Field Measurement of Total Gaseous Mercury and Its Correlation with Meteorological Parameters and Criteria Air Pollutants at a Coastal Site of the Penghu Islands

This study investigated the seasonal and daily variations of the total gaseous mercury (TGM) concentration in the ambient atmosphere, the correlation of TGM concentration with meteorological parameters (e.g., temperature, humidity, and wind speed) and criteria air pollutant concentrations (e.g., SO2, NOx, CO, O3, PM10, and PM2.5), as well as the transportation routes, at the Penghu Islands. The field measurement results showed that the average TGM concentration during the monitoring period was 3.17 ± 1.17 ng/m, within the range of 1.17–8.63 ng/m, with the highest concentration being observed in spring, while the TGM concentration typically increased in the morning, reached its peak concentration, and then started to decrease at nightfall. Moreover, the lowest average TGM concentration of 1.81 ± 0.15 ng/m was observed in summer, and this figure is close to the background TGM concentration of the Northern Hemisphere (1.6–1.8 ng/m). The correlation analysis indicated that TGM concentration correlated positively with SO2, NOx, CO, O3, PM10, PM2.5 and negatively with ambient temperature, relative humidity, and wind speed. In addition, the transportation route analysis showed that elevated TGM concentrations could be transported from either North China, East China, or South China to the Penghu Islands, while those originating from the South China Sea had the lowest contribution to the TGM levels at the Penghu Islands. Therefore, local sources and open burning might be mainly influenced by the long-range transportation of air masses, as the prevailing wind direction and air mass transportation routes potentially play critical roles in the variation of TGM concentration at the Penghu Islands.


INTRODUCTION
Mercury (Hg), one of the hazardous air pollutants (HAPs), is a persistent, toxic, and bio-accumulative heavy metal, and is currently regulated by the Environmental Protection Agency of USA (USEPA) and the United Nations Environment Programme (UNEP) (Nater and Grigal, 1992;Mason and Sheu, 1994;Schroeder and Munthe, 1998;Lin and Pehkonen, 1999;Boening, 2000;Mason and Sheu, 2002;Lin et al., 2005;Clarkson and Magos, 2006;Fu et al., 2010).Mercury and its derivatives are emitted from various natural and anthropogenic sources, as the latter being typically considered as the major sources of atmospheric mercury.Mercury emission from nature sources is arising from the changes in the crust, such as volcanic eruptions, rock weathering, forest fires, lakes, mercury mining, and evaporation from oceans (Kim and Fitzgerald, 1986;Schroeder and Munthe, 1998;Lin and Pehkonen, 1999;Fu et al., 2009), while mercury emission from anthropogenic sources is arising from high temperature combustion, including point sources (e.g., municipal and industrial waste incinerators, fossil-fuel combustors, iron/steel smelting plants, petroleum refineries, cement plants) and area sources (e.g., industrialized region, mercury-contaminated site) (Lin et al., 2008;Jen et al., 2010;Pirrone et al., 2010).UNEP (2003) reported that the emission of mercury from natural and anthropogenic sources are estimated to be about 2,000 and 4,000 tons/yr, respectively.Particularly, coal-fired power plants and waste incinerators accounted for approximately 70% of total anthropogenic mercury sources globally (Gratz et al., 2013), and China plays an important role in global anthropogenic mercury emission (Li et al., 2013).
Atmospheric mercury exists primarily in three forms based on its chemical and physical properties, including gaseous elemental mercury (GEM), reactive gaseous mercury (RGM), and particulate mercury (Hg p ) (Schroeder and Munthe, 1998;Lin and Pehkonen, 1999;Poissant et al., 2005;Feng and Qiu, 2008;Jen et al., 2012;Huang et al., 2013).Collectively, sum of GEM and RGM are referred as total gaseous mercury (TGM), while sum of GEM, RGM, and Hg p are referred as total atmospheric mercury (TAM).Among them, the TGM could be transformed back and forth between elemental and oxidation states, ceaselessly (Sommar et al., 2001;Pal and Ariya, 2004).Even though, GEM is the major species in the ambient atmosphere, which accounts for 95-99% of TAM with a residence time of 1.5-2.0 years in the atmosphere and can transport over great distances across continents (Schroeder and Munthe, 1998;Lin and Pehkonen, 1999;Fu et al., 2010;Jen et al., 2012).As such, long-range transportation and marine dissipation of mercury in the atmosphere has been identified as the predominant source at the background and the coastal regions (Ebinghaus et al., 1999;Blanchard et al., 2002;Sheu et al., 2010).
The levels of TGM concentration are different in ambient air, range of 1-3 ng/m 3 , 3-5 ng/m 3 , and > 5 ng/m 3 in the rural, urban, and industrial regions, respectively, while its background concentrations are typically ranged from 1.6 to 1.8 ng/m 3 in the Northern Hemisphere and from 1.1 to 1.4 ng/m 3 in the Southern Hemisphere (Keeler et al., 1995;Baker et al., 2002;Lamborg et al., 2002;Jen et al., 2010;Sheu et al., 2010;Jen et al., 2012).The variations of meteorological parameters and criteria air pollutant concentrations could increase or decrease the TGM concentration and its chemical composition (Feng et al., 2002;Sheu et al., 2002).Such as an environment of high humidity and high wind speed could promote the TGM concentration in ambient air to remove, as O 3 concentration become higher, TGM concentration tended to decrease (Fu et al., 2009(Fu et al., , 2010)).Besides, the regional sources and its characteristics might also influence the temporal variation and spatial distribution of atmospheric mercury speciation, concentration, and deposition flux (Mason andSheu, 1994, 2002;Poissant and Hoenninger, 2004;Poissant et al., 2005).
To date, not many field studies have been conducted for atmospheric mercury measurements in Taiwan (Kuo et al., 2006;Sheu et al., 2010;Fang et al., 2010;Jen et al., 2010;Fang et al., 2012, Jen et al., 2012), yet none have been done for measuring atmospheric mercury speciation and concentration at Taiwan's offshore islands.In this study, continuous field monitoring of TGM was conducted to investigate its seasonal and daily variations at the Penghu Islands, as the correlations of TGM concentrations with the meteorological parameters and several criteria air pollutants being further examined and discussed.More importantly, a backward trajectory simulation model was applied to further investigate the transportation of TGM to the Penghu Islands with respect to the transportation routes for those observed at 500 meter above the sea level during the monitoring seasons.While the TGM has been a continuing issue of great concern worldwide, the results of this study would help develop more effective management strategies to control the adverse influences associated with the presence of TGM on the environment and public health in the areas of the Penghu Islands and those possibly affected by the longrange transportation of TGM.

