014;Chuang et al., 2016;Hsu et al., 2016).
The study site is in Kaohsiung (22°30'-22°45'N, and 120°14'E, 120°23'E), in south-western Taiwan.The municipality has an area of 2,947 km 2 and a population of approximately 2.8 million.It is situated in a unique location for observation of the long-range transport of pollutant outflows from mainland China in the winter northeastern monsoon period and from Southeast Asia in the summer southwestern monsoon period and for assessment of the impact of transboundary transport of atmospheric aerosols on the local air quality (Hsu et al., 2005;Tsai et al., 2012).
Previous studies of the transboundary movement of pollutants have adopted metal ratios (Hsu et al., 2005), lead isotope ratios (Hsu et al., 2006;Lee et al., 2007), and backward trajectory analysis to identify the origin of air masses (Abdalmogith and Harrison, 2005;Lee et al., 2007;Huang et al., 2012;Tan et al., 2014;Lai et al., 2016).Backward trajectory analysis can reveal an association between the pathway of air masses and the atmospheric concentrations of metallic elements.Nevertheless, the characteristics of atmospheric metallic elements with respect to different transport pathways can still be difficult to interpret.Only a limited number of studies used the backward analysis to trace the origin of air masses in Taiwan (Chuang et al., 2016;Hsu et al., 2016).Li et al. (2016) reported the seasonal variation and chemical characteristics in PM 2.5 on both sides of the Taiwan Strait.However, these reports did not discuss the metallic characteristics under the different transport routes.Although several investigations have examined the characteristics and sources of PM 2.5 , PM 2.5-10 , and PM 10 in Kaohsiung over the past decade (Lin, 2002;Chen et al., 2003;Tsai and Chen, 2006;Hsu et al., 2008;Tsai et al., 2010;Tsai et al., 2011).These previous studies were conducted only over short sampling periods (several months); longer campaigns (e.g., one year) to determine the chemical characteristics of metallic elements in PM 2.5 and PM 2.5-10 are still rare.The present study was designed to identify the metallic element character, sources, and effects of long range PM transport during the summer and winter monsoons.
Sixteen metallic elements in the PM 2.5 and PM 2.5-10 fractions were examined in the urban coastal area of southern Taiwan.The particulate mass, size fraction ratio (PM 2.5 and PM 2.5-10 in PM 10 ), elemental concentrations, enrichment factors (EF c ), and principal component analysis (PCA) were used to elucidate the seasonal differences in chemical characteristics, to identify possible sources of metallic elements in PM, and to differentiate the local pollution emissions from the longrange transport (monsoon-driven).The different chemical characteristics between northeast monsoon (winter) and southeast monsoon (summer) was also investigated.In addition, the seasonal variations in the potential cancer risk of the PM-bound carcinogenic metals in PM 2.5 and PM 2.5-10 delivered by the different routes was assessed.

Sampling Site
The sampling site (22°37'N, 120°18'E) is placed on the campus of National Sun Yat-sen University, which is adjacent to the coast in Kaohsiung, located at the southwestern part of Taiwan (Fig. 1).Weather condition belongs to the subtropical zone.Rainy season runs from May to September while from October to April for dry season.Kaohsiung is an industrial city with a population of 2.8 million.There are four processing export zones.Steel, metal processing and petrochemical production are the major industries.National Sun Yat-sen University is surrounded by Shoushan Mountain from the north to east and by Taiwan Strait in west.Ambient particles were sampled on the rooftop of Marine Environment and Engineering Building, approximately 25 m above the ground level and 100 m away from the shoreline of the Taiwan Strait.Kaohsiung harbor (the biggest harbor in Taiwan) is located on south of the site (1.2 km distant).Power plants are located to the north (28 km distant) and south (13 km distant) of the sampling site; a steel and petrochemical industrial park is located to the southeast (13 km distant).

