The Taiwan air quality model incorporating a dust module was applied to calculate masses of supermicron (diameter greater than 1 μm) and submicron particles (diameter less than 1 μm) and their dust fractions during the ACE-Asia airborne experiments over the northwestern Pacific. The results showed that the calculated vertical profiles of supermicron particle concentrations matched reasonably well with the observations obtained from 19 research aircraft missions. During dust storm events and at dust concentrated altitudes, the calculated dust fractions in the supermicron particles were usually greater than 90%, and the dust was concentrated in the lower troposphere mainly below 6 km. Without dust storm, dust was still the major component of the supermicron particles above boundary layer. In contrast to supermicron particles, the model results showed that the major component of the submicron particles observed during aircraft experiments was mostly from pollution. The calculated vertical profiles of submicron particle concentrations were sensitive to the emission inventory of air pollutants over east Asia. The correlation between observed anthropogenic volatile organic compound and submicron particles were used to identify the pollution fraction in the submicron particles, and the results were consistent with the model calculations of dust fraction. The model results showed that the dust fractions in the submicron particles were usually less than 28% in the boundary layer. During dust storm events the dust fractions were usually greater than 40% but can be as low as 24% when significant amount of pollutants were present.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 The vertical distribution of dust can extend several km up in the free troposphere. For example, lidar and satellite observations over Japan indicated that the dust cloud originated from 2 different sources, consisting of 2 different layers, one at 6 km and the other at 2 km [Iwasaka et al., 1983]. In addition, Uno et al.  simulated dust distributions and showed that the dust layer was located primarily below 3 km during long-range transport across Pacific Ocean due to strong subsidence at the backside of a cutoff vortex. Uno et al.  also showed that most of the dust observed in Beijing was below 6 km during the ACE-Asia period. From airborne measurements obtained during ACE-Asia, Anderson et al.  showed that the distribution of coarse-mode dust and fine-mode pollution were different. Dust was found throughout the lower troposphere, indicating that they were either originated at high altitude or was uplifted during transport; whereas pollution was found to concentrate in the boundary layer.
 Dust events can significantly affect not only particle concentrations but also size distribution over downwind areas. Through lidar observations, Zhou et al.  summarized that dust contributed mostly to particles with radius of 1–3 μm (μm = 10−6 m). Chun et al.  also observed increases in particles with diameters in the range of 1.35–10 μm during a dust storm event in Korea. Cahill et al.  observed that there were significant amount of fine particles presented during the dust period, and as the dust passed, the particles in the range of 2.5 to 5 μm were observed. Aerosol mass concentration including crustal elements and nitrate were observed to increase over the downwind areas during dust events [Duce et al., 1980; Uematsu et al., 1983; Prospero and Savoie, 1989].
 While dust contribution on coarse particles is better recognized, its contribution on fine particles is not well understood. Most of the previous aerosol observations paid attention to dust effects on total particle or particle less than 10 μm. In addition, a significant component of fine particles come from pollution, including direct emissions and gas to particle conversion [Seinfeld and Pandis, 1997]. The emission inventory of sulfur dioxide has been compiled over the east Asia in earlier studies [Kato and Akimoto, 1992; Bai, 1995; Benkovitz et al., 1996; Streets et al., 2000]. However, the emission inventories of nitrogen oxides and especially volatile organic compound (VOCs) over east Asia were highly uncertain. The emission inventory of VOCs over east Asia has been developed recently [Streets et al., 2003], but the uncertainty remains to be quantified. Therefore, due to lack of observations and large uncertainty in model calculation of fine particle mass, dust contribution on fine particles is difficult to quantify.
 The Aerosol Characterization Experiments (ACE) were designed to better understand the influences of atmospheric aerosol particles on the global climate system. The intensive field experiments of ACE-Asia were conducted during the spring of 2001 off the coast of China, Japan and Korea. One of the major objectives of the experiments was to quantify the chemical, physical and radiative properties of aerosols in the ACE-Asia study area and assess the spatial and temporal variability of these properties. An important issue involved was Asian dust outflow. Measurements from several types of experiments on ships, aircraft, and ground-based stations contributed to the regional characterization of Asian dust properties. Aircraft experiments on board of the NCAR C-130 aircraft were conducted during the period, and the measurements provided a set of extremely important data for understanding the characteristics of the dust as well as pollutants in the western Pacific and east Asia.