Selection and Description of Sampling Sites
Located at the middle of Taiwan Strait between eastern China and Taiwan Island, the Penghu Islands have the area of 127.97 km 2 and the population of about 90,000.The islands are in the subtropical weather mainly influenced by East Asian monsoons.According to the meteorological data obtained from 1985 to 2011, the dry season with the rainfall of about 800 mm started from April to September.Table 1 summarizes the meteorological data measured at the Penghu Islands during the TGM monitoring seasons, indicating that the prevailing winds were blown from ENE, NW, and S, with the ambient temperatures of 13.4-31.2°C,the relative humidity of 53.2-91.3%, and the wind speeds of 0.5-9.7 m/s.The TGM sampling site was located on the roof of a four-floor building, which was approximately 12 m above the ground and 500 m and 50 m far from the coastline and the major roads, respectively.
Overall, the TGM monitoring site was located at the marine boundary layer (MBL), reducing the possible interferences resulted by other factors.The atmospheric TGM was monitored at the Hsiaomen site (23°38′471′′North latitude, 119°30′316′′ East longitude) located at the north-western coastline of the Penghu Islands (see Fig. 1) for 15 continuous days in each season from March of 2011 to January of 2012 at the Penghu Islands.In this study, four seasons are defined as March to May (spring), June to August (summer), September to November (fall), and December to February (winter), respectively.