Sampling and Analytical Procedures
The previous report showed contamination may derive from sampling, storage, and the analysis processes (Windom et al., 1991).Therefore, EPA method 1669 (USEPA, 1996) was adopted for sampling and analysis to reduce the contaminants in this study.All labware was soaked in prepared 10% HNO 3 solution for 24 hours, rinsed with deionized water (18MΩ), and dried in a clean bench and then stored in a clean area prior to use.
A Universal Air Sampler (Model 310, MSP Corporation, USA) was adopted to obtain PM 2.5 (fine) and PM 2.5-10 (coarse) samples from the study area.The flow rate was set at 300 L min -1 .Particle sampling campaign conducted from March 2009 to February 2010, with a sampling frequency of about three times per month on non-raining days.A problem with the sampling equipment resulted in collection of only six samples in the spring.The northeast monsoon often occurs in winter and leads to a deterioration in local air quality, so sampling was conducted more frequently during this season.A total of 88 twenty-four hour samples (PM 2.5 and PM 2.5-10 ) were collected on quartz fiber filters (Pall, 2500 QAT-UP) for each sampling period.After sampling, the filters were placed on pre-cleaned plastic petri dishes, sealed in plastic bags, and put into a clean plastic box.The filters were subsequently conditioned for 48 h in an electric desiccator at relative humidity 45 ± 5% and temperature 25 ± 3°C prior to weighing.
Filter samples were cut with ceramic scissors and extracted with a mixture of 8 mL nitric acid (Merck, Tracepure 69%) and 2 mL hydrofluoric acid (Merck, Proanalysis 40%) (Yeh et al., 2015) in a Teflon screw-cap vial (Savillex Corp., UK).Samples digested for 3-4 h to evaporate the extracted solution to near dryness on a hotblock (SC 154, Environmental Express Corp., SC) at 160°C.The digested solution was diluted to 30 mL with 1% nitric acid in a vessel for analysis.Metallic standard solutions were all ordered from the Merck Company (Germany).Major elements (Al, Ca, Fe, Mg and Na) were determined by inductively coupled plasma atomic emission spectrometry (ICP-AES, Thermo IRIS Intrepid II XSP) and trace elements (V, Cr, Mn, Co, Ni, Cu, Zn, As, Se, Cd, and Pb) by inductively coupled plasma mass spectrometry (ICP-MS, HP4500).

Quality Assurance and Quality Control
To enhance the reliability of data quality, method blanks, quality control, duplicate analysis, and matrix spiked samples were conducted during laboratory analysis.Standard reference material (SRM 1648a) from the National Institute of Standards and Technology was used for performance evaluation.All the control samples were treated in the same way as the actual samples.In addition, method detection limits were determined as the concentration equivalent of three times the standard deviation of seven replicate measurements of the analyte in reagent water, and then converted to an atmospheric concentration in ng m -3 .In this study, all data reported were corrected using the method blank.The percent recoveries of quality control, matrix spike, and standard reference material (SRM 1648a) were from 97.2% (Pb) to 101.9% (Mg), 80.6% (Cu) to 110.0% (Na), and 87.7% (Cd) to 109.9% (Zn) [except for Cr, which was 33.7%], respectively.The low recovery of Cr might due to its presence in silicate matrices that are difficult to digest (Paode et al., 1999;Yatkin and Bayram, 2008).The relative percent differences (RPD) for all duplicate analysis were from 5.84% (Mn) to 20.1% (As).The method detection limits were between 0.0028 ng m -3 (As) and 1.9 ng m -3 (Ca).

Meteorological Conditions
The PM concentration and its chemical composition can be influenced by meteorological conditions (Chen and Kwok, 2001;Tao et al., 2012).The seasonal distributions of wind roses are shown in Fig. S1.During the sampling periods, the prevailing winds in the spring, summer, autumn, and winter were northwest, south, north, and north, respectively.The mean wind speeds were 1.6, 1.2, 1.4, and 1.7 m s -1 , respectively.The mean rainfall was 147, 1447, 192, and 13 mm, respectively.A much higher rainfall occurred in summer than in other seasons.The mean relative humidity was 70, 77, 76, and 70 %, respectively.

