Air Quality Measurements from the Southern Particulate Matter Supersite in Taiwan

This study introduces the Southern Particulate Matter Supersite in Taiwan, which began operating on April 1, 2005. The supersite has one core station and three satellite stations for monitoring the properties of particulate matter (PM) and emission sources in southern Taiwan. High-time resolution (1–30 minutes) data for physical and chemical properties of ambient PM are acquired continuously. Measurement data are as follows: (1) PM2.5 (PM with aerodynamic diameters < 2.5 μm) and PM10 (PM with aerodynamic diameters < 10 μm) mass concentrations; (2) PM2.5 compositions of sulfate, nitrate and carbon; (3) particle light scattering and absorption; (4) particle number concentrations in various size fractions between 10 nm and 20 μm; (5) related precursor gases such as NOy, H2O2, and NH3; and, (6) meteorological variables. Most measurements are unique to the study area and can be used to elucidate the causes of PM pollution and evaluate PM exposure and adverse health effects. In addition to describing the sampling location, measurements and data archiving, future challenges for the supersite are discussed as well.

Monitoring Network (TAQMN) since 1993 (Chang and Lee, 2006).The TAQMN consists of more than 70 air stations distributed throughout Taiwan (Taiwan EPA, 2007).Each air station measures hourly concentrations of criteria pollutants (i.e., SO 2 , CO, NO 2 , O 3 , PM 10 ) and meteorological variables (i.e., wind speed, increase may be due to the increase in the number of vehicles, 6-10% per year (Lin et al., 2002).An improved understanding of the causes of excessive air pollution and source-receptor relationships is needed.The need for enhanced temporally, chemically, and size-resolved PM data prompted the establishment of Taiwan Particulate Matter Supersites (Chan, 2000).Currently, Taiwan has two supersites (Fig. 2), the Northern Particulate Matter Supersite (Northern Supersite) and Southern Particulate Matter Supersite (Southern Supersite).The Northern Supersite started monitoring in March, 2002 (Lee, 2002;Lee et al., 2006;Chang et al., 2007), and the Southern Supersite began in April, 2005 (Wu et al., 2005;Lin et al., 2006a).

SOURCES AND METEOROLOGICAL CHARACTERISTICS
The Southern Supersite (Fig. 2) is located in the Kao-Ping air basin in southern Taiwan (Wu et al., 2005;Lin et al., 2006a)

Pollution sources
Mobile and stationary sources are important in the Kao-Ping air basin.
According to the Taiwan Emission Database 6.1 (Taiwan EPA, 2006)

Meteorology
Southern Taiwan is located in a subtropical, coastal area.Fig. 3 shows the monthly variations in temperature, relative humidity and rainfall, based on a 30-year Rainfall is principally concentrated in summer.Temperature and humidity are relatively high in summer.However, variations in temperature (18.9-29.9°C),and humidity (75-82%) are relatively stable due to the coastal environment.
Air pollution levels are low in summer and high in winter (Chen et al., 2004a).
During summer, the synoptic weather is

Monitoring stations
The Southern Supersite has one core and three satellite stations.

Measurements
Both PM 2.5 and PM 10 mass concentrations are measured by a tapered-element oscillating microbalance preceded by a sample equilibrium system (SES-TEOM) (Meyer et al., 2000).The SES-TEOM, which is operated at 30°C to minimize volatilization, has a diffusion drier that    (2005a; 2005b; 2005c) reported that the SES-TEOM and federal reference method (FRM) PM 2.5 are highly correlated (R 2 = 0.89-0.95) and their regression slopes were near unity (0.86-1.07) at the Atlanta, Baltimore and Pittsburgh Supersites.
Notably, conversion efficiencies from calibration solutions are not necessarily the same as those for ambient particles.(μg/m 3 ) using an appropriate conversion factor (16.2 m 2 /g as the default for an aethalometer) (Magee Scientific, USA).
However, the conversion factor depends on absorption wavelength and the physical and chemical properties of particles.In Fresno, the conversion factor is 10 m 2 /g, which was estimated by comparing light absorption with integrated filter analysis of EC (Park et al., 2006a).This comparison can be used to Collocated tests between the TSI scanning mobility particle sizer (SMPS) (TSI, USA) and Grimm SMPS (GRIMM, Germany) at the Fresno Supersite reveal that the two sizers are similar in the 30-50 nm particle range (Chow et al., 2008).The Grimm 1.108 and Grimm SMPS can also report particle mass concentrations derived from the measured size distribution when particle shape and density are used.Therefore, measurement of local particle density is important to obtaining reliable PM concentrations.

