Dynamic variations of ultrafine, fine and coarse particles at the Lu-Lin background site in East Asia

(1)

Dynamic variations of ultra

ﬁne, ﬁne and coarse particles at the

Lu-Lin background site in East Asia

Sheng-Chieh Chen

a,f

, Shih-Chieh Hsu

b

, Chuen-Jinn Tsai

a,*

_{, Charles C.-K. Chou}

b

_,

Neng-Huei Lin

c

, Chung-Te Lee

d

, Gwo-Dong Roam

e

, David Y.H. Pui

f

a_{Institute of Environmental Engineering, Nation Chiaotung University, Hsinchu 300, Taiwan} b_{Research Center for Environmental Changes, Academia Sinica, Taipei 115, Taiwan} c_{Department of Atmospheric Sciences, National Central University, Jhongli 320, Taiwan} d_{Institute of Environmental Engineering, National Central University, Jhongli 320, Taiwan}

e_{National Institute of Environmental Analysis, Environmental Protection Administration, Jhongli 320, Taiwan} f_{Particle Technology Laboratory, Mechanical Engineering, University of Minnesota, Minneapolis, MN 55455, USA}

h i g h l i g h t s

< Atmospheric PM0.1, PM2.5and PM10were studied at a background site.

< The concentration of biomass burning (BB) makers, K and Mn, elevated signiﬁcantly. < Free troposphere around Taiwan is impacted by BB plumes from Southeast Asia.

a r t i c l e i n f o

Article history: Received 3 January 2012 Received in revised form 8 May 2012 Accepted 12 May 2012 Keywords: Atmospheric aerosol Ultraﬁne particle Biomass burning Biogenic aerosol Chemical mass closure Eastward transport

a b s t r a c t

The characteristics of atmospheric ultrafine particles (i.e. <100 nm, nanoparticles or PM0.1), PM2.5and PM10were studied at the Lulin Atmospheric Background Station (LABS, 2862 m a.s.l., Taiwan) as part of the 7SEAS/Dongsha campaign. Sampling was conducted in July and August of 2009 and September to November of 2010, during which two 96-h and four 72-h PM samples were taken. Real-time particle size distributions were measured continuously from July to August of 2009 and July to November of 2010. PM0.1, PM2.5and PM10were collected by using two MOUDIs (micro-orifice uniform deposit impactor, MSP 110) and a Dichotomous PM10sampler (Andersen SA-241) while real-time size distributions of particles of 5.5e350 nm in diameter were measured by an SMPS (scanning mobility particle sizer, TSI 3936). Filter samples were analyzed for gravimetric mass and chemical compositions, including organic carbon (OC), element carbon (EC), water-soluble ions and trace elements. Meteorology parameters and gaseous O3and CO concentrations were also monitored along with the SMPS data for studying particle nucleation, condensation, SOA (secondary organic aerosol) formation and long-range air pollutant transport at the LABS. SMPS data showed that nanoparticle concentrations at the LABS remained rela-tively stable at low level (w300e500 #/cm3_{) during the nighttime (22:00}_{e04:00), increased during} daytime, and reached a maximum (w2000e4000 #/cm3_{) in the afternoon (12:00e16:00). The NMD} (number median diameter) showed an opposite trend with the peak number concentrations observed in the afternoon corresponding to the smallest NMD (20e40 nm). These results indicate the dominance of local sources rather than the transport from other atmospheric air because that the lifetime of nano-particles was only few minutes. Chemical analysis offilter samples showed that the concentrations of trace elements K and Mn, which serve as biomass burning markers, were elevated in thefine particle fractions during November 9e12th when the air mass passed through South and Southeast Asia prior to reaching the LABS. The concentrations of K and Mn would have been low if the aerosols had local origins The biomass burning derived K was found in allfine particle samples at the LABS suggesting that the free troposphere around Taiwan is frequently impacted by the long-range transport of biomass burning plumes via the westerly winds.

* Corresponding author. Tel.: þ886 3 573 1880; fax: þ886 3 572 7835. E-mail address:cjtsai@mail.nctu.edu.tw(C.-J. Tsai).

Contents lists available atSciVerse ScienceDirect

Atmospheric Environment

j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / a t m o s e n v

(2)

1. Introduction

Atmosphericﬁne and ultraﬁne particles are largely contributed

by the burning of fuels (fossil fuels and biofuels used in power

plants, industry and motor vehicles), forestﬁre and crop residue

burning (the so-called biomass burning, any deliberate burning of agricultural residues for land clearing and land-use change). Particle emissions due to the burning of fuel may not be well controlled due to weak regulations (Li et al., 2011a,b) while those from the burning of biomass are completely unregulated. These combustion particles may act as the cloud condensation nuclei (CCN) responsible for the“cloud-climate” effect and, locally, could

be hazardous to public health (Hsu et al., 2009; Kavouras et al.,

1998; Li et al., 2011a,b). Scientists have estimated that humans are responsible for about 90% of biomass burning with only a small

percentage of natural ﬁres contributing to the total amount of

vegetation burned (http://earthobservatory.nasa.gov/Features/

BiomassBurning/).Streets et al. (2003)estimated that the annual total amount of biomass burned in South and Southeast Asian countries including China, India and Indo-China peninsula coun-tries was about 500 Tg which was comparable to the total annual biofuel burned in this area, 1300 Tg. In addition to biomass burning, the industrialization of East China has also contributed massive quantities of anthropogenic pollutants into the troposphere and inﬂuenced the air quality of East Asian countries (Hsu et al., 2009; Li et al., 2011a,b).

