Impact of Asian dust and continental pollutants on cloud chemistry observed in northern Taiwan during the experimental period of ABC/EAREX 2005



[1] Observations of particulate matter (PM), vertical cloud and aerosol structure and cloud water chemistry in northern Taiwan were conducted during the ABC/EAREX 2005 period. Five Asian continental outflow regimes reaching Taiwan were identified. One was coupled with a dust storm observed not only at Gosan, Korea, but also over Taiwan, suggesting the scope of its regional impact. The arrival of the dust event was determined by lidar, cloud water, and surface PM measurements. When continental outflow events correspond to the presence of significant dust concentrations, air quality can be drastically worsened due to high levels of PM. PM10 (PM with aerodynamic diameters < 10 μm), pH, conductivity, and ion concentrations of cloud water increased drastically near the dissipating stage of the frontal passage/cloud event for the dust case. Cloud water may have become acidified by pollution from industrial and urban regions along the coast of eastern China. Nevertheless, abundant Ca2+ contributed to the neutralization of acidic cloud water during the dust stage. The much higher aerosol and chemical loading injected into these clouds caused an enrichment effect in the cloud water, which can double the cloud loading of total ions, when Ca2+ increases by approximately 7 times.

1. Introduction

[2] In recent decades, gas-phase and particulate-based pollutants from East Asia have been released into the atmosphere in increasing amounts due to rapid regional economic development [e.g., Akimoto, 2003; Wild and Akimoto, 2001]. It has been estimated that one fourth to one third of the total global emissions of SO2, organic matter, soot, and dust result from urban industrial output, large desert areas, and extensive forest, and agriculture fires in northern China [e.g., Chin et al., 2000, 2002; Ginoux et al., 2001, and references therein]. It has become evident that air pollution, both industrial and anthropogenic, plays an important role in modifying the atmospheric chemistry, radiation and climate on global [Intergovernmental Panel on Climate Change (IPCC), 2007] and regional [e.g., Ramanathan et al., 2001; Li et al., 2007, and references therein] scales.

[3] Aerosol transport, including dust from Asia, across the Pacific Ocean, is well documented from low-lying ground-based observations [e.g., Prospero et al., 1985, 2003; Zieman et al., 1995; Arimoto et al., 1996; Huebert et al., 2001; Liu et al., 2006] and elevated mountain sites [Wai et al., 2008]. Asian mineral dust can be transported by the prevailing westerlies in the spring [e.g., Zhang et al., 1997], increasing particulate matter concentrations [e.g., Liu et al., 2006; Park et al., 2008] and pH values of rainwater [Wang et al., 2002] and cloud water in downwind areas. Aircraft observations through Asian outflow over the Asian continent [Dickerson et al., 2007] and the northwest Pacific [Arimoto et al., 1997; Jordan et al., 2003; Maxwell-Meier et al., 2004], and in transpacific Asian plumes over the northeast Pacific [Andreae et al., 1988; Clarke et al., 2001; Price et al., 2003], have provided evidence of aerosol transport in the lower free troposphere. They reveal the spring months to be the time of maximum trans-Pacific transport of outflow and anthropogenic emissions associated with Asian dust, resulting in combined dust and pollution signatures in measurements of nominal air mass characteristics.

[4] Dust and biomass burning aerosols provide surface area for the uptake of gas species (i.e., sulfates and nitrates). Over time, these aerosols become more hygroscopic. As cloud condensation nuclei (CCN), this effect may enhance cloud formation potential during transport [Roberts et al., 2006]. The formation of cloud droplets depends, primarily, on the size and chemical composition of the particles and the supersaturation of water vapor. Dust particles, in particular, are well known to be active heterogeneous ice nuclei at relatively warm temperatures [DeMott et al., 2003; Sassen, 2005]. This attribute can enhance their indirect effect on cloud microphysical properties by influencing the thermodynamic phase state of hydrometeors. Clouds can contain high concentrations of acids and other dissolved ions, depending on recent trajectories [Lin and Saxena, 1991; Mohnen and Vong, 1993]. Therefore, the chemistry and microphysics of clouds have become increasingly important in understanding the influence of clouds on atmospheric chemistry, as well as air pollution on a local, regional or global scale [Menon and Saxena, 1998].

[5] The phenomenon of Atmospheric Brown Cloud (ABC) is an example of long-range transport and the merging of anthropogenic aerosols and gas phase emissions. During the INDOEX field campaign in February 1999, the global threat of ABC, which is most commonly associated with urban haze, was documented to span an entire continent and ocean basin [Ramanathan and Crutzen, 2003]. The East Asian Regional Experiment 2005 (EAREX 2005), conducted in March−April 2005 in the East Asian region, was organized under the UNEP/ABC-Asia project. It was followed by the ABC Maldives Monsoon Experiment 2004 (APMEX 2004), which focused on the climatic and environmental effects of air pollution in the Asian region [Nakajima et al., 2007].

[6] In this paper, we describe the only cloud chemistry measurements collected during the experimental period of ABC/EAREX 2005. We investigate a dust storm event that noticeably influenced air quality and cloud composition over northern Taiwan. This dust storm event was connected with a synoptic-scale weather system that developed during EAREX 2005 [Lee et al., 2007; Sawa et al., 2007; Takami et al., 2007; Choi et al., 2008]. Lidar and cloud chemistry measurements in northern Taiwan serve as ancillary data sets to help characterize the dust event. We also refer to previous studies reported for EAREX 2005 to illustrate the regional impact of this storm. An additional nondust continental outflow event during EAREX 2005 is described to contrast the cloud-chemical impact on northern Taiwan.