PM Concentrations and Size Fraction Ratios
Table 1 shows the atmospheric PM concentrations from the samples collected on non-raining days over the four seasons, as well as the PM 2.5 /PM 10 ratios.The annual mean concentrations of PM 2.5 , PM 2.5-10 , and PM 10 were 34.9 ± 20.6, 20.8 ± 11.3, and 55.8 ± 28.5 µg m -3 , respectively.The annual mean concentration of measured PM 2.5 (based on 88 twenty-four hour samples) compared favorably with the value of 44.2 ± 21.2 µg m -3 obtained from a nearby Taiwan EPA monitoring station (based on hourly data) during the sampling campaign.(http://erdb.epa.gov.tw/DataRepository/EnvMonitor/AirQualityMonitorHourData.aspx) and 43 µg m -3 at Tainan, Taiwan (Lu et al., 2016).The PM 2.5 data from our study and from the air quality monitoring station both exceeded the Taiwan EPA regulatory annual limit (15 µg m -3 ) (http://air.epa.gov.tw/Public/suspended_particles.aspx#t4).The mean PM concentrations for the four seasons for PM 2.5 , PM 2.5-10 , and PM 10 at the sampling site ranged from 8.2 ± 4.9 (summer) to 47.2 ± 16.6 µg m -3 (winter), 11.1 ± 8.3 (summer) to 28.5 ± 16.0 µg m -3 (autumn), and 19.3 ± 10.2 (summer) to 69.5 ± 21.8 µg m -3 (winter), respectively.The lowest PM concentrations for all three fractions occurred in summer; while the PM 2.5-10 , PM 2.5 , and PM 10 fractions showed the highest concentrations in autumn, winter, and winter, respectively.Note that 50, 0, 67, and 75% of the samples exceeded the regulatory daily limit for PM 2.5 concentration (35 µg m -3 ) in spring, summer, autumn, and winter, respectively.The low PM concentration in summertime may be due to the wet season, a higher mixing layer, and clean marine air masses driven by the southwest monsoon (section 3.1).By contrast, the high PM concentration in winter time could be due to the dry season, a low mixing layer, and long-range transport via the northeast monsoon (Ho et al., 2006;Hao et al., 2007).
Table 1 shows that the highest mean daily PM 2.5 /PM 10 mass ratio occurred in winter (0.69), whereas the lowest value was observed in summer (0.46).This indicates that PM 2.5 represents the largest fraction of PM 10 , except in summer.The high PM 2.5 /PM 10 ratio in winter indicates increased anthropogenic sources for fine particles, which might be caused by long-range transport carried by the northeast monsoon.In addition, note that PM 10 concentrations are significantly correlated with PM 2.5 , with a correlation coefficient (R 2 ) of 0.89 (Fig. S2).This value is comparable to the value of 0.84 reported for the southeastern United States (Parkhurst et al., 1999) and 0.91 for Guangzhou, China (Wang et al., 2006a).This hints at similar sources of both the PM 10 and PM 2.5 that influence the sampling site.
The annual mean of the PM 2.5 /PM 10 ratio (0.60) in this study was comparable with the value of 0.61 to 0.67 reported for central and southern Taiwan, 0.54 to 0.59 for northern Taiwan (Chen et al., 1999), and 0.58 and 0.60 for urban coastal sites (Almeida et al., 2005;Rodrı'guez et al., 2008).Yang et al. (2017) also showed that PM 2.5 /PM 10 ratios were mostly higher than 0.5 in southern Taiwan.By contrast, the PM 2.5 /PM 10 ratio was 0.26 for SdeBoker, an arid area, and it ranged from 0.6 to 0.80 for European and Western Mediterranean sites (Andreae et al., 2002).This may be attributed to large changes in natural and anthropogenic influences at these various sites.