Monitoring of Atmospheric TGM Concentration and Correlation Analysis
In this study, the concentration of atmospheric TGM was continuously measured by a TGM monitoring system (Tekran, Model 2537B) in a principle of cold vapor atomic fluorescence spectrometry (CVAFS) (Fig. 2).Previous studies have presented the regarding principle, set-up, and operation of the Tekran TGM monitoring system (Landis et al., 2002;Steffen et al., 2002;Landis et al., 2004;Fu et al., 2010;Jen et al., 2010;Sheu et al., 2010;Ci et al., 2011).The TGM monitoring system, with a sampling flow rate of 1.5 L/min, continuously measured atmospheric TGM every 5 min interval by using pre-concentration on alternating gold traps, thermal desorption, and quantification by the cold vapour atomic fluorescence spectrometry (CVAFS).Its technique is based on the adsorption of TGM on gold traps, followed by thermal desorption at 600°C, and detected the wavelength (λ = 253.7 nm) of Hg 0 by CVAFS.The system provides continuous analysis of TGM at sub-ng/m 3 levels (equivalent to parts per trillion (ppt) or parts per quadrillion (ppq)).Moreover, the instrument provides a mechanism whereby residual mercury is desorbed from the gold cartridge before it is used for analysis.This process is automatically performed on each cartridge before a calibration or a run begins.Regular cleaning, together with short cycle times  and small cartridge loadings, virtually eliminated the memory effect previously associated with pure gold cartridges.In this study, the instrument was calibrated daily using an internal mercury source verified quarterly by manual or auto injection.Briefly speaking, the instrument collected air samplings and trapped mercury vapor into a cartridge containing an ultra-pure gold adsorbent.The amalgamated mercury was thermally desorbed and detected by using CVAFS.A dual cartridge design allowed alternate adsorption and desorption, resulting in continuous measurement of Hg 0 in the air.
In order to understand the correlation of atmospheric mercury with meteorological parameters and criteria air pollutants, this study collected the meteorological data and criteria air pollutant concentrations from the Makung Air Quality Station during the same monitoring periods, and used a linear regression model derived from TGM concentrations with meteorological parameters and criteria air pollutants.Correlation coefficient (R) is used to describe the relationship between TGM concentration, meteorological data, and criteria air pollutant concentrations.When the R value is positive, it indicates a positive correlation; when the R value is negative, it indicates a negative correlation.In this study, the meteorological parameters collected included the ambient temperature (Temp.),relative humidity (RH), wind speed (WS), and wind direction (WD), while the criteria air pollutants of interest included SO 2 , NO x , CO, O 3 , PM 10 , and PM 2.5 .

Backward Trajectory Simulation and Fire Map
In order to identify the long-range transportation of TGM at the monitoring site, 72-hour backward trajectories were simulated by using a NOAA HYSPLIT model with the National Centers for Environmental Prediction's Global Data Assimilation System (NCEP-GDAS) meteorological dataset used as the model input in this study.All the backward trajectories started with an arrival height of 500 m above the sea level.By using the NOAA-HYSPLIT model, the dates with the highest TGM concentration at the Penghu Islands in different seasons were determined and the transportation routes of air masses toward the Penghu Islands during the monitoring periods were then simulated.This information was further applied to examine the possible long-range transportation routes conveying TGM from several long-range emission sources to the Penghu Islands.
Previous studies showed that during the spring, the biomass burning frequently occurs in Southeast Asia and further causes to increase the levels of atmospheric mercury concentration in Taiwan, resulting from long-range transportation (Sheu et al., 2010).This study used the fire map to identify the possible sources of TGM drawn by the FIRMS web (http://firms.modaps.eosdis.nasa.gov/firemap/)with the moderate resolution imaging spectroradiometer (MODIS) data from National Aeronautics and Space Administration (NASA).Each fire site on the map represents an area of a circle with a 1 km diameter containing one or more actively burning fires within the area.