2. Methodologies

2.1. Taiwan Measurement Stations

[7] During the EAREX 2005 period, measurements of PM concentrations and cloud chemical parameters, as well as zenith lidar profiles of cloud and aerosol structure were collected in Taiwan at ground-monitoring stations operated by the Taiwan Environmental Protection Agency (EPA) and National Central University (NCU). The Taiwan EPA has operated a network of 73 ground-based air monitoring stations since 1994 around the island to form the Taiwan Air Quality Monitoring Network (TAQMN) (see more information). These stations continuously measure ambient concentrations of total PM10 and that for specific species, including SO2, NOx, CO, and O3. All stations are equipped with similar instrumentation. For PM10 measurements, this is the Thermo/R&P 1400a, with a detection limit of 0.5 μg m−3. In our work here, data were obtained from three air quality stations in northern Taiwan during the EAREX 2005 and then combined with cloud chemical samples at Mount Bamboo (25.18°N, 121.52°E; 1.100 km MSL) and the lidar profiles collected from NCU (24.97°N, 121.18°E; 0.133 km MSL).

[8] Wan-Li air quality station (25.18°N, 121.68°E; 0.040 km MSL) is near sea level, and serves as an early detection site for continental outflow from Asian continent during the winter−spring monsoon season. The Mount Bamboo station is used to collect measurements of cloud water chemistry from an elevated locale and, therefore, serves as a contrasting site relative to Wan-Li station. Dust storm events encountered in Taiwan are normally associated with northeasterly flow during the winter monsoon, which is common between March and May. On average, there are four to five dust events and six dust days in a year in Taiwan [Liu et al., 2006]. The Wan-Li and Mount Bamboo sites, on the northeast coast and north of Taiwan, respectively, often experience the greatest air quality impact during dust storm season. Data sets from two other EPA air quality stations, i.e., Wu-Chiuan (WC; 24.95°N, 121.2°E; 0.141 km MSL) and Shi-Tuen (ST; 24.16°N, 120.62°E; 0.069 km MSL), are also considered to characterize the spatial/regional impact of the air mass intrusion. ST is located near an industrial area in central Taiwan, and data collected there are considered very sensitive to local emission influence. WC is near the NCU lidar station in a suburban area where less urban influence is expected. This information is summarized and detailed in Figure 1.

Figure 1.

Locations of monitoring stations at Gosan and across northern Taiwan for this study. Air quality stations (urban) are marked by a circle; air quality stations (background) are marked by squares; cloud chemistry station is marked by a triangle; and lidar station is marked by a star.

2.2. Cloud Chemistry

[9] During EAREX 2005, cloud water samples were collected at Mount Bamboo. This site, marked by a triangle in Figure 1, is frequently immersed in clouds. Lin and Peng [1999] concluded that cloud formation at this site during winter corresponds primarily with northeasterly monsoonal flow and frontal passages. Marine stratus and stratocumulus clouds are commonly observed. Marine stratus clouds are frequent near 2.0–3.0 km MSL, and stratocumulus clouds can extend to as high as, or even higher, than 5.0 km MSL. The formation of clouds due to orographic lifting effects may also occur at the site [Lin and Peng, 1999], and this scenario is characterized by the Froude number. When the index exceeds 0.5, flow is forced over the hill and lifting occurs [Smolarkiewicz and Rotunno, 1989]. However, Lin and Peng [1999] suggest that when northerly flow prevails during cloud events, cloud water collected at this site is typically not contaminated by local anthropogenic emissions, which originate primarily from the Taipei basin situated to the south of Mount Bamboo.

[10] Cloud water samples were collected approximately every hour using an ASRC-type (Atmospheric Science Research Center, State University of New York) collector [Castillo et al., 1983]. Some samples were collected with a longer sampling time. The samples were immediately prepared for pH and conductivity measurements after their collection. Water was filtered through a 0.45 μm pore sized mixed cellulose ester filter and collected directly into a sampling bottle. The bottle was then sealed and refrigerated at 4°C for later ionic analysis via conventional ion chromatography consistent with GAW/WMO (Global Atmosphere Watch/World Meteorological Organization [see Mohnen et al., 1993]) sampling protocols. The pH of the cloud water sample was measured by a pH meter (Mettler Toledo, Model MPC 227). Samples were analyzed for ion concentrations of Cl, NO3, SO42−, NH4+, Na+, K+, Mg2+, and Ca2+ using Ion Chromatography (Dionex Model DX-100). Laboratory quality assurance procedures included the routine running of blanks and control samples, as well as replicated samples. A description of these procedures is summarized by Lin et al. [1999].

[11] Cloud liquid water content (LWC; g m−3) can be estimated using an empirical equation,

equation image

where M (in g h−1) is the amount of cloud water collected within sampling time. To parameterize this equation, it was determined that a = 0.0016 (in h m−3) and b = 0 for fixed interception at origin based on a regression analysis between the LWC detected by PVM-100 (Particulate Volume Monitor Model 100 [Gerber, 1991]) and corresponding M collected hourly using the ASRC collector at Mount Bamboo during 1996–1998 winter seasons. R2 for the linear fit question was 0.93. A description of this relationship is given by Peng [1999].