Taiwan is located on the edge of the western Paciﬁc Ocean and

adjacent to the Southeast and South Asian countries. The southwest monsoon prevails in summer in Taiwan, whereas the northeast monsoon prevails between fall and spring. At higher elevations around Taiwan the westerlies, wind originating from the Mediter-ranean Sea and Northern Africa which then passes through the south of Tibet Plateau, prevail at most the time, except in summer. After entering China the westerlies bring moist air to Southern China and Indochina Peninsula. Because of geographical location and meteorological conditions, it has been demonstrated that the

air quality of Taiwan can be inﬂuenced by South and Southeast

Asian biomass burning and industrial and power plant emissions from East China (Andrews et al., 2011; Hsu et al., 2009; Lee et al., 2011; Li et al., 2011a,b; Lin et al., 2009; Ou Yang et al., 2012; Sheu et al., 2010; Wai et al., 2008).

In order to understand the long-term free atmospheric Hg produced by biomass burning from South and Southeast Asian

countries, Sheu et al. (2010) measured the gaseous elemental

mercury, reactive gaseous mercury and particulate mercury at a high alpine site in Taiwan, the Lulin Atmospheric Background Station (LABS, 2862 m a.s.l.). Because of its elevation, the local

anthropogenic pollution can only weakly inﬂuence the LABS thus

allowing one to differentiate pollutants from long-range transport (Cozic et al., 2008). They concluded that sampled Hg at the LABS might be attributed to the biomass burning of Southeast Asia. Yet, there was still no direct evidence that could demonstrate the collected Hg was from biomass burning. In a recent study biosmoke pollutants from East China, including those originating from biomass burning, were found to raise atmospheric K, Mn, Pb and As concentrations, which have been designated to be good Asian biosmoke markers, in theﬁne particles of Taiwan (Hsu et al., 2009). Therefore, by conducting size resolved chemical concentration measurements for particles in LABS more direct information of South and Southeast Asian biomass burning as well as the regional pollution from China may be obtained.

Because of the unique location of Taiwan and the high elevation of the LABS, it is worthwhile to conduct a comprehensive measurement of both particles and gaseous pollutants in the LABS. For particle measurement, size-resolved sampling and further

chemical analysis was conducted. Meteorological and gaseous pollutants data obtained from the LABS were used to correlate the

size distributions of ultraﬁne particles to potential sources. In

addition, back-trajectories of the air masses reaching the LABS were analyzed to correlate the characteristics of PMs sampled at LABS. The results of this study could help in the development of a long-range pollutant transport model and in the establishment of control methods by the Taiwan government.

2. Experimental methods 2.1. Site descriptions

The ﬁeld study was conducted at the LABS (23₂₈0₀₇00_N,

1205202500E), which is a two-story building on the summit of Mt. Front Lulin in Yu-Shan National Park in central Taiwan as shown in

Fig. 1. This station was established by Taiwan EPA in 2006 for complementing the global network of Global Atmospheric Watch (GAW) in the East Asia region, where no high-elevation baseline station was available before the appearance of the LABS.

The Tsuga and Picea coniferous forests dominate the

surrounding area of the site. In a conifer forest, aerosols can be produced directly from the conifer leaves (epicuticular wax) or indirectly through photooxidization and gas-to-particle conver-sion. These organic particles, the so-called SOA (secondary organic

aerosol), may act as cloud condensation nuclei (Kavouras et al.,

1998, 1999). Thus, biogenic aerosols may be produced in the vicinity of the station. Hiking is the only way to access the summit

and the nearest main road, which has a very low trafﬁc, is about

2 km (perpendicular distance) away from the LABS. In any case, the

existingﬂourishing forest sheltered the LABS from the pollutants

emitted by vehicles. Therefore, the local motor emissions were neglected in this study.

2.2. Sampling protocol

In this study, PM sampling was conducted in July and August of 2009 and September to November of 2010, during which two 96-h and four 72-h PM samples were taken. Real-time particle size distributions were measured continuously from July to August of 2009 and July to November of 2010, except when sporadic shutdown of power occurred. PM samples were collected by two 10-stage MOUDIs (Model 110, MSP Corp., MN, USA) and a

Dichoto-mous PM10sampler (Dichot, Model SA-241, Andersen Inc., Georgia,

USA) operated in parallel. In the MOUDIs, the 3.2

m

m cutsize stages

were replaced with 2.5

m

m cutsize stages and the nozzle plates of

stage 10 with 56 nm cutsize were removed. Thus, the cutsizes of the MOUDIs were 18, 10, 5.6, 2.5, 1.8, 1.0, 0.56, 0.32, 0.18, and 0.1

m

m.

That is, the after ﬁlters sampled particles smaller than 100 nm

(PM0.1). Because pressure and temperature at the LABS, which

averaged about 0.7 atm and 10C, were low the actual operational

ﬂow rates of the samplers were adjusted based on the ambient conditions while the concentrations of PMs were calculated based on the standard condition of 25C and 1 atm. The PM2.5and PM10

concentrations were calculated based on the mass distributions of

MOUDI and the collection efﬁciency curves of the EPA PM2.5Well

Impactor Ninety-Six (Peters et al., 2001) and the Hi-Vol Sampler

(McFarland et al., 1984), respectively. The Dichot was used to collect PM_2.5e10and PM2.5samples. An SMPS (Model 3936, TSI Inc., MN,

USA) equipped with a Nano- or Long-DMA (TSI Model 3085 or

3081) and an Ultraﬁne Condensation Particle Counter (UCPC, TSI

Model 3776) was used to measure the size distribution of particles from 8.5 to 350 nm with the Long-DMA or, alternatively,

(3)

In the_{ﬁrst MOUDI (M1) silicone grease coated aluminum foil}

was used as the impaction substrate in 0e9 stages to reduce solid

particle bounce (Chen et al., 2011) and 2 quartzﬁlters (Tissuquartz 2500QAT-UP, 7201 & 7202, Pall Corp., New York, USA) in series were

used as the afterﬁlters for ultraﬁne particle OC and EC sampling

(QBQ method). In the second MOUDI (M2), Teﬂon ﬁlters

(TeﬂoR2PL047, Pall Corp., New York, USA) were used in every stage

and in the afterﬁlter. In the Dichot, quartz ﬁlters were used in the

both ﬁne and coarse particle channels (QBQ method). In every

sampling run, at least two Teflon filters, two quartz filters and two uncoated and grease coated Aluminum foils were used as labora-tory andfield blanks for gravimetric or chemical analysis.

