An Orchestrated Attempt to Determine the Chemical Properties of Asian Dust Particles by PIXE and XRF Techniques

2. 1 Particle Collection

In April 2001, several big scale dust storms developed over the Gobi desert in northern China. Both satellite remote sensing data and analyses of meteorological conditions demonstrated its long-range transport from China to the United States via Korean Peninsula over a period of several weeks (Szykman et al., 2003).

Particle collections were carried out at Gosan located at the western tip of Jeju Island, Korea (Fig. 1). The surface-based sampling site at Gosan, which had been in operation as one of the super sites during ACE-Asia, is about 70 m above sea level and 100 m southwest from Jeju upper air meteorological station. This site has been investigated intensively as one of the most optimal locations for AD research in Korea as well as throughout the world. Its geographical location as the outbound transport pathways of AD took have the advantage of offering useful information on the AD occurrence, as described in elsewhere (Kang et al., 2007; Park et al., 2004).

Fig. 1.

Maps showing a particle sampling site (a filled circle: 33.17°N and 126.10°E) along with three desert sand collection points in China.

The size discrimination of ambient particles has been generally recognized as fine and coarse particles. In general, fine particle (PM_2.5) is defined as airborne particles which are smaller than coarse particle (PM_10-2.5). This size classification of ambient particles is basically originated from a link between PM_2.5 and mortality (in more detail, the relationship between human respiratory system and particle size (e.g., the thoracic particulate matter (PM₁₀) and penetrating particle (PM_2.5))) and their formation processes.

Meanwhile, in the study of AD particles, the results of previous field studies (Hwang et al., 2008; Ma et al., 2005) showed that particles larger than 1 μm were remarkably increased during AD event. In addition, the particle mass concentration on AD event showing the bimodal distribution was enriched in relatively large particles (D_p>1.6 μm) (Ma and Choi, 2007). For these reasons, in this study, a two-stage filter pack sampler (Tokyo Dylec Co.) which can collect particles in both large (>1.2 μm) and small size fractions (<1.2 μm) was operated to collect size-resolved AD particles on April 12, 2001, when the outbreak of a thick yellowish AD was recorded by the ground-based in-situ monitoring system (and aircraft measurements). In addition, as an ancillary sample of AD, the collection of non-AD samples was executed in the same way (April 30, 2001).

This sampler collects the large (>1.2 μm) and small size (<1.2 μm) fractions of aerosols separately on a prefilter (a 47 mm diameter, 8 μm rated pore size Nuclepore ^® filter with 1×10⁵ cm^-2 pore density and 1 mg cm^-2 nominal wt.) and a back-up filter (a 47 mm diameter, 0.4 μm rated pore size Nuclepore^® filter with 1×10⁸ cm^-2 pore density and 1 mg cm^-2 nominal wt.), respectively. The 50% cut-off diameter of the prefilter with 25 L min^-1 flow rate was estimated to be 1.2 μm equivalent aerodynamic diameter (Ma et al., 2001a; Kasahara et al., 1996). The mechanisms by which particles are arrested on the Nuclepore^® filter are Brownian diffusion, interception, and inertial impaction. Efficiencies for the separation processes relate to the governing variables such as particle size, aerosol approach velocity, and filter pore size (Smith et al., 1976).

The duration of sampling was adjusted to 20 minutes (April 12, 08:40 to 09:00 (UTC)) and 3 hours (April 30, 06:00 to 09:00 (UTC)) in AD event and non-AD days, respectively.

After sealing Petri dish (50×9 mm) containing filter with Teflon tape and wrapping with aluminum foil, every sample was placed in a cold storage bag during air transportation.

During the first sampling time, the wind speed was recorded in the range of 15.2-20.5 m s^-1 with the dominant westerlies. Temperature and relative humidity in the sampling period averaged 13.6°C and 56%, respectively.

Meanwhile, on a non-AD day (April 30, 2001) wind speed was comparatively weak (5.7 m s^-1) and it was blown form the south. The mean temperature and relative humidity were around 15.7°C and 73%, respectively.

2. 2 Elemental Analysis

3. 1 Simulated Trajectories of AD

The movements of air parcel in two field events were simulated through backward trajectories starting at Gosan, Jeju based on the Meteorological Data Explorer (METEX) (Center for Global Environmental Research, 2010). Three-day backward trajectories (origin 33.17°N; 126.10°E) were computed at three different altitude levels of 500, 1,000, and 3,000 m. Fig. 2 shows the result of backward trajectories calculated at those three altitude levels in two in-situ measurements. The analysis of air parcel movement pathways suggests that air parcels were transported to the Gosan site through the desert areas on April 12, 2001 (Fig. 2). On the other hand, on April 30, 2001 the air parcel transport pathways stemmed from southwest and southern directions.