Seasonal and Daily Variation of TGM and Criteria Air Pollutants
Table 2 summarizes the average and standard deviation of TGM and criteria air pollutant concentrations measured at the Penghu Islands.The highest concentrations of PM 10 and PM 2.5 observed in spring were 54.60 ± 15.56 and 31.07 ± 10.59 µg/m 3 , respectively, while the lowest concentrations of PM 10 and PM 2.5 occurred in summer were 36.63 ± 8.69 and 17.70 ± 6.48 µg/m 3 , respectively.Spring and winter are two major seasons frequently blowing Asia dusts from northern China to the Penghu Islands (Yuan et al., 2004;Tsai et al., 2012), which could significantly increase the concentrations of PM 10 .Overall speaking, the concentrations of SO 2 and NO x measured at the Penghu Islands were much lower than most cities in Taiwan, indicating there are no significant highly polluted industries in the Penghu Islands.Moreover, O 3 concentration was relatively higher compared to other criteria air pollutants measured at the Penghu Islands.
Field monitoring of TGM at the Penghu Islands showed that the average concentration of TGM was 3.17 ± 1.06 ng/m 3 within the range of 1.17-8.63ng/m 3 .The concentration of TGM in four season monitoring were ordered as spring (4.34 ± 1.87 ng/m 3 ) > winter (3.51 ± 0.67 ng/m 3 ) > fall (3.03 ± 0.40 ng/m 3 ) > summer (1.81 ± 0.15 ng/m 3 ), and the concentration observed amongst the seasons were significantly different by the analysis of ANOVA (Analysis of Variance) at the confidence level of 95%.Moreover, the average TGM concentration (3.17 ng/m 3 ) at the Penghu Islands was approximately two times of the background TGM concentration of Northern Hemisphere (1.6-1.8 ng/m 3 ).The Penghu Islands is likely to be influenced by long-range transportation of TGM in spring and winter, causing higher TGM concentrations in spring and winter than those in summer and fall at the Penghu Islands.The lowest average TGM concentration of 1.81 ± 0.15 ng/m 3 was observed in summer, which was slightly close to the background TGM concentration of Northern Hemisphere, suggesting that the air masses blown from the South China Sea were relatively clean during this season.
Fig. 2 illustrates the variation of TGM concentrations during four seasons at the Penghu Islands.The daily concentration of TGM in spring increased significantly from April 2 nd to the peak TGM concentration (8.63 ng/m 3 ) observed on April 10 th .The lowest TGM concentration ranging from 1.50 to 2.70 ng/m 3 was consistently observed in summer.Moreover, the concentrations of criteria air pollutants were also the lowest in summer compared to other seasons (Table 2), suggesting that the ambient air quality in summer was relatively better than other seasons at the Penghu Islands, and the TGM concentrations was 1.7-2.8times lower than other seasons.
Unlike other seasons, the TGM concentration fluctuated in fall and winter.In winter, it increased rapidly from 3.74 to 5.08 ng/m 3 on December 28 th , and then decreased to 3.38 ng/m 3 on December 30 th .The highest peak concentration (4.35 ng/m 3 ) occurred on January 2 nd and stabilized after January 3 rd .Long-term TGM monitoring at Mt. Lulin background air quality monitoring station showed that Taiwan was highly influenced by atmospheric mercury and gaseous pollutants from China and Southeast Asia in spring and winter (Sheu et al., 2010).Backward trajectory simulation indicated that atmospheric mercury detected in Taiwan was significantly increased due to the effects of biomass burning from Southeastern Asia and industrial emission from North China in spring and winter.It suggested that the TGM concentrations at the Penghu Islands might be also influenced by the mercury emitted and transported from these regions.While, the air blown from the Pacific Ocean had relatively low TGM concentration.