Meteorological parameters at the LABS, including temperature, relative humidity, solar radiation, wind direction and wind speed, were measured continuously by various research groups (Lee et al., 2011; Ou Yang et al., 2012; Sheu et al., 2010) and were used in the current study. In addition, gaseous pollutant data monitored at the

station, such as O3 and CO, were also used to help elucidate the

current experimental data. 2.3. Sample analysis

A detailed description of the sample analyses can be found else-where (Chen et al., 2010) and is described brieﬂy in the following. All

aluminum foil and Teﬂon ﬁlter PM samples were analyzed

gravi-metrically by a microbalance (Model CP2P-F, Sartorius, Germany).

Further chemical analysis was conducted for the Teﬂon (water

soluble ions and elements) and quartz (OC and EC) samples. Before weighing,ﬁlters were conditioned at least 24-h in a temperature and

relative humidity controlled room (22 1_{C, 40}_{3% RH). The PM}

mass of each stage, 0e9, of M1 and M2 was compared to each other to

examine whether particle bounce occurred in M2, using M1 as the reference MOUDI (Chen et al., 2011). After gravimetric analysis, the Teflon filter was cut into equal halves using a Teflon coated scissors. One half was analyzed by ICP-MS (Elan 6100, PerkinElmer, Waltham, Massachusetts, USA) for elements while the other half was analyzed by ion chromatography (IC, Model DX-120, Dionex Corp, Sunnyvale, CA) for water soluble ionic species. The quartzfilter samples of the M1 (PM0.1) and Dichot (PM2.5and PM2.5e10) were analyzed by the

thermal-optical reﬂectance (TOR) method for OC and EC

concen-trations. In accordance withTurpin and Lim (2001),Kavouras et al. (1999)andKavouras and Stephanou (2002), the POM (particulate organic mass) to POC (particulate organic carbon) ratio at LABS was assumed to be 1.4 for all PMs size fractions since pinic acid and pinonic acid (both belong to carboxylic acids which contain less organic oxygen) were the dominant organic compounds in the conifer forest. In order to correlate the aerosol collected at the LABS to biomass burning emission from South and Southeast Asian countries and other sources, air mass back-trajectory analysis was conducted using the NOAA HYSPLIT model.

3. Results and discussion

3.1. PM concentration and mass distribution

The gravimetric analysis of all laboratory and ﬁeld blanks

showed that the weight differences between pre- and

post-sampling were less than 1

m

g, which was much lower than the

lowest PM0.1mass in this study. In addition, particles collected in

each impaction stage always had a higher mass than that at the afterﬁlter (PM0.1). Therefore, the data of the gravimetric analysis in

this study were reliable. Fig. 1. Location of Taiwan and the Lulin Atmospheric Background Station (LABS).

(4)

Throughout the majority of the sampling campaign, a close agreement between PM mass collected on the same stages of M1 and M2 was observed in all samples, with relative differences being <10%, due to the relatively high RH at the LABS which signiﬁcantly

reduced solid particle bounce. However, during Nov. 9thwNov.

12th, lower RH of less than 65% persisted for about 40 h during the 72-h sampling, resulting in a slightly larger difference between M1 and M2. Nevertheless, the maximum difference was only 15%, which occurred at stage 3 (PM2.5e5.6), indicating that the effect of

solid particle bounce in M2 could be neglected in this study.

Table 1summarizes the sampling date, start/end time, weather condition and the corresponding PM10, PM2.5and PM0.1

concentra-tions of the six samples. The PM0.1was 0.08e0.2

m

g m3at the LABS,

much lower than that in urban areas (1_e3

m

g m3,Chen et al., 2010). Low PM0.1concentrations, 0.08

m

g m3, as well as relatively low PM10

and PM2.5concentrations were observed in samples (b) and (e),

which was due to precipitation scavenging. For the samples taken in sunny days (samples a, c, d and f), the average concentrations of PM10,

PM2.5and PM0.1were 7.18, 5.46 and 0.15

m

g m3, respectively. The

average PM10concentration of the four sunny measurements was

very close to that measured by the R&P TEOM 1400a (Rupprecht & Patashnick Co., Inc., East Greenbush, NY, USA) of the LABS, which was 8.14

m

g m3, demonstrating that the PM10was accurately sampled by

the MOUDIs. However, the current PM10concentration was slightly

higher than that observed at a similar high alpine site (Jungfraujoch) inCozic et al. (2008), which was only 4.5

m

g m3. This could be attributed to the fact that the LABS received more long-range transport pollutants than Jungfraujoch.