Concentration of Metallic Elements
Table 2 lists the observed seasonal concentrations of sixteen metallic elements in PM 2.5 and PM 2.5-10 .The annual mean of metallic elements ranged from 0.27 ± 0.14 ng m -3 (Co) to 390 ± 231 ng m -3 (Na) for PM 2.5 and 0.10 ± 0.08 ng m -3 (Cd) to 1648 ± 1019 ng m -3 (Na) for PM 2.5-10 .Higher concentrations were observed in autumn and winter for most of the major elements (Al, Ca, Fe, Mg, and Na).The highest concentrations in PM 2.5-10 were observed in winter for trace elements and in autumn for major elements.Most of the trace and major elements in PM 2.5 and PM 2.5-10 were at their lowest concentrations in the summer.Again, as previously mentioned, the low metallic concentrations in summertime may be due to the wet season, a higher mixing layer, and clean marine air masses in the study area.The higher metallic concentrations in winter time may be due to the dry season, a low mixing layer, and long-range transport via the northeast monsoon (Hao et al., 2007).
The PM 2.5 showed the highest concentrations of trace elements in winter, except for V.The annual mean mass concentrations of V, Cr, Mn, Ni, Cu, Zn, As, Se, Cd, and Pb were higher in the PM 2.5 than in the PM 2.5-10 fraction (Table 2).This indicates an enrichment of these elements in the PM 2.5 fraction (Marcazzan et al., 2001;Almeida et al., 2005;Kong et al., 2010;Tan et al., 2014).Some of the trace elements (Cu, Zn, As, Se, and Cd) showed seasonal variations in their ratios (ratio of maximum and minimum seasonal average concentration) of 7.0-to 13.3-fold in PM 2.5 and 6.7-to 9.4-fold in PM 2.5-10 , while major elements (Al, Ca, Fe, Mg, and Na) exhibited much lower seasonal variation ratios, from 2.0-to 2.9-fold in PM 2.5 and 1.9-to 3.1-fold in PM 2.5-10 .Major elements were mostly affected by crustal and marine sources and exhibited less seasonal variation than was observed for the trace elements (V, Cr, Mn, Ni, Cu, Zn, As, Se, Cd, and Pb); these were mostly likely a contribution from anthropogenic sources (e.g., coal combustion, industrial sources, and traffic emissions).
Seasonal PM variations can be especially intensified due to coal combustion in cold weather for heating (Zheng et al., 2005).However, Taiwan has no such heating activity; therefore, we believe the substantial increase in seasonal variation ratios was caused by long-range transport (Lai et al., 2016).Interestingly, vanadium showed the lowest seasonal variation ratios (1.3 in PM 2.5 and 1.2 in PM 2.5-10 ) among all elements investigated.This revealed that V was not significantly affected by season, with the inference that local emissions, rather than long-range transport, are the main contributors of V. Yu et al. (2013) reported the annual mean concentration Ni/V ratio was 1.84 in Beijing (China).However, this value was only 0.95 in the present study.
The seasonal variation ratio (max/min) of Ni was 3.5 and 3.1 in PM 2.5 and PM 2.5-10 , respectively (Table 2).Higher seasonal variation was therefore noted for Ni than for V in the present study area.In addition, V showed the highest concentrations in the summer, when the prevailing wind was southerly in the study area.Kaohsiung Harbor is located on the south side, 1.2 km away from the study area.Harbor activities involving marine diesel fuel could be sources of V (Mueller et al., 2011;Zhao et al., 2013;Yeh et al., 2015).
The most abundant elements were Al, Ca, Fe, Mg, and Na, accounting for 78.5% (winter) to 92.0% (summer) in PM 2.5 and 96.6% (winter) to 98.9% (autumn) in PM 2.5-10 , respectively.This result aligns with crustal and marine aerosols as major elemental sources for the PM 2.5-10 fraction, which would result in a quite constant percentage.Natural elements such as Al, Ca, Fe, Mg, and Na predominate in the PM 2.5-10 fraction, while the trace elements Se, As, V, Cd, Pb, Zn, Cu, and Cr are mainly associated with PM 2.5 .This finding is consistent with previous studies (Marcazzan et al., 2001;Ny and Lee, 2011).The seasonal distribution of metallic elements of PM 2.5 and PM 2.5-10 in PM 10 is illustrated in Fig. 2. A similar distribution pattern was observed, except in summer.This result may be related to the distinct sources and meteorological conditions, such as wind rose shown in Fig. S1, if the local emission remains unchanged seasonally, as observed for summer.

Enrichment Factors (EF c )
Enrichment factors (EF c ) are used to differentiate elements in aerosols originating from natural (crustal and marine) or anthropogenic sources.In this study, elemental EF c values above 10 typically consider as an anthropogenic source (Chester et al., 2000;Lin et al., 2016).The EF c value of element X is obtained from the following equation: where (C X /C Al ) a is the concentration ratio of element X and Al in ambient particle and (C x /C Al ) c is the concentration  ratio of analyzed element in the crust (Wedepohl, 1995).
The EF c values of the fifteen elements in PM 2.5 and PM 2.5-10 during the sampling periods are shown in Fig. 3.The EF c values of Cu, As, Zn, Pb, Cd, and Se were from 11 (Cu) to 7429 (Se) in PM 2.5 and from 11 (As) to 387 (Se) in PM 2.5-10 , which suggested that these elements would more likely originate from anthropogenic sources across the four seasons in the study period.The metallic elements Cu, As, Zn, Pb, and Se with EF c greater than 100 were also found highly enriched in PM 2.5 in the Western Taiwan Strait region, China (Xu et al., 2013).Therefore possibility of cross Strait transport of contaminants exists.Moreover, the EF c of these elements were generally higher in PM 2.5 samples were generally higher than in PM 2.5-10 samples (Almeida et al., 2005), indicating that those elements from anthropogenic sources, such as traffic, coal combustion and heavy oil combustion, predominated in the PM 2.5 fraction (Espinosa et al., 2001;Xu et al., 2012).In addition, the EF c values of V, Cr, and Mn in PM 2.5-10 were all below 10, while they were above 10 in PM 2.5 , suggesting that V, Cr, and Mn arose from mixed origin sources.The EF c values of Ca, Fe, Mg, and Co were below 10 and were relatively constant, which suggested these elements likely originated from crustal sources (Gao et al., 2014).The EF c values of Na were below 10 in PM 2.5 but above 10 in PM 2.5-10 .The high EF c values of Na in PM 2.5-10 could reflect an effect of ocean seawater (Kubilay et al., 1997)  when compared with the finer PM 2.5 fraction.Therefore, the local pollution sources, like harbors and coal power plants, along the transport pathway of the southwest monsoon (summer) could be one of the reasons for the high EF c values observed in summer.
The comparable EF c patterns (Fig. 3) suggested similar sources for anthropogenic elements in PM 2.5 and PM 2.5-10 .However, natural elements demonstrated distinct patterns that revealed different sources for natural elements in PM 2.5 and PM 2.5-10 .For example, Na and Mg exhibited higher EF c values in PM 2.5-10 than in PM 2.5 , indicating a greater contribution of marine aerosols to PM 2.5-10 than to PM 2.5 in this study area, as expected (Chow et al., 1994).