Quality assurance and data validation
Monitors are maintained and calibrated by consulting firms according to standard operating procedures established by the Taiwan EPA (Taiwan EPA, 2005).
Independent audits are performed by academic institutes (Wu et al., 2005;Lin et al., 2006a).Measurements are transferred to data servers located at the main office of the Taiwan EPA in Taipei.Data are given validation codes (Table 3) (Wu et al., 2005;Lin et al., 2006a).Daily averages of each measurement are also calculated.The calculated daily data for the most recent year are available on the Taiwan EPA website (http://taqm.epa.gov.tw/emc/default.aspx?m od = PsiAreaHourly).
The PM 10 mass levels before April 2007 at the TZ station were higher than those at the other stations (Fig. 4a), while PM 2.5 mass levels at the TZ station were close to those at the other stations (Fig. 4b).The increased coarse PM concentrations at the TZ station are unclear and need further investigation in the future.
Notably, SO 2 is the primary precursor of sulfate and can be oxidized to form sulfuric acid (H 2 SO 4 ) via gas-and liquid-phase reactions.Once H 2 SO 4 forms, it can react quickly with NH 3 and thereby form nonvolatile ammonium sulfate ((NH 4 ) 2 SO 4 ), which is mostly found in fine PM.The gas-phase oxidation rate is determined by levels of the ambient hydroxyl radical (OH), which is associated with photochemical activity (Seinfeld and Pandis, 1998).
Liquid-phase oxidation is fast and occurs in clouds, rainwater, and within the water fraction of ambient aerosols.The liquid pathway is likely important in southern Taiwan as the monthly average humidity exceeds 75% (Fig. 3).Considerable sulfate differences among stations exist (Fig. 4c), indicating that sulfate is formed locally.
Notably, SO 2 is primarily emitted from coal-fired power plants and large point sources in southern Taiwan.The relatively high levels of sulfate in winter at the northern TO station (Fig. 4c) is likely influenced by the SO 2 emissions originating from the upwind Shingda coal-fired power plant, 14 km from the TO station.However, further investigation is required to validate this hypothesis and evaluate other possible sources.in October (major) and May (minor).The October ozone peak coincided with elevated PM 2.5 mass and sulfate (Fig. 5), implying that the increasing PM 2.5 mass resulted from excess production of sulfate through the gas-phase pathway due to strong photochemical activity.
The primary precursor of nitrate is NO x .
Both mobile and point sources are important NO x sources in southern Taiwan.Similar to the gas-phase pathway for sulfate, NO x can be oxidized by OH and thereby form nitric acid (HNO 3 ) (Seinfeld and Pandis, 1998).
Additionally, NO 2 can be oxidized to nitrous oxide (N 2 O 5 ) by NO 3 radical during the nighttime and, subsequently, HNO 3 can be formed homogeneously or heterogeneously when N 2 O 5 reacts with water (Seinfeld and Pandis, 1998).Notably, HNO 3 can react with NH 3 to form ammonium nitrate (NH 4 NO 3 ); NH 3 can come from livestock and fertilizers.Considerable differences in nitrate concentrations exist among stations (Fig. 4d), indicating that NH 4 NO 3 is produced locally.The relatively high nitrate concentrations in winter at the southern CZ station (Fig. 4d) likely resulted from abundant NH 3 , with sulfate at the lowest levels among the four stations (Fig. 4d).Therefore, small NH 3 was consumed by sulfate, explaining the elevated NH 4 NO 3 concentration observed at the CZ station.
Nitrate concentrations at the urbanized TZ station are much lower than those at the other three stations (Fig. 4d), possibly due to the lack of NH 3 in the city.Nitrate concentrations at the southern CZ and northern TO stations peak in January (Fig. 4d).These peaks are likely due to the high ambient NH 3 concentrations (Fig. 5) and low ambient temperature, which favor particle phase of NH 4 NO 3 .
Notably, EC is a primary pollutant and is generated by incomplete combustion of local sources (Fine et al., 2008), while OC can be from primary and secondary sources.Elevated wintertime PM 2.5 OC and EC concentrations (Figs.4e and 4f) can be attributed to the abundant vehicles in southern Taiwan.High amounts of carbonaceous aerosol at low temperatures suggest low combustion efficiency for internal combustion engines and enhanced condensation of organic vapors.The EC concentrations peak in January (Fig. 4f) can be explained by enhanced emissions and a relatively low mixing depth in cold months.
Notably, the OC levels peak in December, one moth earlier than EC, likely due to the contribution of secondary OC (Fig. 5).The production rate of secondary OC peaks in October and drops off after October, as inferred based on ozone variations (Fig. 5).
Table 4 shows the annual PM 10 mass and PM ).This is similar to levels observed in the eastern U.S. (Fine et al., 2008), but differ from those in the western U.S. where only NO 3 -and OC dominate (Chow et al., 2008).The annual PM 10 averages (Table 4), ranging from 47-57 μg/m 3 , are lower than the Taiwan PM 10 annual NAAQS (65 μg/m 3 ), but the annual PM 2.5 averages (Table 4), ranging from 32-25 μg/m 3 , are much higher than the U.S. PM 2.5 annual NAAQS (15 μg/m 3 ).According to Tsai et al. (2006), when relatively humidity is > 85%, the BAM reading will increase as humidity increases.
These situations frequently occur during nighttime in southern Taiwan due to the nature of high humidity (Fig. 3).The low For PM 10 mass, PM 2.5 mass and PM 2.5 OC and EC concentrations, the major peaks were near morning rush hours, indicating that these are related to vehicle emissions.
The TZ station, located in Kaohsiung City, has higher morning peaks of PM 10 mass, and