Fig. 2compares the mass distributions of samples (a), (c), (e) and (f) which represent the distributions obtained on sunny days in summer (sample a), rainy days in fall (sample e) and sunny days in fall (samples c and f), respectively. Distributions of samples (b) and (d) are not shown because they have similar distributions to those of samples (e) and (c), respectively. On sunny days, distributions from samples (a), (c) and (f) all have a single mode with a MMAD

(mass median aerodynamic diameter) close to 1

m

m. The

general-unimodal distribution observed at LABS was in agreement with

that was found byLeaitch and Isaac (1991)who attributed this to

the precipitation scavenging and deposition of smaller and larger particles in the middle troposphere above the clouds. Although the size distribution of sample (c) demonstrated a goodﬁt with a single mode, a slight increase in the concentrations of both coarse andﬁne particle fractions was observed. This was attributed to the LABS

receiving aged maritime airs from the Paciﬁc Ocean (chemical data

will be shown later). In contrast, the distributions obtained on rainy days had both accumulation and coarse modes (samples b and e).

This was becauseﬁne particles contained inorganic salts,

under-went deliquesce under high RH and grew in size. 3.2. Size distribution of ultraﬁne particles

Based on the SMPS data, this study found that the size distri-bution of ultraﬁne particles at LABS varied dramatically with time

and was a function of various factors, including atmospheric air mass, temperature, RH, solar radiation, wind direction and speed,

O3 concentration, etc. However, a quantitative evaluation of the

relative importance of these factors is not possible at present.Fig. 3

represents three typical size distributions observed at the LABS, which were obtained in the sunny daytime (thick line) and night-time (thin line) at LABS on July 7th and August 24th of 2009 and Nov. 11th of 2010.

The three daytime distributions all show a nuclei mode at about 15 nm. Because the lifetime of these small particles is very short (normally on the order of minutes), the data suggest the existence of a local source. The nuclei mode particles can be produced only by combustion or via homogeneous or heterogeneous gas-to-particle conversion. However, no combustion event occurred in the vicinity of the sampling site. Thus, the nanoparticles could only be attributed to in-situ formation via gas-to-particle conversion from biogenic emissions. The differences in the peak concentrations of the nuclei modes between the daytime distributions were mainly due to the varying intensities of the photochemical reaction and the

wind speed which affects dilution. An obviously higher O3

consumption (O3 concentration 19/ 10 ppb in 3 h) and lower

wind speed (<1 m s1) was observed on Aug. 24th relative to that

observed on the other two days. Higher consumption of O3would

enrich the formation of organic acid particles (Kavouras et al., 1999;

Kavouras and Stephanou, 2002) and the weaker dilution effect facilitated the accumulation of newly formed particles, explaining the much higher PM0.1concentration on Aug. 24th. The distribution

Table 1

PM concentrations (mg m3) of all six samples.

Sample Date Start/end

time

Weather PM10 PM2.5 PM0.1 (a) July 6the9th, 2009 18:00 Sunny 8.33 6.17 0.20 (b) Aug. 4the8th, 2009 18:00 Rainy 2.36 1.44 0.08 (c) Sep. 29theOct. 2nd, 2010 12:00 Sunny 7.87 5.86 0.11 (d) Oct. 2ndeOct. 5th, 2010 13:00 Sunny 6.38 4.91 0.12 (e) Nov. 5theNov. 9th, 2010 16:00 Rainy 2.21 1.55 0.08 (f) Nov. 9theNov. 12th, 2010 17:00 Sunny 6.12 4.90 0.18

Fig. 2. Comparison of mass distributions sampled on a sunny day in summer (a), rainy day in fall (e) and sunny day in fall (f).

Fig. 3. Typical size distributions and total number concentrations of ultraﬁne particles at the LABS.

(5)

of Nov. 11th (14:00e15:00) represented a special case when both

nucleation and accumulation modes (peaks atw15 and w100 nm,

respectively) existed simultaneously. This was due to a dramatic

increase of RH (50/ 80%) and a reduced temperature (14 / 10_C)

which, due to the condensation effect, enriched the accumulation

mode (100e300 nm) with water droplets and other organic vapors

(Kavouras et al., 1999).

The particle number distribution of the July 7th, which repre-sents a case typical for the afternoon (13:00e16:00) of sunny days,

had an average total PM0.1 concentration of w1000 #/cm3. In

comparison, the particle number concentration was relatively lower at nighttime than that at daytime. On average, the

concen-tration was onlyw500 #/cm3_{. Moreover, a lower than typical PM}

0.1

concentration of_{<200 #/cm}3_{was observed on the nighttime of}

Aug. 24th when it rained.

Fig. 4shows the general time series of the number concentra-tion of PM0.1, NMD (number median diameter), temperature, RH, O3

concentration and solar radiationﬂux during Sep. 29th to Oct. 2nd, 2010. It is seen that the PM0.1concentration, temperature and solar

intensity peaked at around noon while the O3concentration, RH

and NMD was at a minimum near the noon. The noon peak concentration of PM0.1was 1000e1500 #/cm3and it was reduced to

200e500 #/cm3 _{near midnight. The NMD was reduced to}

20e40 nm near noon while it was increased with the increasing RH

to 80e100 nm near midnight. That is, the PM0.1 concentration

increased with the increasing temperature and solar intensity, but

NMD and O3showed a reverse trend. The observation of a reduced

O3concentration at the daytime (6 ame6 pm), also seen byOu Yang

et al. (2012)at the LABS, demonstrates that high solar radiation promoted a photo-oxidative reaction which produced the SOAs (secondary organic aerosols) at the LABS.