 To quantify the impacts of dust on coarse and fine particles during ACE-Asia aircraft experiment, this study applied a regional model to calculate the vertical distributions of supermicron particles, submicron particles and their dust fractions; and compared the model results with the observations. The aircraft measurements were conducted over the northwestern Pacific during ACE-Asia. The area was downwind of Asian dust outflow and the measurements were taken over a period of more than one month when a number of intensive dust storms occurred.
 Since pollution is a significant component of submicron particles, the model calculation of submicron particle mass and its dust fraction during ACE-Asia aircraft experiments is sensitive to the emissions inventory of air pollutants. In order to understand the pollution fraction of the submicron particles, this study applied the VOCs observed during the ACE-Asia aircraft experiments. A large number of VOCs are emitted primarily by anthropogenic activities, and are good indicators of air pollution. Strong correlation of the concentrations between submicron particle and anthropogenic VOCs suggests that the two species originate from similar anthropogenic sources. We realize that the positive correlation can't be a proof of a common source because it can result from other causes such as dilution effect or photochemical reaction. Thus we will use the correlation between VOCs and submicron particles as an indicator of a significant pollution fraction in the submicron particles.
2. Model Descriptions
 The regional chemistry and aerosol model used in this study is based on the Taiwan air quality model (TAQM) [Jeng et al., 2000], which was previously modified from the Regional Acid Deposition Model (RADM) [Chang et al., 1987; Chang, 1990]. With further modification, the current version of TAQM, version 2, was developed mainly based on the San Joaquin Valley Air Quality Model (SAQM) [Chang et al., 1997] developed for the California Air Resources Board. As an extension and adaptation of the existing SAQM, TAQM allows the use of nonhydrostatic meteorological data, and can be applied to simulate the time-varying three-dimensional distributions of trace gases and particles over Asia. It includes quantitative description of emissions, advection, diffusion, gas phase chemistry, aerosol transformation, cloud processes, and dry and wet depositions. The domain of the model can cover the whole east Asia, and can be adjusted to cover smaller area with a nesting technique. Vertically, the model divides the troposphere from surface to 100 hPa into 15 layers using variable-spaced sigma coordinates. The surface layer of the model can be further subdivided into 3 sublayers to improve the prediction ability on surface. The resolution for this 15-layer model is adjustable depending on the altitude of the interest.
 A dust module was incorporated into the current version of TAQM in order to assess the dust impacts during the ACE-Asia aircraft experiments. Emissions, transport, as well as dry and wet depositions of dust particles were included in the modules. To simplify the interpretation of the results, no chemical or physical transformation of dust particles was included.
 Gas phase emissions including anthropogenic and natural sources for 2001 were obtained from several emission inventories (M. Zhang, personal communication, 2002). The anthropogenic gas phase emissions of NOx (NO + NO2), SO2, CO and VOCs over Asia were based on the emission inventory of Streets et al.  and Olivier et al. . Natural emissions of NOx and VOCs from soil or biology were taken from the Global Emission Inventory Activity (GEIA) monthly global inventory [Benkovitz et al., 1996] for the month of March. SO2 emissions from volcanoes estimated by Streets et al.  were also included. It was assumed that 5% of these SO2 emissions were in the form of H2SO4.
 Anthropogenic emissions of particulate matter less than 10 μm (PM10) were compiled in the emission inventory over Taiwan (Taiwan EPA), but not over the rest of the Asia. In this study, the PM10 emissions over the rest of the Asia were derived from NOx emissions according to their emission ratio over Taiwan. Since the emitted particle size was not available from the emission inventory, it is assumed that 50% of these PM10 emissions were from particles less than 2.5 μm (PM2.5). The PM2.5 contained 30% of particles less than 1 μm, including 1% in Aitken (nucleation) mode and 99% in accumulation mode, and the rest were coarse particles. The emissions of black carbon and organic carbon were also included in the model and were derived from the emissions of particular matter according to their ratio over the United States.