Conversion efficiencies for the RP 8400S
and RP 8400N , are an alternative.Good agreement existed between on-site ICs and filter measurements at several U.S. supersites (Hogrefe et al., 2004;Grover et al., 2006) and those in Taiwan (Chang et al., 2006;2007).

Measurements of particle density and water
Particle density facilitates conversion of number distributions to mass distributions and aerodynamic diameters related to Stokes diameter.Bulk particle density can be estimated from the PM chemical composition.Alternatively, effective particle density can be calculated when the following combinations are known: mobility size-aerodynamic size; mobility size-particle mass; or, aerodynamic size-particle mass (Ristimaki et al., 2002;McMurry et al., 2002).Particle water may account for some unresolved PM mass (Chow, 1995;Rees et al., 2004).Rees et al. (2004) reported that water can contribute 8-16% of FRM PM 2.5 mass at the Pittsburgh Supersite.One of the most common methods for measuring particle water is to use hygroscopic tandem differential mobility analyzers (H-TDMAs) (Cocker et al., 2001).Additionally, automatically measuring particle water is possible.One method has been successfully tested at the Pittsburgh Supersite (Khlystov et al., 2005).Further investigation of seasonal variations of particle water is necessary to explain the unresolved PM mass in Taiwan, where numerous studies about PM compositions are based on filter analyses in the laboratory (Lin and Tsai, 2001;Lee and Chang, 2002;Lin, 2002;Lin and Lee 2004;Lin et al., 2005b;2007;Fang et al., 2006;Chen et al., 1999;2001;2003;2004b;Tsai and Cheng, 1999;2004;Tsai and Kuo, 2005;Chio et al., 2004;Chang et al., 2006;Chiang et al., 2005).

Organic speciation
Organic compounds are emitted from combustion processes and natural sources, and formed in atmosphere as secondary organic aerosols (SOAs).Some organic compounds are good source markers, such as levoglucosan for biomass burning, cholesterol for meat cooking, 1,2-benzenedicarboxylic acid for SOA, and iso-nonacosane for cigarette smoke (Schauer and Cass, 2000;Chow et al., 2007a;2007b;Watson et al., 2008).Therefore, organic species analysis can enhance source apportionments and correlate combustion aerosols to adverse health effects (Mauderly and Chow, 2008).Analytical techniques for PM organic species in Taiwan are labor intensive and use water or solvent extraction followed by gas chromatography/mass spectrometry (GC/MS) (Lin and Lee, 2004).
The recent advancement in thermal desorption following GC/MS analysis was examined by Chow et al. (2007b) and can be used as an alternative, cost-effective method.