Hourly Variation of TGM Concentration during Four Seasons
Fig. 3 illustrates the hourly variation of TGM concentrations during four seasons at the Penghu Islands.It showed that the variation of hourly TGM concentrations were relatively stable in summer.The TGM concentration increased from AM 6:00, gradually reached to its concentration peak at AM 11:00, and then decreased after noontime.The increase of TGM concentration resulted from the following two processes: (a) UV radiation could temporally promote Hg + , Hg 2+ and Hg p to volatile Hg 0 , and subsequently emitted to the atmosphere (Schroeder and Munthe, 1998); (b) The downward mixing enhanced TGM aloft may increase the levels of TGM concentration as the destroying of nocturnal inversion layer (Stamenkovic et al., 2007).Except for the morning time, the TGM concentration was less variable through the whole day.The average TGM concentrations in the daytime (AM 8:00-PM 12:00) were typically higher than that at nighttime (AM 12:00-AM 2:00).These findings might be attributed to the effects by the height variation of atmospheric boundary layer.Additionally, it might be also influenced from several local anthropogenic activities.
The magnitude of hourly variation (difference between the maximum and minimum TGM concentrations) was lower in summer and winter with the relative percent differences (RPD) of 61.9 and 88.1%, respectively, and higher in spring and fall with the RPD of 130.7 and 107.8%, respectively.As unexpected, different findings reported from previous studies showed the monitored TGM concentration variations at low altitudes in the typical rural areas (Mao et al., 2008;Nguyen et al., 2010;Sheu et al., 2010).This study revealed that the concentration of TGM monitored at the Penghu Islands was mainly influenced by local wildland open burning and those from long-range transportation.
Fig. 4 illustrates the frequency distribution of TGM detected during the four monitoring seasons.It showed that TGM concentrations followed a lognormal distribution pattern in the range of 2.0-4.5 and 6.0-7.5 ng/m 3 , accounting for approximately 80.00% of total frequency in spring.The values ranging from 1.5 to 2.5 ng/m 3 accounted for approximately 89.44% of total frequency in summer.Similarly, the values in the range of 2.0-4.5 and 3.5-5.5 ng/m 3 accounted for 93.33% and 61.67% of total frequency in fall and winter, respectively.However, the episodes with extremely high TGM concentrations (> 9.0 ng/m 3 ) were frequently observed in spring, fall, and winter.It is worth noting that the frequency distribution of TGM appeared to follow two different trends, as shown in Fig. 4. The distributions of TGM levels in summer and fall were unimodal, while those in spring and winter followed a multi-modal distribution.The concentrations of TGM in summer and fall were relatively stable.
The burning of coal and oil could not only increase the concentrations of SO 2 and NO x , but also the TGM concentration in the air.Another explanation might be mobile sources (e.g., diesel trucks and boats) which could also emit SO 2, NO x , and Hg.Previous studies reported that the GEM concentration tends to decrease, while RGM or Hg p concentration is likely to increase, at elevated O 3 concentration (Selin et al., 2007;Sommar et al., 2010)  was attributed to the fact that Hg 0 can be oxidized to Hg + and Hg 2+ by O 3 , OH•, and reactive halogens (e.g., Cl, Br, ClO, BrO) in the atmosphere, and further adsorbed onto the surface of atmospheric aerosols (Mason and Sheu, 1994;Schroeder and Munthe, 1998;Lin and Pehkonen, 1999;Lu et al., 2001;Mason and Sheu, 2002;Poissant et al., 2005;Feng and Qiu, 2008).However, if the aerosol particles contained high content of sea salts from oceanic spray, O 3 could be adsorbed onto the surface of these aerosol particles, and thus reducing the O 3 concentration, and more importantly, further oxidizing gaseous elemental mercury (GEM) to its oxidized forms (Hg + and Hg 2+ ) by the photochemical oxidation reactions with OH• and reactive halogens (Berg et al., 2001;Feng,et al., 2002;Couillard et al., 2008).
Moreover, carbon monoxide (CO) could reduce Hg + and Hg 2+ to Hg 0 , causing higher TGM concentration as the CO concentration increased (Feng et al., 2002;Fu et al., 2010;Sommar et al., 2010).It suggested that TGM was highly positively correlated with both PM 10 and PM 2.5 .When the concentrations of PM 10 and PM 2.5 were increased, the surficial contained divalent mercury (Hg + and Hg 2+ ) of aerosol particles could be restituted to elemental mercury (Hg 0 ) being released to the atmosphere by photo-reduction.Moreover, the TGM could be transported by Asian dusts with high PM 10 and PM 2.5 levels through long-range transportation in the specific seasons, which thus increased the concentrations of PM 10 , PM 2.5 , and TGM consistently.
Unlike criteria air pollutants, TGM concentration had either moderate or strong negative correlations with meteorological parameters including ambient temperature, RH, and wind speed.The prevailing winds of spring and winter were dominated by northeastern monsoons when the long-range transportation of Asian dusts occurred.The higher the wind speed was, the better the atmospheric dispersion could be, potentially reducing the atmospheric mercury concentration.Moreover, the oxidized form of mercury on the surface of the atmospheric aerosol particles would be more stable especially at high O 3 concentration environment.If RH became higher, it would cause a large amount of RGM and Hg p to be scavenged by wet deposition, and thus reduced the concentration of TGM in the ambient air.