Other than the typical particle distribution and meteorological data shown inFig. 4, an unusually high CO concentration coinciding with a dramatic decrease of RH was observed and is presented in

Fig. 5. It can be seen the CO concentration was as high as 220 ppb in the evening of Nov. 9th, which was about two times the annual

average concentration ofw130 ppb measured at the LABS in 2010

(data were not published). Accompanying the high CO concentra-tion was a dramatic variaconcentra-tion of RH, which increased to 100% then very quickly reduced to as low as 10%. This was due to the subsi-dence of the upper tropospheric air characterized by a low RH. That is, the LABS could receive air pollutants transported from the upper free troposphere. InFig. 5, it is seen that the number concentrations

of PM0.1increased with increasing RH, which was different from

that observed inFig. 4. This unusual distribution can be found in

Fig. 3(14:00e15:00 of Nov. 11th). Moreover, the lowest NMD in this

time period was found to be relatively large (w60 nm,

condensa-tion effect). It is to be noted again, the current ultraﬁne particles basically were SOA and generated from local biogenic source mainly.

3.3. Chemical compositions of PMs

Based on the QBQ method, this study found the average OC positive artifact to be 240_{90% for PM}0.1, 18.6 12.7% for PM2.5,

and 13.4 9.3% in the PM10. Higher artifact was found in PM0.1and

on rainy days when PM mass was low.

Accounting for the correction of the positive artifact,Table 2

shows the comparison of the chemical compositions (%) including OM, EC, ions and elements for PM0.1, PM2.5and PM10between the

six experimental campaigns at the LABS. It can be seen that a good

chemical mass closure (OMþ EC þ ions þ elements) was obtained

for all PMs, where PM0.1 was 70.7e85.6%, PM2.5was 54.8e73.5%

and PM10 was 59e75.6%. The lower mass closure in PM2.5 and

PM10could be attributed to a relatively lower OM content. The use

of the OM/OC ratio of 1.4 for PM2.5and PM10may underestimate the

OM concentration because theﬁne and coarse particles collected in

LABS could contain a certain amount of atmospheric aged aerosols which were designated as high polarity, water soluble organic acids. The OM/OC ratio was found to be 1.8 inCozic et al. (2008). After adopting 1.8 for the OM/OC ratio, the present mass closure would increase by 11e21% for both PM2.5and PM10.

For July 6e9th (sample a), OM and EC data are not shown

because the quartzﬁlter samples were not taken. In this sample,

water soluble ions contributed 5.3, 12.2 and 11.4% and elements contributed 0.02, 7.3 and 10.1% to PM0.1, PM2.5and PM10,

respec-tively. From the air mass back-trajectory analysis, it was found that before reaching the LABS the air passed through the Southeast Provinces of China, where were highly polluted industrial areas.

Therefore, the polluted air impacted the LABS and signiﬁcantly

elevated the concentration of elements in PM2.5and PM10. This

ﬁnding is in agreement with which was observed on Mt. Tai (Li

et al., 2011a,b). The anthropogenic pollutants were sampled when the air mass passed through industrial areas before reaching the sampling site.

On the rainy days of Aug. 4e8th (sample b) and Nov. 5e9th

(sample e), OM was the most abundant constituent of the PMs.

This reﬂected that the local biogenic source was the most

predominate of the LABS particles and that most of the atmospheric particles were scavenged by rain. Unusually high EC content was

observed in PM0.1on Nov. 5e9th and in PM2.5and PM10on Aug.

4e8th, which might be due to the relatively low wet scavenging

efﬁciency for EC. Fig. 4. A typical time series of the PM0.1number concentration and NMD (number

median diameter) along with temperature, RH, O3concentration and solar radiation for Sep. 29theOct. 3rd, 2010.

Fig. 5. Time series of the number concentration of PM0.1, NMD, RH and CO concen-tration when the upper tropospheric air mass descended to LABS.

(6)

On the sunny days, Sep. 29th_{eOct. 2nd (sample c) and Oct.}

2nde5th (sample d), OM was the most abundant species in PM0.1

while water soluble ions dominated PM2.5and PM10. The above

data were in very good agreement with the results of the air mass back-trajectory analysis which showed that the air mass passed

over the Paciﬁc Ocean before reaching LABS. Therefore,

atmo-spheric aged aerosols inﬂuenced the LABS during the two sampling

periods while locally produced SOA still dominated PM0.1.

On Nov. 9e12th (sample f), when a high elevation air mass (from

trajectory data) inﬂuenced the LABS, OM was still the most

abundant species in PM0.1. It is to be noted that the concentrations

of the elements Al, Fe, K, Mn, Ba, Zn, Cu, Pb, Rb etc. were elevated,

accounting for 7% of PM0.1mass, which was 4.5 times higher than

the average (w1.5%) of other ﬁve samples. The results indicated

that the downward airﬂow brought a non-negligible amount of

aged nanoparticles enriched in these elements to the LABS. In the PM2.5and PM10, OM was still the most abundant species while ions

and elements were elevated. The ions were 24.6 and 23.5% and elements were 5.18 and 6.73% in PM2.5and PM10, respectively.

3.4. Chemical characteristics of PMs at LABS

Based on the data shown above, it can be concluded that the aerosols present at the LABS had the same origins as those were observed in Cozic et al. (2008). That is, a high alpine site could receive vertically transported aerosols from the boundary layer (in summer) and from the free troposphere (from fall to spring). In addition to the above two sources, locally produced biogenic

aerosols, which predominated in ultraﬁne fraction which

pre-dominated in ultraﬁne fraction, were another important source at

the LABS. Thus, the mixing of varying sources resulted in the complex chemical characteristics of PMs at the LABS and it is very

Fig. 6. Air mass back-trajectory analysis during Sep. 29theOct. 2nd, 2010 (NOAA HYSPLIT).

Table 2

Chemical composition (%) of PM0.1, PM2.5and PM10at LABS.