2.3. Lidar

[12] Lidar measurements are frequently used to profile the vertical structure of tropospheric aerosol, cloud and water vapor [Welton et al., 2000; Schmid et al., 2003]. They can also be used to determine the height of the boundary layer [e.g., Parikh and Parikh, 2002]. In this study, normalized relative backscatter (NRB) was collected using a ground-based lidar at NCU. The EPA/NCU Micro-Pulse Lidar (MPL; 0.527 μm) system is a member of the National Aeronautics and Space Administration (NASA) Micro-Pulse Lidar Network (MPLNET) project [Welton et al., 2001]. The instrument is joint-operated by the Taiwan EPA and NCU. The standard instrument design, basic processing algorithm, and the NRB parameter are described by Campbell et al. [2002]. Our system supports long-term measurements with 1.0 min temporal and 0.075 km vertical resolutions.

2.4. Trajectory Analysis

[13] Trajectory analysis is often used to diagnose the movement of pollutants. Here, the Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model developed by NOAA Air Resources Laboratory (R. R. Draxler and G. D. Rolph, HYSPLIT4 (HYbrid Single-Particle Lagrangian Integrated Trajectory) model, 2003, is applied to calculate air mass backward trajectories. The meteorological data used to initialize HYSPLIT is obtained from the NCEP Global Data Assimilation System (GDAS) [Kalnay et al., 1996]. This archived data set is referred to as the final (FNL) archive with a horizontal resolution of 191 km and a temporal resolution of 6 h on 13 mandatory pressure levels and the surface level. Based on the current knowledge prescribed to the application of air mass back trajectories, corresponding errors approaching 20% of the transported distances are suggested [Stohl, 1998].

3. Results and Discussion

[14] We first describe the meteorological conditions present during the study period, including synoptic weather observed and station meteorological data. This will form the basis for subsequent description of continental outflow regimes in springtime. We then describe observations of a dust storm event that occurred during the EAREX 2005 period. Variations of PM10 at Wan-Li, WC, and ST combined with lidar and cloud chemistry measurements at NCU and Mount Bamboo, respectively, are discussed to characterize a postfrontal passage dust event during the EAREX 2005 period. Meteorological data and trajectories are used to track the movement of the air masses during each event. Finally, a nondust event is described to distinguish the characteristics of our measurements with Asian continental outflow types.

3.1. Overall Meteorological Conditions During EAREX 2005

[15] Hourly temperatures, sea level pressure (SLP), relative humidity (RH), precipitation, and winds in March 2005 over northern Taiwan are shown in Figure 2. Northeasterly winds prevail over northern Taiwan at this time. In contrast, winds from other directions are weaker and more irregular. At the same time, these sites show consistent trends in temperature, SLP and precipitation, with higher SLP, lower temperatures and more prolonged precipitation occurring during the first half of the month. The passage of frontal systems initiates rainfall, low temperatures and northeasterly winds.

Figure 2.

Hourly temperature, sea level pressure (SLP), RH, rainfall, and wind field measurements at (a) Wan-Li, (b) Mount Bamboo, and (c) NCU in March 2005. A–E denote five cases defined in section 3.1.

[16] Five cases of continental outflow following frontal passages over northern Taiwan can be identified from Figure 2. The time period for each case is determined according to the following two conditions: (1) stable northerly winds in Wan-Li and (2) backward trajectories at Wan-Li that previously did not advect over Taiwan. The five continental outflow periods are, thus, classified as: Case A: 0400 UTC 2 March to 1200 UTC 4 March; Case B: 0700 UTC 11 March to 2100 UTC 13 March; Case C: 1900 UTC 17 March to 0400 UTC 19 March; Case D: 0000 UTC 23 March to 0900 UTC 25 March; Case E: 1300 UTC 28 March to 1700 UTC 30 March. Identification of these outflow cases corresponds with those of previous studies [Lee et al., 2007; Sawa et al., 2007; Takami et al., 2007].

3.2. Case Study: Case C (17–19 March) Frontal Passage Coinciding With Dust Storm

[17] Although there were five continental outflow regimes preceded by frontal passages over Taiwan in March during the EAREX 2005 period, only one case (Case C) was associated with a dust storm observed both at Gosan, Korea and at the Taiwan sites. This event was also described by Lee et al. [2007] and Choi et al. [2008]. In this study, we investigate Case C to illustrate the scenario of pervasive continental outflow covering most of East Asia.

3.2.1. Meteorological Conditions

[18] During Case C, surface and 850 hPa level (∼1500 gpm) wind fields revealed similar patterns with little time lag, as seen in Figure 3. A frontal boundary was situated between an anticyclone and cyclone near Japan, which both later advanced eastward over the western Pacific. Combined visible and infrared satellite imagery, shown in Figure 4, show that the southern branch of the frontal system, with shallow associated cloudiness, passed Taiwan near 1200 UTC 17 March. By comparing the cloud features in these images, a homogeneous stratocumulus and stratus cloud can be seen covering southern China and extending toward northern Taiwan. Cloud top pressure data, also shown in Figure 4, indicate lower values at 1200 UTC 17 March (∼ 600 hPa) in Figure 4b, which increased to around 700 hPa by 0000 UTC 18 March in Figure 4c. This suggests some dissipation of the cloud system over time. From sounding profiles (not shown here) launched at the Taipei weather station (∼10 km south of Mount Bamboo), saturation with respect to liquid water (i.e., cloud base) was observed near 0.80 km at 1200 UTC 17 March, which then elevated to near 1.00 km in the 0000 UTC 18 March measurement. NCU lidar data depict a concurring cloud base distribution (Figure 5).

Figure 3.

(a−d) Mean sea level pressure (hPa) and surface wind flows (m s−1) and (e−h) mean geopotential high (gpm) and wind flows (m s−1) at 850 hPa during 17–18 March 2005.