Principle Component Analysis (PCA)
Various types of receptor models are widely used in source apportionment studies.However, principal component analysis (Almeida et al., 2005;Lee and Hieu, 2011) is a useful tool that can help identify potential source categories and has been frequently used (Viana et al., 2008).In the present study, PCA was applied to the PM 2.5 fraction and identified four main factors, accounting for 77.1% of the overall variance (Table 3).In principle component 1 (PC1), elements with high scores, including Mn, Co, Ni, Cu, Zn, As, Se, Cd, Pb, and Fe, indicated that these elements were mainly associated with traffic emission and coal combustion (Fung and Wong, 1995;Lin et al., 2005;Wu et al., 2007;Lee and Hieu, 2011;Xu et al., 2013;Hsu et al., 2016).The high scores for V, Cr, and Ni in principle component 2 (PC2) analysis were attributed to heavy oil combustion (Fung and Wong, 1995;Viana et al., 2008).The high scores for Al, Mg, and Na in principle component 3 (PC3) analysis could be considered as contributions related to soil and sea salt (Almeida et al., 2005;Viana et al., 2008).Calcium, with a high score, was dominant in principle component 4 (PC4) analysis, which could possibly be related to cement plants or the construction industry (Fung and Wong, 1995;Yang et al., 2015;Lin et al., 2016).
Three components were extracted from the PM 2.5-10 fraction.These sources accounted for 79.7% of the total variance (Table 3).PC1 included Mn, Co, As, Al, Ca, Fe, Mg, and Na, with high loadings.This component was associated with soil and sea salt (Almeida et al., 2005;Viana et al., 2008).PC2 had high loadings for V, Cr, Ni, Cu, and Zn.This factor is likely to represent heavy oil combustion and traffic emissions (Wu et al., 2007;Viana et al., 2008).PC3 had high scores for As, Se, and Cd, which may be derived from coal combustion (Fung and Wong, 1995;Lin et al., 2005;Ranville et al., 2010;Xu et al., 2013).Fig. 4 presents the score plots of the PCA.The grouping resulted in three groups with distinct seasonal variations in particulate metallic composition.For example, the coal combustion signal in the winter group might represent, in part, material carried by long-range transport from Mainland China (Lai et al., 2016).