Transportation Routes of TGM
Fig. 5 illustrates the pollution rose of TGM for four monitoring seasons at the Penghu Islands.In this study, high TGM concentrations were observed mainly in the wind directions of 0-90 degrees and followed by 270-360 degrees in spring, while in the wind direction of 60-120 degrees in fall and winter.There were no significant mercury emission sources at the northwestern or southwestern part of the Penghu Islands, suggesting that the high concentrations of TGM might be transported remotely from China or southeastern Asia in spring.In summer, the major prevailing winds were blown from the southwest with low wind speeds, thus the TGM observed at the Penghu Islands was likely to be transported from the South China Sea.Additionally, the high concentrations of TGM were blown from the northeast in fall and winter, possibly attributed to local sources from Taiwan and/or longrange transportation from China.However, the comparisons of the wind direction at the near-surface with the TGM concentration detected in the atmosphere might be insufficient to predict the potential trajectories of the atmospheric TGM at the Penghu Islands.
An increasing number of studies have shown that biomass burning, industrial combustion, and ocean evaporation are three major emission sources of TGM in the regional scale (Blanchard et al., 2002;Fu et al., 2010;Ci et al., 2011).Fig. 6 illustrates the backward trajectories and TGM concentration percentage of the air masses arriving the Penghu Islands during four seasons.During the monitoring period in spring, the TGM concentrations for the air masses conveying from routes (1) and (2) ranged from 3.55 to 7.12 ng/m 3 , accounting for approximately 89% of TGM, which was possibly dominated by those transported from local stationary sources and open burning with the air masses toward the Penghu Islands.Thus, South China were another possible sources making contributions to the TGM levels at the Penghu Islands.
In summer, the TGM concentrations for the air masses conveying from routes (4) and (5) ranged from 1.48 to 2.39 ng/m 3 , accounting for approximately 97% on TGM, which were transported from the South and the East China Sea to the Penghu Islands, dramatically increasing the TGM concentrations at the Penghu Islands during the summer monitoring period.Consequently, the TGM concentrations in summer were relatively lower than other seasons, and the average concentration of TGM was close to the background TGM concentration of Northern Hemisphere (approximately 1.6-1.8ng/m 3 ).In fall, the TGM concentrations for the air masses conveying from routes ( 7) and (8) ranged from 3.02 to 3.73 ng/m 3 , accounting for approximately 96% on TGM, which seemed to be mainly transported from northern China, Korea, and Japan with the air masses toward the Penghu Islands, resulting in approximately 1.67 times higher TGM concentration in fall than those in summer.In winter, the TGM concentrations for the air masses conveying from routes (11) and ( 12) ranged from 3.67 to 3.84 ng/m 3 , accounting for approximately 85% on TGM.The possible sources conveying the TGM toward the Penghu Islands in winter were from southeast Asia, Northern China, and Mongolia, increasing the TGM concentrations up to 1.94 times of those in summer.High humidity and rainfall frequency at the Penghu Islands in winter might also explain relatively lower TGM concentrations measured during this monitoring period.Previous study reported that the occurrence of biomass burning such as forest fires from February to April in the Southeast Asia and the Indochina Peninsula emitting high concentrations of mercury-containing pollutants to the atmosphere (Sigler et al. 2003).Fig. 7 illustrates the fire maps obtained from the FIRMS web fire maps and air mass transportation routes in four seasons.These fire maps might be explained why the high TGM concentration occurred in spring at the Penghu Islands.During the monitoring periods, the fire hot spot occurred densely in spring than other seasons, thus it might emit a large amount of TGM to the atmosphere, and further transported from the long-distance sources to the Penghu Islands.
As illustrated in Fig. 8, the TGM concentration in the ambient air at several coastal areas and islands in the East Asia were compared to ascertain the extent of the TGM pollution at the Penghu Islands.The TGM concentrations were decreased in the order of An-Myun > Jeju > Penghu > Okinawa Islands.It is worth noting that the seasonal effect on the ambient mercury concentration at the coastal areas and islands in the East Asia appeared to be similar, as higher concentrations being frequently observed in spring.Previous studies have reported various types and extents of mercury pollution in the atmosphere from different regions in the world.The mercury levels in winter were generally higher than those in summer, suggesting that the height of mixing layer at these areas was lower in winter due to a relatively more stable atmosphere, which plays an important role in elevating the atmospheric mercury concentrations based on mass balance (Lindqvist et al., 1991;Burke et al., 1995;Kock et al., 2005).However, different findings observed at the Penghu Islands and selected locations in the East Asia were possibly attributed to the influences of biomass burning and mercury transport from adjacent areas in certain seasons (Nguyen et al., 2007;Nguyen et al., 2010;Chand et al., 2008;Fu et al., 2010;Tsai et al., 2012).In summary, in addition to the local sources and open burning, the concentration of TGM at the Penghu Islands was mainly influenced by the long-range transportation of air masses, as the prevailing wind direction and air mass transportation routes potentially playing the critical roles on the variation of TGM concentration in the atmosphere.