Figure 4.

GOES-9 satellite visible images at 0000 and 1200 UTC; GOES-9 satellite IR images at 0000 UTC; and MODIS cloud top pressure (hPa) at daytime and nighttime during 17–18 March 2005. The GOES-9 images were provided by Korea Meteorological Administration. The cloud top pressure was obtained from the Giovanni online data system (, which is produced by MODIS-Aqua infrared retrieval methods both day (1200–1400 local time) and night (0000–0200 local time).

Figure 5.

Normalized relative backscatter (PhE km2μs−1μj−1) measured by the Micro-Pulse Lidar Network instrument at NCU during 16–19 March 2005.

[19] Conditions at Mount Bamboo showed prevailing northeasterly flow with an average wind speed of 9.1 m s−1, decreasing temperatures (to near 8°C), increasing SLP (from 1019 to 1023 hPa), high RH (>90%) and rainfall of 3 mm during the frontal passage of Case C (Figure 2b). In some cases, ASRC cloud water collections may occasionally include raindrops or drizzle in their samples. However, for lighter rainfall events, such as Cases C and D, any contamination can be ignored. The Froude number was estimated to be less/greater than 0.5 before/after 1600 UTC 17 March, implying that any orographic influence on cloud formation was enhanced during the last half of Case C. In contrast, the Froude number of Case D was estimated to be less than 0.3, indicating that the enhancement of cloud formation due to orographic lifting can be ignored.

3.2.2. PM and Lidar Measurements

[20] When a dust storm occurs, its transport accompanies southeastward moving stable continental high pressure (the so-called “Siberian High”), and the air mass accumulates anthropogenic surface pollutants during the process. Air quality drastically worsens due to increasingly high PM levels in the boundary layer in these scenarios. During EAREX 2005, PM10 measured at Gosan (33.29°N, 126.16°E; 0.072 km MSL) exceeded 350 μg m−3 at 0200 UTC 18 March [Choi et al., 2008]. On the same day, air quality stations in northern Taiwan also detected high levels of PM10. These data are shown in Figure 6. The time lag in detecting PM10 signatures, a result of air mass transport between Gosan and Taiwan, reflects the regional impact and distributions of a continental outflow regime.

Figure 6.

Time lag in PM10 peaks at three selected sites distributed from the northern tip to central Taiwan during the dust event (Case C).

[21] Previous studies from in situ measurements have reported that anthropogenic pollutants arrive earlier (∼12 h) than dust-laden pollutants either in separate air masses [e.g., Uematsu et al., 2002] or in a mixing air mass [e.g., Choi et al., 2008] during dust outflow events over East Asia. In our case, a distinct PM10 concentration peak related to anthropogenic pollutants was observed at WC and ST before the arrival of dust-laden pollutants, as shown in Figure 6. However, we did not see the first peak at Wan-Li site, because the air mass source format this time was oceanic, as suggested by back trajectory analyses shown in Figure 7a.

Figure 7.

Five day backward trajectories of Gosan (green), Wan-Li (red), and Mount Bamboo (blue) starting at (a) 0600 UTC 17 March, (b) 1200 UTC 17 March, (c) 1800 UTC 17 March, (d) 0000 UTC 18 March, (e) 0600 UTC 18 March, and (f) 1200 UTC 18 March.

[22] The dust-laden air mass advected from north through central Taiwan after the cold front passed. PM10 data presented in Figure 6 reveal the temporal and spatial evolution of dust particle transport from Wan-Li, in the north, to WC ∼50 km south, to ST station in central Taiwan, another ∼100 km farther south, with successive time lags. These observations are consistent with the description of Liu et al. [2006], where dust particles commonly arrive 12 h before the time of the peak of PM10 concentration and endure for 36 h. The potential for anthropogenic modification of a dust-laden air mass exists for this synoptic scenario. Maximum PM10 concentrations observed in northern Taiwan were in the range of 150–300 μg m–3. In addition, both trajectories coinciding with maximum PM10 measurements observed at Gosan (Figure 7b) and Wan-Li (Figure 7d) originated near Shanghai (30°N, 120°E) on 15–16 March. We speculate that the air mass may have become a mixture of preexisting dust and continental pollutants, which later sequentially reached Gosan and northern Taiwan.

[23] Lidar measurements depict the evolution of the aerosol vertical structure for Case C. Figure 5 shows the temporal distribution of the NRB parameter observed at NCU during 16–19 March 2005. This is a qualitative parameter. We interpret the nature of the particulate scatterers present in the profiles based on levels of signal transmission during the period. From roughly 1200 UTC 17 March to 1200 UTC 18 March, low cloudiness, associated with the frontal boundary passage, was apparent from the lidar composite image. We conclude that the layer near and above 1.5 km MSL is cloud since liquid hydrometeors rapidly attenuate the source laser pulse at a visible optical depth exceeding 3.0 [Sassen and Cho, 1992]. Maximum NRB values below cloud base were measured around 0000 UTC 18 March. A lack of strong signal attenuation in the subcloud base layer likely corresponds with aerosol loading. Though likely deliquesced given nearby water bodies, since relatively large aerosols scatter visible laser energy in proportion to their cross-sectional area, high concentrations of such particles result in relatively high yet ultimately transmissive NRB signals. This signal maximum and the overall temporal trend agree well with PM10 concentrations observed at WC (Figure 6). The profile also corresponds well with the monthly vertical profile of extinction cross section derived from lidar observations at Amami, Japan [Nakajima et al., 2007]. After 1200 UTC 18 March, cloud presence dissipated. The height of the boundary layer, which is depicted well by the strong gradient in aerosol scatter with height, decreased gradually with time as colder and more stable air moved over the island.