LABS Sample PM (mg m3) OM % EC % Ions % Elements % Total %

PM0.1 (a) 0.20 ea e 5.3 0.02 e (b) 0.08 70.0 7.3 5.1 1.3 83.7 (c) 0.11 55.6 14.0 8.1 2.5 80.2 (d) 0.12 46.7 10.8 11.5 1.7 70.7 (e) 0.08 54.4 17.0 9.0 2.2 82.6 (f) 0.18 61.8 5.4 11.4 7.0 85.6 PM2.5 (a) 6.17 e e 12.2 7.3 e (b) 1.44 48.0 12.1 5.4 2.5 68.0 (c) 5.86 28.4 3.5 30.9 2.1 64.9 (d) 4.55 24.9 5.7 40.9 2.0 73.5 (e) 1.55 43.2 5.1 4.6 1.9 54.8 (f) 4.50 31.7 3.4 24.6 5.2 64.9 PM10 (a) 8.33 e e 11.4 10.1 e (b) 2.36 40.0 14.2 4.3 2.1 60.6 (c) 7.87 26.5 3.3 29.2 2.5 61.5 (d) 5.79 28.5 4.9 39.5 2.7 75.6 (e) 2.21 47.4 4.3 5.2 2.1 59.0 (f) 6.12 35.4 2.9 23.5 6.7 68.5 a_{No data.}

Fig. 7. (a) Five-day air mass backward trajectory analysis and (b)fire events (dots indicate the fire locations retrieved from satellite, based on Fire Information for Resource Management System (FIRMS),http://earthdata.nasa.gov/data/nrt-data/firms

(7)

important to understand their contributions for the development

of regional and global transport models in the future (Li et al.,

2011a,b).

In samples (c) (Sep. 29theOct. 2nd) and (d) (Oct. 2nde5th),

a signiﬁcantly high content of ions in PM2.5 and PM10 and

a moderate content of ions in PM0.1were observed. Thisﬁnding

could be attributed to the origin of air mass.Fig. 6shows the air

trajectories, from the NOAA HYSPLIT Model, of Sep. 29theOct 2nd

(sample c), in which two-day backward trajectories with an interval of 12-h are presented. The air trajectories of sample (d) were not shown since they were similar to those of sample (c). InFig. 6, it is seen that the air mass came from the Paciﬁc Ocean. Therefore, LABS received marine aerosols during the sampling periods. Ionic anal-ysis found SO42was the most abundant species in the water soluble

ions, which agrees with our previousﬁnding (Lee et al., 2011). SO42

contributedw80% of the total ions in PM2.5for both samples (c) and

(d) and w40 and w50% in PM2.5e10 for samples (c) and (d),

respectively. In addition, it was found that coarse particles (PM2.5e10) had relatively higher concentrations of Cl, Naþand Kþ

than those ofﬁne particles in the two samples. These ﬁndings are in

accord with the observation inFitzgerald (1991)whereﬁne

parti-cles in the clean marine air were shown to be composed predom-inantly of non-sea-salt sulfate whereas sea salt was the most abundant species in coarse particles.

As mentioned earlier, LABS was inﬂuenced by a high elevation

air mass during Nov. 9e12th (sample f), when the dry and dense air

subsided to the LABS. Fig. 7(a) shows the _{ﬁve-day backward}

trajectories obtained from the NOAA HYSPLIT Model with an interval of 24-h during the sampling period of sample (f) and

Fig. 7(b) shows theﬁre events during these days. It can be seen that

the air mass trajectories have a direction similar with the

west-erlies. In addition, they rose to a high elevation ofw5000e6000 m

a.s.l. in the middle China areas and then descended to w3000e4000 m a.s.l. in the southeast China areas before reaching the LABS. Therefore, it can be concluded that the air mass was driven by the westerlies and then impacted the LABS during the

measurement (f). In addition, it can be seen fromFig. 7(b) that

southeast China was riddled withﬁre events, which was helpful in

interpreting why the sample (f) contained high concentrations of biomass burning makers, as will be shown later. Furthermore, according to the Fire Information for Resource Management System (FIRMS), it was seen that smoke covered southeast China (data not shown) during the sampling (f).

Fig. 8 depicts the size distributions of K, Mn, Pb and As concentrations for the sample (f) and the average distributions for other samples. All these elements in sample (f) show a bimodal pattern, but the former two elements display a typical distribution

with the peak concentrations at 2.5e5.6

m

m and around 1

m

m. The

latter two elements are similar to a unimodal pattern with a pronounced peak at 1

m

m and a minor peak in the coarse fraction. The size distribution of K observed in the sample (f) is very similar to that measured in northern Taiwan during the northeastern

monsoon when the Asian outﬂows prevail and has been applied as

an indicator of Asian biosmoke aerosols (Hsu et al., 2009). The

average distributions of the other samples resembled that of sample (f) except that some distinctive features were found. First,

the average distribution seemed to be relatively_{ﬂat in comparison}

to sample (f). Second, the average distribution of Mn had the largest peak at the smaller size; while the averaged distribution of potas-sium peaked at the coarse size. In addition, it was noted that there

(8)

were higher concentrations for these selected metals (K, Mn, Pb and As) in all size fractions of sample (f). Sea salt and mineral dust aerosols were also the signiﬁcant contributors for aerosol K apart from biomass burning (Andreae, 1983; Park et al., 2007; Hsu et al.,

2009). Thus, the size distribution patterns of K can reﬂect the

relative dominance of various sources. For instance, natural sources would preferentially contribute to the coarse fraction while

anthropogenic sources would reside inﬁner sized aerosols (Hsu

et al., 2009). Finer particles containing Mn might also be attribut-able to biomass burning (Viana et al., 2008). Accordingly, we may conclude that particle emissions resulting from biomass burning

existed throughout the sampling periods, with the inﬂuence being

most evident in sample (f). Along with biomass burning, As and Pb in the samples represented the contributions of other

anthropo-genic sources such as coal burning and industrial emissions (Hsu

et al., 2009; Mkoma et al., 2009; Christian et al., 2010).