CONCLUSIONS
The hourly, daily, and seasonal variations of the TGM concentration at the Penghu Islands were investigated.The average TGM concentration at the Penghu Islands during the monitoring seasons was 3.17 ± 1.06 ng/m 3 within the   (Nguyen et al., 2007); (b) Yellow Sea (Ci et al., 2011); (c) Jeju Island (Nguyen et al., 2010); (d) Okinawa Island (Chand et al., 2008); (e) Penghu Islands (This study); (f) South China Sea (Fu et al., 2010).range of 1.17-8.63ng/m 3 , approximately twice higher than the background TGM concentration (approximately 1.6-1.8ng/m 3 ) in the Northern Hemisphere.While the hourly variation, TGM concentration typically increased in the morning (AM 8:00-13:00), reached its peak concentration, and followed by a reduction beginning in the late afternoon (after PM 14:00).For the average TGM concentrations during four monitoring seasons were ordered as spring > winter > fall > summer.Summer is the only season close to the background TGM concentration of Northern Hemisphere at the Penghu Islands.
The results of correlation analysis showed that the TGM concentration was correlated positively with SO 2 , NO x , CO, O 3 , PM 10 , PM 2.5 and negatively with ambient temperature, relative humidity, and wind speed.The air masses transported from southern and northern China, southern Asia, Korea, Japan, and Mongolia might affect the Hg levels at the Penghu Islands during the monitoring seasons.At the Penghu Islands, the concentrations of TGM might be influenced by the mercury-polluted air masses to be transported remotely from areas or local stationary combustion and mobile sources.While the air mass toward the Penghu Islands was dominated by that transported from the South China Sea in summer, the TGM concentration levels at the Penghu Islands appeared to be lower during the monitoring period than other seasons, possibly attributed to the low Hg levels in the air masses.In addition, the high TGM concentration observed at the Penghu Islands during the spring monitoring period might be resulted by the following three reasons: (a) wildland burning, local stationary combustion, and mobile sources emission; (b) long-range transportation from biomass burning in the Southeast Asia or neighboring coastal cities, and (c) long-range transportation through Asia dusts from the North China.

Fig. 2 .
Fig. 2. Variation of TGM concentration during four monitoring seasons at the Penghu Islands.

Fig. 3 .
Fig. 3. Hourly variation of TGM concentrations during four monitoring seasons at the Penghu Islands.

Fig. 4 .
Fig. 4. Frequency distribution of TGM concentration covering the four monitoring seasons.

Fig. 5 .
Fig. 5. Pollution rose of TGM during four seasons at the Penghu Islands.

Fig. 6 .
Fig. 6.Backward trajectories of air masses and TGM concentration percentage for different routes during the monitoring seasons at the Penghu Islands.

Fig. 7 .
Fig. 7. Fire maps of air mass during four monitoring seasons in the East Asia in (a) spring, (b) summer, (c) fall, and (d) winter.

Table 1 .
Meteorological data monitored at the Penghu Islands during the TGM sampling periods.
. It

Table 3 .
Correlation of TGM concentrations with meteorological data and the concentrations of criteria air pollutants.