3.2.3. Cloud Chemistry

[24] A frontal boundary preceding a high-pressure system reached northern Taiwan around 1000 UTC 17 March and swiftly passed over the island over the course of the day. The first and last cloud water samples were collected at 1100 UTC on 17 March and at 0000 UTC 18 March, respectively. Nine samples were obtained during the frontal passage.

[25] Table 1 lists the mean pH, conductivity, and ion concentrations of cloud water samples for Case C. The average pH value (3.9) falls within a typical range of 3.6–4.3 for the winter season in this region [Lin and Peng, 1999]. The average cloud water conductivity (∼ 356 μΩ−1 cm−1) is high compared with similar measurements (177–244 μΩ−1 cm−1) reported from Mount Mitchell, North Carolina, USA [Saxena and Lin, 1990], (43 μΩ−1 cm−1) Mount Washington, New Hampshire [Houghton, 1995], and (28–300 μΩ−1 cm−1) in the European subcontinent [Acker et al., 1995]. Our measurement of relatively high conductivity is the result of large amounts of dissolved ions (averaged ion concentration of 5248 μeq L−1) existing in the stratus cloud system during the sampling period. The sea salt ions Cl and Na+ (1141 and 1163 μeq L−1, respectively) accounted for the highest contributions, 44%, to the total ion concentration, due to the cloud system having traversed the marine environment.

Table 1. Statistical Summary of Cloud Chemistry for the Cases C (C1 and C2) and D Observed at Mount Bambooa
CaseParameterNpHConductivityClNO3SO42−nss-SO42−bH+cNH4+Na+K+Mg2+Ca2+Σ Ion
  • a

    Conductivity in μΩ cm−1, and ion concentration in μeq L−1. N is the number of collected samples, μ is the mean value followed, and σ is the standard deviation.

  • b

    With [nss-SO42−] = [SO42−]-0.121*[Na+].

  • c

    [H+] is derived directly from field measured pH value.

Case C (dust event)μ93.8835611415697466052266091163633064255248
σ 0.6733211488038086651738471188823247065801
% of sum   2211141241222168 
Stage C1 (samples 1–7)μ73.6821866117438130225319765824165752588
σ 0.2921674016841532617425373629186882779
% of sum   267151210825163 
Stage C2 (samples 8–9)μ24.57838281719522022166813020522932201801165114558
σ 1.385516426559511801260121274104
% of sum   19131411114201611 
Case D (nondust event)μ134.1062818412111399836421216560
σ 0.33386077777655835721515344
% of sum   14152220181511023 

[26] Our measurements show a ratio of Cl/Na+ of 0.98 (compared to the seawater ratio of 1.16), which suggests a slight loss of Cl (∼4% of total ions) that may reflect either the leaching of MgCl2 from sea salt aerosols [Watanabe et al., 2006] or anthropogenic sources encountered during transport [Watanabe et al., 2001]. When the former mechanism dominates, the equivalent ratio of (Cl+lossMg2+)/Na+ must equal the ratio of Cl/Na+ in seawater. However, in our data, this ratio equaled 1.01, indicating only slight change. Therefore, we believe that anthropogenic influences were the dominant mechanism responsible. In this case, Cl loss was likely caused by the formation of HCl via the reaction of acidic gases (HNO3 and H2SO4) with sea salt, which then deposited to the sea surface. [Watanabe et al., 2001, and references therein].

[27] The acidic and alkaline species of non-sea-salt SO42− (nss-SO42−) and NH4+ (605 and 609 μeq L−1, respectively) are the most abundant anion and cation (12% of the total concentration for each), followed by NO3 (569 μeq L−1; 11% of the total concentration), corresponding to air samplings of fine mode particle for urban regions [Chuaybamroong et al., 2006; Li et al., 2009; Shen et al., 2009]. There is a linear relationship (r = 0.98) between nss-SO42− and NH4+ concentrations as well as the NH4+-NO3 pair (r = 0.99), which indicated that these species are associated in the likely form of NH4NO3, (NH4)2SO4 or NH4HSO4. Shen et al. [2009] and Li et al. [2009] indicated water-soluble ions (i.e., ammonium, sulfate and nitrate) were the major components of fine mode particles in Chinese urban regions, and evidenced that these three secondary ions exist mostly in the form of NH4HSO4 and NH4NO3. However, the linear regression of mole concentration of SO42− and NH4+ for 13 samples showed [SO42−] = 0.39 [NH4+] + 67.29, suggesting that NH4+ and nss-SO42− are present in the form of (NH4)2SO4 in cloud water.

[28] The concentration ratio of (nss-SO42−+NO3)/(H++NH4+) is 1.4. Consequently, other cations such as Ca2+, Mg2+, and K+ (425, 306, and 63 μeq L−1, respectively) are responsible for neutralizing excessive anions. For instance, Ca2+ is estimated to neutralize 8% of the cloud water acidity. After including Ca2+, the concentration ratio of (nss-SO42−+NO3)/(H++NH4++Ca2+) decreases to 0.93. We further consider the loss of Cl, 4%, whose ratio became 0.97 indicating that some undetected acids may need to be included to balance the excessive cations. These undetected anions are probably organic in nature [Lin and Saxena, 1991]. Overall, we suggest that the cloud system passing over the sea and arriving over Taiwan should have been appreciably processed by sea salt, and the large amount of pollutants dissolved in the cloud are likely to have their origin on the Asian continent.