SUMMARY
The Southern Supersite will increase an

Fig. 1 .
Fig. 1.Annually occurring frequencies of PSI > 100 averaged over the whole of Taiwan and southern Taiwan, respectively, from the Taiwan Air Quality Monitoring Network (TAQMN) air stations during 1994-2006.

Fig. 2 .
Fig. 2. (a) Locations of the Northern and Southern Particulate Matter Supersites in Taiwan, and (b) wind direction, temperature, relative humidity).The daily pollutant standards index (PSI) at each air station is calculated by the Taiwan EPA based on daily concentrations of criteria pollutants at each station.The calculated PSI value is proportional to pollutant concentrations with PSI = 100 equal to Taiwan's NAAQS.Therefore, the percentage of times the PSI > 100 is the frequency that Taiwan's NAAQS are violated.Fig. 1 shows annual percentage PSI > 100 during 1994-2006, averaged for all TAQMN air stations and in southern Taiwan.The annual percentages decreased from 1994-2003 (Fig. 1) as the Taiwan EPA has implemented many control strategies (e.g., air pollution fees, emission permits, and environmental impact assessments) over the last decade.However, these percentages have increased after 2003, suggesting the need for further control measures.This

Fig. 3 .
Fig. 3. Monthly averages of temperature, relative humidity, and rainfall based on a 30-year record for 1971-2000 at the Kaohsiung meteorological station in southern Taiwan.
dominated by typhoons and Pacific anticyclones.The air is typically clean during typhoon periods due to the strong winds and heavy showers that impact Taiwan.Pollution levels are also low during Pacific anticyclones due to clean southwesterly winds and deep daytime mixing depths.During winter, the synoptic weather is dominated by continental anticyclones traveling from China to the East China Sea.Western Taiwan experiences northerly winds due to channeling from the Central Mountain Range.Pollutants emitted in central and northern Taiwan can be transported southward.When the synoptic flow is easterly, western Taiwan is on the lee side of the Central Mountain Range.In this circumstance, upwind air pollution transport and accumulation of local emissions are conducive to pollution episodes in the Kao-Ping air basin.
, and EC measured in Southern Supersite are analyzed to demonstrate the possible benefit of the Supersite.The analysis of other measurements will be

Fig. 4 .
Fig. 4. Seasonal and hourly variations of: (a) PM 10 mass, (b) PM 2.5 mass , (c) PM 2.5 sulfate, (d) PM 2.5 nitrate, (e) PM 2.5 organic carbon (OC), and (f) PM 2.5 elementary carbon (EC) between April 1, 2006, and June 30, 2007, at the Southern Supersite.Station codes are shown in Fig. 2. Notably, OC and EC at the FY station between February 1, 2007 and June 30, 2007 are removed due to failing to pass the data validation.As well, OC and EC at the CZ station between May 5, 2006, and October 23, 2006, are not measured.
work, hourly data coupling with a filter period of T = 90 days with t = 40.25 days (966 hours) and p = 5 are used to produce seasonal tends for each considered measurement.Fig. 4 shows the calculated seasonal trends along with hourly concentrations for PM 10 mass, as well as PM 2OC concentrations have considerably seasonal variation with low values during summer and high values during winter.The strong ventilation and heavy showers via summer typhoons explain the low pollutant levels in summer, while relatively less ventilation and upwind pollution transport increase pollutions in

Fig. 5
Fig. 5 compares the monthly variations of PM 2.5 mass, sulfate, nitrate, ozone and ammonia at the FY core station.The monthly averaged ozone concentration (Fig. 5) shows that photochemical activity peaks