Fig. 2.

Backward trajectories started at 33.17°N and 126.10°E. Starting times were at 09 UTC, Apr. 12 (top three) and Apr. 30 (bottom three) in 2001. The start heights of the air parcel were 500, 1,000, and 3,000 m.

3. 2 Transport and Spatial Distribution of PM during AD Event

To account for the origin of enhanced PM levels encountered at the Gosan sampling location on April 12, 2001, the National Oceanic Atmospheric Administration (NOAA) HYbrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) dispersion-trajectory model “backwards” (http://www.arl.noaa.gov) was applied. Fig. 3 displays the simulated backward aerosol dispersion started at Gosan receptor site for 72 hours. In Fig. 3, the area scales indicate the integrated aerosol concentration between in altitudes 100 and 1,000 m. As shown in Fig. 3, the aerosol dispersion on April 12, 2001 initially followed a southeastern route toward central and eastern China which then subsequently turned toward the western coast of the Korean Peninsula. Thus, the high PM value at the Gosan receptor site was undoubtedly affected by long-range transport of the particulate matter driven from the desert areas in China such as Gobi.

Fig. 3.

Simulated backward aerosol dispersion by the NOAA HYSPLIT model. Aerosol concentration (kg m-3) averaged between 100 and 1,000 m. Release start (YY MM DD HH): 01 04 12 09 (UTC) (backward), Duration: 72 hrs.

3. 3 Characteristics of Bulk Particles during AD

A list of particle mass, elemental concentrations, and their ratios (e.g., (large size PM/small size PM) and (b of AD/b of n-AD)) of bulk particles is made for two different cases: (1) the intensive dust storm event (i.e., an unusually huge dust storm developed on April 12, 2001) and (2) a non-dust storm day (April 30, 2001) (Table 1). Unfortunately, as the analytical sensitivity of our PIXE is relatively low for Al, Al data were excluded from further evaluation in the present study.

Table 1.

Mass and elemental concentration (ng m-3) of particles collected during Asian dust (AD) storm event (Apr. 12, 2001) and a non-Asian dust (n-AD) storm day (Apr. 30, 2001).

The large size particle fraction in the AD event, as expected, showed mass concentrations that are roughly 14 times higher than those measured on a non-AD day. In particular, the mass concentration of small size particles in company with large size particles was also found to be significantly high in AD relative to the non-AD period. The definite increase in small size particle masses along with AD was previously mentioned (Ma et al., 2001b). In addition, the small size particle mass in the AD event was elevated (e.g., approximately times relative to that of urban area). Hence, it confirms the potential role of AD in determining the budget of small size particles.

The mass concentration of the most representative soil components (Si, K, Ca, and Fe) in the large particle mode on AD day marked 37.2 μg m^-3. These soil-originated elements accounted for 17.4% of total particle mass. In contrast, those of the non-AD day showed 0.4 μg m^-3 (2.5% of total particle mass).

The ratio of elemental mass concentration (b/a) suggests that mineral origin components (e. g., Si, Ca, and Fe) were mainly distributed in large size particles during AD event. Among them, the ratios (b/a) of S and Fe during AD are higher about 2.7 and 4.6 times than those of non-AD period, respectively. The ratio (b/a) of Cl showing absolutely high level (704.5) was probably caused by the sea-salt from the bursting of sea water bubbles by strong wind speed (15.2-20.5m s^-1) during AD event. As shown in the last column of Table 1, the mass concentrations of Si, Ca, and Fe in large size particles collected during AD event show overwhelmingly high levels (Si; 102.9, Ca; 83.9, and Fe; 199.5) compared to those of non-AD particles. Furthermore, the ratios of Mn and V [(b in AD)/(b in n-AD)] suggest that the episodically strong AD storm carried not only dust, but also anthropogenic pollutants to our sampling site.

As mentioned earlier, an air parcel was seen to pass over central China (i.e., our sand collection points) and then it turned toward the east. Through the characterization of the factors affecting the elemental properties between the desert sand and AD particles at a receptor area, one may collect evidence to account for the source profile and chemical aging of AD particles. Fig. 4 shows the crustal enrichment factors [(Z/Si)_particle/(Z/Si)_{desert sand}] of elements in large size particles (>1.2 μm) collected on the AD day at a ground-based site on the west coast of Jeju Island, Korea. The elemental characteristics of the local desert sand had already been referred to by Ma et al. (2008). Although Al is the most representative soil component, Si was used as reference for enrichment factor computation because, as mentioned already, Al could not be reasonably analyzed by our PIXE technique.