Backward Trajectory Analysis
Taiwan is often affected by southwest monsoons in summer and northeast monsoons in winter.Therefore, the pollutants (e.g., Pb, Cd, Zn, and PCDD/F) could be delivered by the monsoons (via long-range transport) and affect the local air quality in Taiwan.Although some previous work Winter Spring/Autumn Summer Fig. 4. Principle component score plot of PM 2.5 size for metallic elements in the sampling site.
has investigated the impact of the northeast monsoon on the air quality in Taiwan (Hsu et al., 2005;Hsu et al., 2006;Chi et al., 2014;Lai et al., 2016), no studies have reported the effect of different transport routes on chemical characteristics of particulate material during the southwest and northeast monsoon periods.The present study implemented backward trajectory analysis developed by NOAA.Detailed information regarding the trajectory model and these data sets can be found on the NOAA Web site (http://www.arl.noaa.gov/ready/hysplit4.html).This type of analysis has been widely applied in the environmental sciences; for instances, to investigate the long-range transport of pollutants and to establish the source-receptor relationships of air pollutants (Stohl, 1998;Lai et al., 2016;Xin et al., 2016;Ommi et al., 2017;Yang et al., 2017).Different air masses travel through different regions and therefore bring different chemical constituents with them, leading to different air quality characteristics in downwind areas (Abdalmogith and Harrison, 2005;Wang et al., 2006b;Kong et al., 2010).Fig. 5 shows the five-day backward trajectories at our sampling site during summer and winter.In general, nine samples were classified into two categories of air masses in summer.The SR1 (n = 3) air masses originate from the ocean and pass over the Southeast Asia (e.g., Philippines; Vietnam), while the SR2 (n = 6) air masses are derived from the ocean.Nineteen of the 20 samples grouped into three categories of air masses in the winter: air masses that came from north China and passed over the coastal cities to reach the sampling site (WR1, n = 11), air masses originating from northeast China that passed over the ocean to arrive in Taiwan (WR2, n = 6), and air masses that derived from the ocean and passed over the Philippines to reach Taiwan (WR3, n = 2).
Table 4 shows the mean concentrations of PM 2.5 and PM 2.5-10 at the sampling site derived from different air masses.In summer, the mean concentrations of Zn, Cd, and Pb of PM 2.5 in SR1 were 37.5 ± 7.6, 0.18 ± 0.03, and 32.4 ± 11.2 ng m -3 , respectively.These values were 2.5, 1.7, and 3.2 times higher than in SR2.The mean concentrations of Cu, As, and Pb for PM 2.5-10 in SR1 were 2.64 ± 2.17, 0.13 ± 0.13, and 10.7 ± 7.2 ng m -3 , respectively.These values were 1.9, 2.5, and 1.6 times higher than in SR2.This result indicated that Cu, Zn, As, Cd, and Pb could accompany the air mass (SR1) transported from some of the Southeast Asia to Taiwan.A sharp economic growth of these countries is expected in the near future, so maybe due to weaker pollution control in vehicle and industrial emission compared with developed countries.It will cause the substantial pollutant emissions to the environment and outflow to receptor areas affecting the human health there.Table 5 summarises the comparison of selected PM 2.5 -bound metallic elements in the Southeast Asia.These studies showed high As, Pb, and Zn concentrations in most Southeast Asia.The sources of As, Pb, and Zn may originate from coal-fired power, lead related industries and vehicle (Santoso et al., 2011;Wimolwattanapun et al., 2011;Khan et al., 2016).
In winter, WR1 exhibited the highest concentrations of Zn, As, Se, Cd, and Pb in PM 2.5 and PM 2.5-10 .The concentrations of these metallic elements in PM 2.5 were 1.8 to 2.3 times higher than in WR3.Among the three air masses, WR1 showed the highest levels of metal pollution.This could reflect the fact that WR1 passed over the coastal cities (urban) and the industrial areas in China and had only a short residence time over the ocean, thereby causing higher levels of metallic elements at our sampling site.The metallic pollutants (Zn, As, Se, Cd, and Pb) from China may originate from coal combustion for power generation, house heating in winter (Huang et al., 2010;Yang et al., 2015), and industrial sources (Hao et al., 2007).WR2 originated from northeast China and had longer residence times over the ocean, resulting in lower concentrations of pollutants than in WR1.However, the highest level of V (37.3 ng m -3 in PM 2.5 and 5.76 ng m -3 in PM 2.5-10 ) and Ni (21.6 ng m -3 in PM 2.5 and 8.32 ng m -3 in PM 2.5-10 ) was observed in WR3.These concentrations were 5.8 and 4.6 times higher for V and 1.5 and 1.7 times higher for Ni than in WR1 and WR2, respectively.These results suggested that V and Ni were mainly derived from local emissions, such as the use of heavy oils in shipping in the harbor (Mueller et al., 2011;Yeh et al., 2015).Kaohsiung Harbor is located on the south side of the sampling site and is only 1.2 km away.Accordingly, these results illustrated the impacts of different pathways in the same season on particulate metal composition in the downwind area.In this study, the effect of transport routes on the EF c of PM 2.5 was also investigated.Table 6 summaries the EF c of PM 2.5 under the different air masses.In summer, all the metallic elements in SR1 showed higher EF c values than in SR2, reflecting that SR1 was a more polluted transport route than SR2.The SR1/SR2 ratio of Zn (4.0) and Pb (4.1) was the two highest metallic elements in the determined elements.This result showed Zn and Pb could originate from Pb-related industries and vehicle emissions in Southeast Asia (Santoso et al., 2011;Khan et al., 2016).In winter, among the WR1, WR2, WR3 routes, Cu, Zn, As, Se, Cd, and Pb showed higher EF c values in WR1 and WR2 than .These results revealed that the transport routes of air masses originating from China carried more anthropogenic pollutants to the sampling site.In contrast, V showed the highest EF c value in WR3 route.This result revealed that V may largely originate from local emission.Among the determined metallic elements, Cr, Mn, Zn, As, Se, Cd, Pb, Ca, and Fe demonstrated higher ratio values in WR1/WR3 and WR2/WR3.The results showed these metallic elements may accompany the air masses transported to the sampling site.Chromium, Mn, and Fe could originate from steel production (Mazzei et al., 2006).Arsenic and Se, and Cd could derive from coal combustion (Fung and Wong, 1995;Hsu et al., 2016).Zinc and Pb could originate from vehicle emissions (Santoso et al., 2011).Calcium was enriched in PM 2.5 fraction during WR1 and WR2 routes, with a higher WR1/WR3 (4.4) and WR2/WR3 (3.1) ratio.Calcium enriched aerosol could be attributed to sources other than natural dust, such as construction and cement plants (Lin et al., 2016).Table 6 also shows the elevated elements (Zn, As, Se, Cd, and Pb) in northeast monsoon routes (WR1 and WR2) revealed obvious higher EF c values than in the southwest monsoon (SR1).The above results (higher EF c and the different elemental sources) showed the different chemical characteristics in ambient aerosol between southwest monsoon (summer) and northeast monsoon (winter) periods.