[29] Mohnen and Vong [1993] suggest that clouds may contain high concentrations of acids and other dissolved ions, depending on their past trajectories. Based on the analysis of a 5 day backward trajectories at Mount Bamboo, northwesterly flow originated over a known dust source region (northeastern China) and gradually gained a northerly component over northern Taiwan during the period corresponding with the last two samples (Figures 7c and 7d). By contrast, as shown in Figure 7b, trajectories ending at Mount Bamboo show disordered routes and are mostly westerly for the first half of the event. This air mass would have traveled over dense industrial areas in China. We suggest that the cloud system may have been influenced by different sources before arriving in northern Taiwan. Consequently, two stages (C1 and C2) during the cloud chemical development (see Table 1 and Figure 8) can be characterized through the event, with the latter being related to dust transport.

Figure 8.

The (a) pH, conductivity, LWC, and (b) selected ions of cloud water samples and (c) meteorological data for Case C observed at Mount Bamboo during 17–18 March 2005.

[30] Figure 8a shows the time series of pH and the conductivity of cloud water from our samples. For the first 3 h beginning 1100 UTC 17 March, the estimated cloud liquid-water content (LWC) depicts higher values of 0.70–0.78 g m−3, when compared to the average of 0.39 g m−3 for the entire sampling period. This corresponds with increased pH (up to 4.06) and a lowering of conductivity due to dilution effects. Afterward, pH dropped dramatically to a minimum of 3.23, and conductivity increased significantly with the decrease in cloud LWC. During the dissipating stage of the cloud event (C2, 1700 UTC 17 March to 0000 UTC 18 March), the last two samples were collected within a 7 h time period, and pH increased sharply to 5.54, due to a lower LWC of 0.07–0.13 g m−3. Meanwhile, conductivity was significantly elevated after 1400 UTC 17 March and reached above 800 μΩ−1 cm−1 at the dissipating stage of C2.

[31] As shown in Figure 8b, concentrations of individual principal ions (i.e., NO3, nss-SO42−, Na+, and Ca2+) are well correlated with conductivity. At the dissipating stage of the cloud event, NO3, nss-SO42−, Na+, and Ca2+ reached their highest concentrations of 1764, 1704, 2974 and 1845 μeq L−1, respectively. Compared with the ratio of 0.04 in the seawater, the ratio of Ca2+/Na+ in this case was observed to be 0.62, suggesting a continental source of elevated Ca2+. Furthermore, analysis of cloud chemistry and an investigation of the surface-based back trajectories show that the mineral dusts originated from continental sources that were simultaneously transported along with anthropogenic aerosols. In general, mineral dusts represent the origin of Ca2+, whereas the source of nss-SO42− and NO3 is anthropogenic. The ratios of nss-SO42−/Ca2+ for stages C1 and C2 ranged between 1.6 and 6.8 and 0.9–1.2, respectively. Evidently, the airflow was contaminated by mineral dust during the C2 stage.

[32] The nss-SO42−/NO3 ratio for stages C1 and C2 ranged between 1.0 and 2.0 and 0.8–1.0, respectively. Sulfate presence is thought to be more related to the process of long-range transport because of its longer chemical lifetime. In contrast, nitrates are mainly referred to as a secondary pollutant resulting from urban photochemical reactions. Thus, a higher ratio of nss-SO42−/NO3 in cloud water for the C1 stage is thought to be due to processes involving long-range transport. However, a relatively lower ratio of nss-SO42−/NO3 (as well as nss-SO42−/Ca2+) was observed for the C2 stage. Since northeasterly and northerly winds (Figure 8c) prevailed at the site, local pollutants from the southern leeside of the island could not have been lifted upward to the mountain peak. Therefore it is implicit that the air masses originated from different source regions, or experienced different traveling times during both stages, particularly during stage C2, which explains why Ca2+ and NO3 were simultaneously enhanced during the period of the dust incursion. In the C2 stage, cloud water may have been acidified by pollution from industrial regions in northern China and urban regions along the east coast China. However, abundant Ca2+ contributed to neutralizing the acidic cloud water during the dust stage. The much higher loading of chemical species within the clouds causes an enrichment effect in cloud water and, consequently, ion concentration increased drastically at this stage.

3.2.4. Regional Impact of Case C in Downwind Areas

[33] In this study, coordinated cloud water chemistry, lidar and ground PM10 measurements combine to yield a thorough assessment of the impact and systematic structure of the air mass transport events described over northern Taiwan. However, variability in the range of pollutants observed at other stations in downwind areas show the regional extent of the Case C [Sawa et al., 2007; Takami et al., 2007]. Therefore, in this section, we reconsider the results from sections, along with data reported in other EAREX 2005 studies, to illustrate the regional impact of Case C in these downwind areas.

[34] Gas and aerosol concentrations were enhanced at Gosan and west Pacific stations in accordance with frontal passages during Case C [Lee et al., 2007; Sawa et al., 2007; Takami et al., 2007; Choi et al., 2008]. These simultaneous increases imply widespread impact on downwind areas during continental outflow. In general, nss-SO42− were the most abundant constituents among the sampled PM, and were well correlated with CO concentrations [Lee et al., 2007]. The surface CO distribution of Case C, simulated by Sawa et al. [2007], can be used for representing the transport process. They describe relatively weak winds over the coastal urban emission regions of China that enhanced accumulation of CO. Relatively high CO concentrations accumulated behind the cold front, which then merged into elongated belts of enriched air before spreading over the western North Pacific. Sawa et al. [2007] also show evidence that CO emissions from southeastern China influenced measurements at Gosan and Yonagunijima (close to northern Taiwan) in this case. Their simulation agrees with our descriptions in section 3.2.2 that the air masses accumulated abundant preexisting dust and continental pollutants in the coastal regions of China (near Shanghai) and later separately advected downwind toward Gosan and northern Taiwan.