Fig. 4.

Crustal enrichment factors of elements in large size particles (>1.2 μm) collected in AD event at ground-based site on the west coast of Jeju Island. Crustal enrichment factors were calculated by the elemental components in three different desert sands ((a): Yinchuan vs. Dulan, (b): Yinchuan vs. Yanchi).

Although there is no large difference in coefficient of determination (R²) between the two plots, the plot of [(Z/Si)_particle/(Z/Si)_{Yinchuan sand}] vs. [(Z/Si)_particle/(Z/Si)_{Yanchi sand}] (R²~0.98) exhibited slightly enhanced correlations relative to [(Z/Si)_particle/(Z/Si)_{Yinchuan sand}] vs. [(Z/Si)_particle/(Z/Si)_{Dulan sand}] (R²~0.90).

According to the result of backward trajectories (Fig. 2), an air parcel was seen to travel in a south-easterly direction close to the Yinchuan and Yanch deserts. This must be the reason why the R square levels are somewhat different from each other. This result also indicates that the chemical nature of the sands distributed at each desert area should be dissimilar to each other. Moreover, as shown in Fig. 5, calculation of crustal enrichment factors allows us to fractionate soil-origin components from all detected elements.

Fig. 5.

Ternary plot of the relative mass ratios of S, Cl, and Ca in the dust particles collected at ground-based site on the west coast of Jeju Island, Korea (Total particle number of 250).

3. 4 Characteristics of Individual Particles

Fig. 5 displays the ternary plot of the relative mass ratios of the three major elements S, Cl, and Ca for the entire 250 individual AD particles. From the ratios of atomic weight for S and the weight % of Cl, S and Ca in ocean, individual AD particles can be classified into four distinct groups. The Type 1-group, namely, the CaSO₄ particle formed from the neutralization of H₂SO₄ by CaCO₃ (or CaSO₄ as natural mineral) accounted for a large portion of the total particles. One of minor groups of particles is Type-2, in which SO₂ was excessively attached to the particles along with the neutralized CaSO₄. In the case of the Type-3, the particles did not go through sufficient neutralization processes.

Fig. 6 shows the particle-to-particle variation of the elemental mass in individual large particles (>1.2 μm). The length of box corresponds to the level of dispersion for elemental mass among individual particles. Almost all elements underwent a substantial change of mass among individual particles. However, there are no a noticeable difference in box length scales between the elements (with the exception of Ni and Cr). It should be noted that this unique property of elemental mass distribution among particles cannot be evaluated from a common bulk sample analysis. The Cu mass concentration, forming a box in a compact mass, probably did not origin from particles but should have been that of the sample holder of XRF system.

Fig. 6.

Box plots of each elemental mass of individual large size particles (>1.2 μm) collected at ground-based site on the west coast of Jeju Island, Korea.

If one assumes that S is derived mainly wholly from CaSO₄, the mass of CaSO₄ can be theoretically estimated by the S mass times the ratio of CaSO_{4molecular weight}/S_{atomic weight}. In the present study, the average S mass is 21.9 ng. CaSO₄ mass is about 93.4 ng on the complete supposition of S with CaSO₄. If one hypothesizes that Ca in minerals generally exists in the form of both CaSO₄ and CaCO₃, the mass of Ca from CaCO₃ can be estimated from the difference in Ca mass between particles (and that calculated from CaSO₄ mass). The average mass of Ca (see Fig. 6) in total particles is 30.04 ng, while the Ca mass in CaSO₄ is 27.5 ng (93.4*Ca _{atomic weight}/CaSO_{4molecular weight}). The mass of Ca stemming from CaCO₃ can be consequently calculated as 2.9 ng (It presumably corresponds to the Type-3 in Fig. 5).

Although single particle analysis has a great many merits, it has also a significant drawback. Because only a certain portion of particles is subject to the analysis, it cannot estimate all the properties of the entire particle. Hence, it should be examined how well the result of single particle analysis can be used to represent the pattern of the whole particle. Fig. 7 shows the plots of the elemental occupation rates (%) in bulk and single particle samples. The results do not show a perfect agreement between bulk and single particles (R²~0.64). However, the entire elements (except Si) tend to maintain relative good correlations (R²~0.81). This result thus suggests that a large number of individual particles are required be analyzed to overcome the discordance with a bulk sample.

Fig. 7.

Plots of the elemental occupation rates (%) in bulk sample and single particles. Total common elements analyzed by both bulk and single particle analyses (left) and the same after excluding of Si (right).