Fig. 9shows the correlation between Mn and K mass

concen-tration in sample (f) and the average of samples (b_{ee). It can be}

seen the concentrations exhibit a linear correlation. For sample (f) the slope of this line is 0.038 which was much higher than that in the average crust composition (slope of 0.05_{e0.16) but was close to} 0.041e0.045 previously reported in Taipei aerosols, likely serving as

a signature of Asian biosmoke particulate pollutants (Hsu et al.,

2009). In comparison, the data for samples (b-e) had a smaller

slope (or Mn/K ratio) of 0.032. The result suggests that the Mn/K ratio of around 0.04 could serve as a tracer of biomass burning aerosols.

Furthermore, another marker of biomass burning, CO, was observed to be elevated to 220 ppb (Fig. 4) during the Nov. 9e12th measurement (sample f). That is, biomass burning from South and

Southeast Asian was able to in_{ﬂuence all PMs at the LABS when the}

westerlies prevailed. Becauseﬁne potassium derived from biomass

burning was measured throughout all samples, it suggested that the air quality at high mountains over Taiwan, as well as that at ground level, is impacted frequently by biomass burning pollution from the long-range transport by the westerlies.

4. Conclusions

In this study, the average concentration of PM0.1, PM2.5and PM10

at the LABS were 0.13, 4.0 and 5.4

m

g m3, respectively. In addition

to the gravimetric analysis, size resolved chemical compositions including OC, EC, water-soluble ions and trace elements were also determined. Good chemical mass closures for PMs were obtained after taking into account the OC artifact using OM/OC¼ 1.4 in PM0.1

and OM/OC¼ 1.8 in PM2.5and PM10, which were 80.6 5.9% for

PM0.1, 81.2 6.8% for PM2.5, and 81.0 6.9% for PM10.

The size distributions of ultraﬁne particles at the LABS were

explained with the aid of meteorological data, including tempera-ture, RH, wind speed and direction radiationﬂux, and gas pollutant

concentrations, including CO and O3concentrations. It was found

that the concentration of ultraﬁne particles was increased while the NMD decreased with increasing temperature, radiation intensity

and O3consumption.

Chemical constituents of PMs showed that the aerosol particles at the LABS originated from both on-site biogenic sources and

long-range transport. Speciﬁcally, the local sources were dominant on

rainy days, corresponding to a high OC content in particles when the atmospheric aerosols were scavenged to a large extent. On the sunny days, the two sources were found to be equally important. Backward trajectory analyses in conjunction with the results of chemical analysis were able to interpret the sources well. When the

air mass passed over the Paciﬁc Ocean ﬁne particles were

predominantly composed of non-sea-salt sulfate, while sea salt was the most abundant species in the coarse particles. In addition, biomass markers K and Mn were found to be elevated in all PM fractions, especially in theﬁne fraction, when the westerlies

pre-vailed. Because the biomass burning derived_{ﬁne potassium was}

measured throughout all samples, it suggests that high elevations (i.e., free troposphere) around Taiwan are perhaps impacted frequently by biomass burning from Southeastern Asia via long-range transport.

Acknowledgement

Theﬁnancial support of the Taiwan EPA (EPA-98-U1U1-02-103

and EPA-99-U1U1-02e103) is gratefully acknowledged.

References

Andreae, M.O., 1983. Soot carbon and excessﬁne potassium: long-range transport of combustion-derived aerosols. Science 220, 1148e1151.

Andrews, E., Ogren, J.A., Bonasoni, P., Marinoni, A., Cuevas, E., Rodríguez, S., Sun, J.Y., Jaffe, D.A., Fischer, E.V., Baltensperger, U., Weingartner, E., Collaud Coen, M., Sharma, S., Macdonald, A.M., Leaitch, W.R., Lin, N.-H., Laj, P., Arsov, T., Kalapov, I., Jefferson, A., Sheridan, P., 2011. Climatology of aerosol radiative properties in the free troposphere. Atmospheric Research 102, 365e393.

Chen, S.-C., Tsai, C.-J., Huang, C.-Y., Chen, H.-D., Chen, S.-J., Lin, C.-C., Chou Charles, C.-K., Lung, S.-C., Roam, G.-D., Wu, W.-Y., Smolik, J., Dzumbova, L., 2010. Chemical mass closure and chemical characteristics of ambient PM0.1, PM2.5and

PM10in a highway tunnel and at a roadside. Aerosol Science Technology 44,

713e723.

Chen, S.-C., Tsai, C.-J., Chen, H.-D., Huang, C.-Y., Roam, G.-D., 2011. The inﬂuence of relative humidity on nanoparticle concentration and particle mass distribution measurements by the MOUDI. Aerosol Science Technology 45 (5), 713e723. Christian, T.J., Yokelson, R.J., Cárdenas, B., Molina, L.T., Engling, G., Hsu, S.-C., 2010.

Trace gas and particle emissions from domestic and industrial biofuel use and garbage burning in central Mexico. Atmospheric Chemistry and Physics 10, 565e584.

Cozic, J., Verheggen, B., Weingartner, E., Crosier, J., Bower, K.N., Flynn, M., Coe, H., Henning, S., Steinbacher, M., Henne, S., Coen, M.C., Petzold, A., Baltensperger, U., 2008. Chemical composition of free tropospheric aerosol for PM1 and coarse mode at the high alpine site Jungfraujoch. Atmospheric Chemistry and Physics 8, 407e423.