Excess Cancer Risk Assessment
Among the metallic elements studied, the elements such as As, Cd, Cr, Ni, Pb, and Co are known potential carcinogenic substances, although they also exhibit other toxic effects.These metal elements may be introduced into the human body by exposure through inhalation pathways.The chromium concentration determined by ICP-MS was total chromium in this study and the concentration ratio of the carcinogenic species Cr (VI) in total chromium is about 1/7 (Park et al., 2008;Hsu et al., 2016).This ratio is applied in assessing its corresponding carcinogenic effect.Excess cancer risk (ECR) is used to assess the potential carcinogenic effects that are characterized by estimating the probability of cancer occurrence in a population of individuals from exposures to specific elements (Hsu et al., 2016).The inhalation unit risks of carcinogenic metallic elements (As, Cd, Cr (VI), Ni, Pb, and Co) are provided by the U.S. EPA (Integrated Risk Information System) (https://cfpub.epa.gov/ncea/iris2/atoz.cfm),California Environmental Protection Agency (http://oehha.ca.gov/tcdb/index.asp),and Provisional Peer Reviewed Toxicity Values for Cobalt (https://hhpprtv.ornl.gov/issue_papers/Cobalt.pdf) and were given in Table 7.The calculation of ECR is as follows: where X i is the concentration (µg m -3 ) of metallic element i in PM 2.5 and PM 2.5-10 .Unit risk (UR j ) listed in Table 7.
Table 7 shows the estimated ECR of carcinogenic metals bound in PM 2.5 and PM 2.5-10 .Hexavalent chromium exhibited the highest ECR among the determined metallic elements in PM 2.5 and PM 2.5-10 .Among the four seasons, the highest  total ECR was found in winter for PM 2.5 (36.6 × 10 -6 ) and PM 2.5-10 (18.1 × 10 -6 ).All the total ECRs were higher in PM 2.5 than in PM 2.5-10 in all seasons.These results revealed that potential carcinogenic metals bound to particulate matter were particularly prevalent in the winter season and in the PM 2.5 fraction in our study area.Compared with other cities in wintertime, our total ECR value (36.6 × 10 -6 ) in PM 2.5 was lower than the 62.7 × 10 -6 reported for Beijing, China (Lin et al., 2016).The carcinogenic risks via inhalation exposure for elements (As, Cd, Cr (VI), Ni, Pb, and Co) in PM 2.5 samples were below the acceptable level (1 × 10 -4 ) established by USEPA's risk management (https://www3.epa.gov/airtoxics/cancer_guidelines_final_3-25-05.pdf).Fig. 6 presents the ECR of carcinogenic metals in PM 2.5 and PM 2.5-10 from the different transport routes.Both particle fractions showed the highest total ECR value in the WR1 route, at 41.3 × 10 -6 in PM 2.5 and 19.5 × 10 -6 in PM 2.5-10 .WR2 and WR3 displayed the same level in PM 2.5 and PM 2.5-10 .The lowest ECR value was observed for the SR2 route, at 17.2 × 10 -6 in PM 2.5 and 4.1 × 10 -6 in PM 2.5-10 .WR1 route showed the highest metallic concentrations (Table 4).Therefore, it revealed the highest ECR value.In summer, the wet season and clean marine air masses from ocean, causing the lower metallic concentrations in PM 2.5 and PM 2.5-10 , such that a lower ECR value observed in SR1 and SR2 route.Our results revealed that the different transport routes resulted in a variation of ECR.