[35] Regarding vertical distributions of Asian air pollutions, Bey et al. [2001] indicate that the highest CO concentrations in Asian outflow regimes are in the boundary layer (0.0–2.0 km) behind cold-frontal boundaries. According to their simulations, Sawa et al. [2007] show that the CO cross section over western North Pacific (20°N−35°N and 120°E−140°E) during Case C exhibited dual CO concentration peaks at around 1.5 km in the boundary layer and around 2.0 km above it. The higher plume reflected emissions from Southeast Asia similarly described by Wang et al. [2007]. In contrast, the plume in the boundary layer, sequestered below the frontal boundary-induced cloud ceiling, originated in southeastern China. However, the well-mixed CO concentrations measured within this layer in Taiwan showed differences with the vertical NRB profile collected with the NCU lidar. A high aerosol NRB signal is apparent in Figure 5 around 0.5 km MSL. The estimated residence time for SO2 is less than 50 h in this region in the boundary layer [Galloway, 1990], and the heterogeneous conversion rate of SO2 to sulfate is 2.0% h−1 [Takami et al., 2007]. In this case, air pollutants spent a relatively long time aloft and were well oxidized [Takami et al., 2007]. The oxidized particle composite augments a preexiting dust core with a coating of organics, sulfate and nitrates [Haywood and Boucher, 2000]. We speculate that the anthropogenic and dust particles below the boundary-induced cloud ceiling acted as CCN that were gradually entrained into the observed clouds. Therefore, we believe that the residual aerosol plume containing the mineral dust and anthropogenic aerosols mixture was present in the subcloud base layer near 0.5 km MSL from the NCU lidar measurements.

[36] A lack of precipitation likely aided in characterizing the evolution of the cloud over the course of its transport. Time variance analysis for LWC and ion chemistry of cloud water provided clues to tracing the air mass sources. Ahead of the cold front, air mass transport dominated by warm and relatively moist southerly flow led to high LWC measurements in the beginning of the C1 stage (1100–1300 UTC 17 March). After this phase of the event, our data show increased ion concentrations owing to the entrainment of aerosol into the accompanying clouds. In the C2 stage, the dust-containing cloud was unambiguously identified by high Ca2+ concentrations. The much higher aerosol and chemical loading injected into these clouds caused enrichment of the cloud water and, consequently, ion concentrations increased. In addition, we suggest that the extremely low LWC in C2 caused not only the dissipating stage of the cloud but also suppressed cloud droplet growth as a result of high CCN loading.

3.3. Impact of Regional Pollutants on Cloud Chemistry

[37] In addition to the dust event (Case C), we selected a nondust event (Case D: 23–25 March) to compare the impact on cloud water chemistry and air quality over northern Taiwan. For Case D, a frontal system initially passed northern Taiwan and was followed by continental outflow associated with a broad anticyclone moving southeastward over the region. Case D shows similar synoptic evolution to that of Case C. However, this case was distinct due to precipitation measured over the Taiwanese research sites. Sawa et al. [2007] describe an interesting difference between Cases C and D. Secondary cyclogenesis occurred over the Sea of Japan on 24 March after the first feature moved in a northeast track away from the Asian continent. Lee et al. [2007] indicate that other than Ca2+, no significant chemical enhancement was observed in the chemical composition of PM2.5 during Case D compared to Case C at Gosan.

[38] During Case D, thirteen cloud water samples were collected between 1200 UTC 22 March and 0700 UTC 23 March. Table 1 lists statistical analyses of cloud water chemistry in a manner consistent with that of Case C. The average pH of 4.10 for Case D represents a typical value generally observed for the winter season at Mount Bamboo [Lin and Peng, 1999]. Further compared to the higher average pH of 4.57 at C2 stage, it is suggested that no major dust-laden clouds were observed during Case D. The average conductivity and total ions of this nondust event (62 μΩ−1 cm−1 and 560 μeq L−1, respectively) are distinct when compared to those of the dust event (356 μΩ−1 cm−1 and 5248 μeq L−1, respectively). Dust enhancement resulted in average total ions to increase nearly tenfold compared to the nondust event. It is worth noting that Ca2+ contributed 3.0% of the total ions in the nondust event, compared to 8.0% in the dust event. Furthermore, by analyzing ion ratios, concentration ratios of (nss-SO42−+NO3)/(H++NH4++Ca2+) were equal to 1.0 in nondust cloud water, indicating that cloud water ions approached unity compared to those of the dust event (0.93). Ratios of nss-SO42−/Na+ were 0.6 and 1.8 for the dust and the nondust events, respectively. This indicates that sea salt aerosols were predominant as CCN during the dust event.