Fitzgerald, J.W., 1991. Marine aerosols: a review. Atmospheric Environment 25A, 533e545.

Hsu, S.-C., Liu, S.-C., Huang, Y.-T., Chou, C.C.-K., Lung, S.-C.C., Liu, T.-H., Tu, J.-Y., Tsai, F., 2009. Long-range southeastward transport of Asian biosmoke pollution: signature detected by aerosol potassium in Northern Taiwan. Journal of Geophysical Research 114, D14301.

Kavouras, I.G., Mihalopoulos, N., Stephanou, E.G., 1998. Formation of atmospheric particles from organic acids produced by forests. Nature 395, 683e685. Kavouras, I.G., Mihalopoulos, N., Stephanou, E.G., 1999. Secondary organic aerosol

formation vs primary organic aerosol emission: In situ evidence for the Fig. 9. Comparison of K and Mn concentrations between sample f and samples bee.

(9)

chemical coupling between monoterpene acidic photooxidation products and new particle formation over forests. Environmental Science Technology 33, 1028e1037.

Kavouras, I.G., Stephanou, E.G., 2002. Direct evidence of atmospheric secondary organic aerosol formation in forest atmosphere through heteromolecular nucleation. Environmental Science Technology 36, 5083e5091.

Leaitch, W.R., Isaac, G.A., 1991. Tropospheric aerosol size distributions from 1982 to 1988 over Eastern North America. Atmospheric Environment 25A, 601e619. Lee, C.-T., Chuang, M.-T., Lin, N.-H., Wang, J.-L., Sheu, G.-R., Chang, S.-C., Wang, S.-H.,

Huang, H., Chen, H.-W., Liu, Y.-L., Weng, G.-H., Lai, H.-Y., Hsu, S.-P., 2011. The enhancement of PM2.5 mass and water-soluble ions of biosmoke transported from Southeast Asia over the Mountain Lulin site in Taiwan. Atmospheric Environment 45, 5784e5794.

Li, W., Zhou, S., Wang, X., Xu, Z., Yuan, C., Yu, Y., Zhang, Q., Wang, W., 2011a. Integrated evaluation of aerosols from regional brown hazes over northern China in winter: concentrations, sources, transformation, and mixing states. Journal of Geophysical Research 116, D09301. http://dx.doi.org/10.1029/ 2010JD015099.

Li, W.-J., Zhang, D.-Z., Shao, L.-Y., Zhou, S.-Z., Wang, W.-X., 2011b. Individual particle analysis of aerosols collected under haze and non-haze conditions at a high-elevation mountain site in the North China plain. Atmospheric Chemistry and Physics 8, 407e423.

Lin, C.-Y., Hsu, H.-M., Lee, Y.-H., Kuo, C.-H., Sheng, Y.-F., Chu, D.A., 2009. A new transport mechanism of biomass burning from Indochina as identiﬁed by modeling studies. Atmospheric Chemistry and Physics 9, 7901e7911. Mkoma, S.L., Maenhaut, W., Chi, X., Wang, W., Raes, N., 2009. Characterisation of

PM10atmospheric aerosols for the wet season 2005 at two sites in East Africa.

Atmospheric Environment 43, 631e639.

McFarland, A.R., Ortiz, C.A., Bertch Jr., R.W., 1984. A 10 mm cutpoint size selective inlet for Hi-Vol Samplers. Journal of the Air Pollution Control Association 34, 544e547.

Ou Yang, C.F., Lin, N.H., Sheu, G.R., Lee, C.T., Wang, J.L., 2012. Seasonal and diurnal variations of ozone at a high-altitude mountain baseline station in East Asia. Atmospheric Environment 46, 279e288. http://dx.doi.org/10.1016/ j.atmosenv.2011.09.060.

Park, R.J., Jacob, D.J., Logan, J.A., 2007. Fire and biofuel contributions to annual mean aerosol mass concentrations in the United States. Atmospheric Environment 41, 7389e7400.

Peters, T.M., Vanderpool, R.W., Wiener, R.W., 2001. Design and calibration of the EPA PM2.5Well Impactor Ninety-Six (WINS). Aerosol Science Technology 34, 389e397.

Sheu, G.-R., Lin, N.-H., Wang, J.-L., Lee, C.-T., Ou Yang, C.-F., Wang, S.-H., 2010. Temporal distribution and potential sources of atmospheric mercury measured at a high-elevation background station in Taiwan. Atmospheric Environment 44, 2393e2400. Streets, D.-G., Yarber, K.-F., Woo, J.-H., Carmichael, G.R., 2003. Biomass burning in Asia: annual and seasonal estimates and atmospheric emissions. Global Biogeochemical Cycles 17, 1099.

Turpin, B.J., Lim, H.J., 2001. Species contributions to PM2.5mass concentrations:

revisiting common assumptions for estimating organic mass. Aerosol Science Technology 35, 602e610.

Viana, M., López, J.M., Querol, X., Alastuey, A., García-Gacio, D., Blanco-Heras, G., López-Mahía, P., Piñeiro-Iglesiasc, M., Sanz, M.J., Sanz, F., Chi, X., Maenhaut, W., 2008. Tracers and impact of open burning of rice straw residues on PM in Eastern Spain. Atmospheric Environment 42, 1941e1957.

Wai, K.M., Lin, N.-H., Wang, S.-H., Dokiya, Y., 2008. Rainwater chemistry at a high altitude station, Mt. Lulin, Taiwan: comparison with a background station, Mt. Fuji. Journal of Geophysical Research 113, D06305. http://dx.doi.org/10.1029/2006JD008248.