Fig. 5 .
Fig. 5. Comparison of seasonal trends of PM 2.5 mass, PM 2.5 sulfate, PM 2.5 nitrate, ozone and ammonia between April 1, 2006 and June 30, 2007 at the core FY station.These seasonal trends are produced by the low-pass Kolmogorov-Zurbenko filter, as described in the section of seasonal variations in the text.
31, 2007.Generally, the concentrations of each measurement among the different stations are compatible, except for relatively high PM 10 mass and low PM 2.5 NO 3 -concentrations at the TZ station, and the high PM 2.5 SO 4 2-concentrations at the TO station.The major components of PM 2.5 are SO 4 2-(24-29%), OC (21-22%) and NO 3 -(10-16% Figs. 4(a) and 4(b) present comparisons of these levels with seasonal trends of PM 10 and PM 2.5 mass.The annual PM 10 averages in 2006 were 88, 76 and 72 μg/m 3 at the TZ, TO, and CZ air stations (Taiwan EPA, 2007), which are much higher than the corresponding 57, 49 and 47 μg/m 3 measured at the satellite stations.The PM 10 monitors used in the TAQMN are Model 650 PM 10 beta attenuation monitors (BAMs) (Thermo Env.Instruments, USA), while the Southern Supersite uses the SES-TEOM.The high PM 10 concentrations measured by the BAMs are likely due to moisture interference.
Figs. 6 and 7 compare hourly averages of PM mass and major components during the cold (October 2006 to March 2007) and warm (April-September 2006) seasons.The diurnal variations of PM mass and components are apparent during cold season and less important in warm months.The diurnal variations at the core and three satellite stations are similar (Figs. 6 and 7).
understanding of PM characteristics in southern Taiwan.Information provided is related to: (1) PM composition, (2) PM optical properties (e.g., light scattering and absorption), (3) PM number concentrations in various size fractions from 10 nm to 20 μm, and (4) PM precursors gases (NO y , NH 3 and H 2 O 2 ).The daily averages of these measurements after validation are available on the Taiwan EPA website (http://taqm.epa.gov.tw/emc/default.aspx?m od = PsiAreaHourly).Receptor modeling analysis of measured data will provide source apportionment results for the development of PM pollution-control strategies.Additionally, these data sets will also benefit researchers investigating PM exposure and adverse health effects.The Southern Supersite should expand its measurements and related research to: (1) loss of semi-volatile substance from the SES-TEOM, (2) conversion efficiencies for the continuous sulfate (RP 8400S) and nitrate (RP 8400N) monitors, (3) measurement of particle density and water, (4) analysis of PM organic compounds, especially source markers, (5) analysis of size-resolved particle composition, (6) application of new receptor modeling techniques, and (7) measurement of source profiles for Taiwan pollution sources.

Table 1 .
Air quality and meteorological variables continuously monitored since April 1, 2005 at the core and three satellite stations of the Southern Supersite in Taiwan.

Table 1 .
Air quality and meteorological variables continuously monitored since April 1, 2005 at the core and three satellite stations of the Southern Supersite in Taiwan.(Continued) Hourly criteria pollutants (i.e., SO 2 , CO, NO 2 , O 3 and PM 10 ) and meteorological variables (i.e., temperature, relative humidity, rainfall, wind speed and direction) are available at the air stations of TAQMN for the three satellite stations.BTEX stands for benzene, toluene, ethylbenzene and xylene; GC/PID for gas chromatograph/ photo ionization detector.
a FY stands for the Fooyin core station and TO, TZ and CZ for the Chiautou, Chenjen and Choujau satellite stations, respectively.b c Not applicable.d * This variable are continuously measured for the first time in Southern Taiwan.

Table 2 .
Summary of minimal detectable limits, precisions, and calibration methods for the measurements at the Southern Supersite in Taiwan.

Table 2 .
Summary of minimal detectable limits, precisions, and calibration methods for the measurements at the Southern Supersite inTaiwan.(Continued)

Table 2 .
Summary of minimal detectable limits, precisions, and calibration methods for the measurements at the Southern Supersite inTaiwan.(Continued) removes particle-bound water.Lee et al.

Table 3 .
Summary of data validity levels and flags.

Table 4 .
Annual PM 2.5 , PM 10 , NO 3 OC and EC concentrations and their ratios relative to PM 2.5 mass at the Fooyin core station (FY) and the Chiautou (TO), Chenjen (TZ), and Choujau (CZ) satellite stations based on measurements betweenApril 1, 2006 and March 31, 2007.
, a Na a Not applicable.