[39] We define the cloud loading, Li (in μg m−3), as the product of cloud LWC (in g m−3) and the mass concentration of ion species i (μg g−1, which is converted from the ion equivalent concentration (μeq L−1)), as

equation image

In Table 2 are comparisons of average cloud LWC, cloud loading of NO3, nss-SO42−, Na+ and Ca2+ for the Cases C (C1 and C2) and D. During Case C, cloud loading of Ca2+ at the C2 stage increased by about 670% when compared with that of the C1 stage. This shows a remarkable influence from the addition of mineral dusts. At the same time, an increase in cloud loading can also be seen for NO3, nss-SO42−, and Na+, at about 330%, 120%, and 80%, respectively. This emphasizes that urban-related pollutants were important during stage C2. For total ions, the cloud loading increment between stages C1 and C2 is about 29 μg m−3 (150%), compared with about 137 μg m−3 (415%) for the PM10 increment between the minimum (33 μg m−3) and maximum (170 μg m−3) values at Wan-Li station (see Figure 6). We suggest that this enhancement (the enrichment effect of cloud water) is due to the abundance of mineral and anthropogenic aerosols being enhanced by atmospheric processing [Zhang et al., 2006] and then injected into the cloud during Case C. The enrichment effect enhanced more than half of the cloud loading of total ions, when cloud loading of Ca2+ increases approximately 7 times.

Table 2. Cloud Liquid Water Content (LWC) and Cloud Loadings of NO3, nss-SO42−, Na+, and Ca2+, and Total Ions for the Cases C (Stages C1 and C2) and D
 LWC (g m−3)NO3 (μg m−3)nss-SO42− (μg m−3)Na+ (μg m−3)Ca2+ (μg m−3)Total Ions (μg m−3)
Case C0.267.365.665.181.6532.31
Stage C10.393.013.713.840.4219.58
Stage C20.1012.948.166.893.2348.69
Case D0.311.451.540.370.094.78

[40] For the nondust event (Case D), cloud loading of Ca2+ is only about 5% of that for dust event (Case C), while it is roughly 20–25% for NO3 and nss-SO42−. It is worth noting that cloud loading of sea salt ions for Case D is much lower than that for Case C (only 7%). This indicates less of an entrainment influence of the marine updraft on cloud water chemistry associated with continental outflow in Case D. The cloud loading of total ions (4.78 μg m−3) for Case D is only about 15% of that for Case C. Even though we compare the nondust influence period of Stage C1 with Case D, the cloud loading of total ions for Case D is 24% of that for Stage C1. We speculate that the lower cloud loading of Case D is due to enhanced in-cloud scavenging and precipitation in the cloud system before arriving at northern Taiwan. On the other hand, the average PM10 at Wan-Li station reached as low as roughly 15 μg m−3 (compared to about 50 μg m−3 of Case C), which reveals strong subcloud scavenging mechanisms near the surface that removed the aerosol particles. As a result, we have evaluated the impact of regional pollutants on local cloud chemistry using the cloud loading of total ions. Cloud loading of the nondust even shows a large range of 4.78–19.58 μg m−3, which reflects the efficiency of scavenging mechanisms. However, in order to reinforce these findings, further case studies of cloud chemistry in Asian continental dust and pollutant-enriched air masses are necessary.

4. Summary and Conclusions

[41] We identify five cases of continental outflow regimes reaching northern Taiwan during ABC/EAREX 2005 in March 2005. One was coupled with a dust storm observed not only at Gosan, Korea, but also over Taiwan suggesting the wide scope of its regional impact. We compare air quality and cloud water chemistry measurements at stations located at northern Taiwan for contrasting dust and nondust events. In particular, cloud water chemistry obtained at the elevated Mount Bamboo site is described to investigate the impact of Asian dust on the characteristics of spring frontal clouds.

[42] A dust episode was characterized using lidar, cloud water and surface particulate matter (PM) measurements. PM observations at Gosan, Korea and Taiwan reveal that the dust storm affected a large area of East Asia. The regional impact of the event is illustrated by integrating our measurements and analysis with previous EAREX 2005 studies. We suggest that the air mass may have originally contained dust, but was modified by the ingesting of continental pollutants while passing over the coastal regions of China. The air mass later reached Gosan and northern Taiwan. The anthropogenic and dust particle mixture influenced bulk CCN characteristics and was entrained into frontal boundary clouds. The much higher aerosol and chemical loading of the resulting clouds caused an enrichment effect and, consequently, ion concentrations increased drastically. Extremely low LWC measurements during the dust stage influenced not only the dissipating stage of the cloud, but also suppressed cloud droplet sizes as a result of relatively high CCN loading. Again, these data suggest a strong correlation between continental outflow regimes, cloud chemistry and air trajectories.

[43] The impact of the continental outflow regimes on cloud water chemistry observed in northern Taiwan is described for dust and nondust events. In this study, we evaluate the impact of regional pollutants on cloud chemistry using cloud loading of total ions. As a result, the enrichment effect observed during the dust event can double cloud loading of total ions, when the cloud loading of Ca2+ increases approximately seven times. During the nondust event, the cloud loading of total ions shows a large range of 4.78–19.58 μg m−3, which varies as a function of aerosol scavenging efficiencies.


[44] This work was supported by the National Science Council of Taiwan under grants NSC92-2111-M-008-018-AGC, NSC93-2111-M-008-016-AGC, NSC94-2111-M-008-018-AGC, and NSC94-2752-M-008-006-PAE and by the Taiwan Environmental Protection Administration under contracts EPA93-U1L1-02-101 and EPA94-U1L1-02-101, and also by Hal Maring, NASA Radiation Science Program. We thank Chi-Wen Liou and Wei-Jing Jian for their help on data processing. The authors gratefully acknowledge the NOAA Air Resources Laboratory (ARL) for the provision of the HYSPLIT transport and dispersion model and/or READY website ( used in this publication. Author J.C. acknowledges the support of J.S. Reid at Naval Research Laboratory, Monterey, California. The NASA Micro-Pulse Lidar Network is funded by the NASA Earth Observing System and Radiation Sciences Program.