SUMMARY AND CONCLUSION
This study presented backward trajectory analysis, EF c , PCA, and ECR to examine the effects of seasonality and transport routes on the chemical characteristics of PM 2.5 and PM 2.5-10 .Distinct seasonal variations were noted for the particulate metallic elements, with higher concentrations of metals during the winter season and lower concentrations during summer time.The EF c analysis indicated that Cu, As, Zn, Cd, and Se originated largely from anthropogenic sources in PM 2.5 and PM 2.5-10 during all seasons.PCA results suggested that traffic emission, coal combustion, heavy oil, soil, and sea salt were major sources of PM 2.5 and PM 2.5-10 .Backward trajectory analysis in PM 2.5 and PM 2.5-10 suggested that the observed elevations in Cu, Zn, As, Cd, and Pb concentrations might occur in air masses passing over Southeast Asia following the southwest monsoon flow to the study site in summer; while Cu, Zn, As, Se, Cd, and Pb might be carried from China, accompanying the northeast monsoon airflow to the sampling site during the winter period.The highest metallic carcinogenic risk of PM 2.5 and PM 2.5-10 was observed in the winter season.In addition, the different chemical characteristics (EF c value, elemental sources, and ECR) in ambient aerosol was also observed between southwest monsoon (summer) and northeast monsoon (winter) periods.nowledge the NOAA Air Resources Laboratory (ARL) for providing the HYSPLIT transport and dispersion model (http://ready.arl.noaa.gov)used in this publication.

Fig. 1 .
Fig. 1.Map of the sampling site, located on the campus of National Sun Yat-sen University, Kaohsiung.

Fig. 5 .
Fig. 5. Backward trajectories of air masses during summer (SR1 and SR2) and winter (WR1, WR2 and WR3) (a) originated from the sea and passed through the Philippines, (b) originated from the sea, (c) north and northeast regions through the inland areas and coastal China, (d) northeast regions through the Shandong Peninsula and outflowed to the sea, and (e) through the Philippines and marine regions.

Fig. 6 .
Fig.6.Excess cancer risk of carcinogenic metals in PM 2.5 and PM 2.5-10 arising from different transport routes.

Table 1 .
Mass concentrations (µg m -3 ) of atmospheric particulate matters in this study.

Table 2 .
Seasonal concentrations of PM . Seasonally, in PM 2.5 , anthropogenic elements (Cu, As, Zn, Pb, Cd, and Se) exhibited the highest EF c values in winter and the lowest in summer.In PM 2.5-10 , Cu and Zn exhibited their highest EF c values in winter.Interestingly, Na, V, Ni, Pb, Cd, and Se showed their highest EF c values in summer.In seasons other than summer, the coarser fraction, PM 2.5-10 , is not as easily carried by long-range transport,

Table 3 .
PCA factor loadings for metallic elements in PM 2.5 and PM 2.5-10 during the sampling period.

Table 4 .
The mean concentrations (ng m -3 ) of PM 2.5 and PM 2.5-10 in the sampling site under different air masses.

Table 5 .
Comparison of selected PM 2.5 -bounded metallic elements average concentrations in the Southeast Asian countries (unit: ng m -3 NA: Non-available.

Table 6 .
The mean (SD) EF c of PM 2.5 in the sampling site under different air masses.

Table 7 .
Seasonal excess cancer risk of carcinogenic metallic elements in PM 2.5 and PM 2.5-10 .Values taken from IRIS (Integrated Risk Information System).b Values taken from Cal EPA (California Environmental Protection Agency).c Values taken from Provisional Peer Reviewed Toxicity Values for Cobalt (USEPA).
a d Inhalation unit risk.