 Dust emissions based on Wang et al.  were included in this study. According to Wang et al. , the onset of the dust emission was controlled by three factors: frictional velocity, surface humidity, and predominant weather conditions. When all the three factors satisfied the prescribed thresholds, the emission intensity were calculated as followed,
where dust emission intensity S was in kg m−2 s−1, C1 was the weighting factor of different surface conditions, C2 was an empirical factor and was determined to be 2.9 × 10−11, u* was the surface frictional velocity, RH was the surface relative humidity, and W was the mass weighting factor for each aerosol bin, derived from the observed bin distribution over the sources areas. The dust particles were divided into 12 size bins ranging from 0.13 to 20.13 μm. In this study, all the dust was emitted into the lowest model level, which is different from Wang et al. , where dust emission is distributed into the boundary layer, 1 to 3 km and 3 to 5 km by 60%, 30%, and 10% respectively. Since a fast vertical transport scheme is applied in the current version TAQM, there is no obvious difference between the results obtained from different vertical distribution of dust emissions. In addition, the emission intensity of the submicron particles adopted from Wang et al.  resulted in densities of submicron particles compared to the previous observations in Taiwan, and was reduced to half of the original values in this study.
 The particles emissions from sea salt aerosol and biomass burning were also included in this study. An empirical formula derived by Gong et al.  was applied to calculate sea salt aerosol emissions. The biomass burning emission inventory estimated using fire count obtained from Advanced Very High Resolution Radiometer (AVHRR) satellite images and Total Ozone Mapping Spectrometer–Aerosol Index (TOMS-AI) were obtained from Streets et al.  for 2001.
 The emission resolutions were 1° × 1° for all the sources, but the emission heights were different. The trace gases and aerosol that were emitted from area sources were incorporated into the lowest model level. Point sources may be emitted into higher model layers depending on the plumerise calculations. In addition to dust, biomass burning and sea salt emissions were also emitted into the lowest model layers.
2.2. Advection and Diffusion
 The advection part of the atmospheric transport is computed using a fourth-order positive definite and mass conservative scheme derived by Bott . The finite difference advection scheme is applied to minimize the problems of numerical diffusion. Vertically, a new second-order scheme with nonuniform grid sizes derived from the Bott's scheme has been applied. The diffusion equation in flux conservative form is solved by applying second-order Crank-Nicolson implicit scheme, taking into account the nonuniform grid size in the vertical direction. In addition to vertical advection and diffusion, the Asymmetrical Convective Model (ACM) obtained from Pleim and Chang  was adopted in the TAQM. In this vertical mixing scheme, emitted species were rapidly transported upward to the free troposphere under convective conditions.
2.3. Gas Phase Chemistry
 On gas phase chemistry, the Regional Acid Deposition Model version 2 (RADM2) mechanisms [Stockwell et al., 1990], which containing 158 reactions among 63 organic and inorganic species, were applied in this study. The RADM2 mechanisms consist of 42 organics, including 26 stable species and 16 peroxy radicals, representing organic chemistry through a reactivity lumped molecular approach. 21 photolysis rate coefficients are derived from a radiative transfer model and are a function of time, latitude and altitude. An efficient chemical solver based on the new Euler Backward Iterative [Hertel et al., 1993] method is used to reduce the computational time while maintaining good accuracy.
2.4. Aerosol Transformations
 Physical transformations of aerosol particles formulated in Binkowski and Shankar's  Regional Particulate Model (RPM) were included in the model. The transformation processes consisting of nucleation, condensation, and coagulation processes were considered for atmospheric aerosols, including Aitken mode (nucleation mode), accumulation mode, and coarse mode. For each mode, a lognormal distribution equation was applied to specify the size distribution. Aqueous chemistry on the aerosol surface was considered based on the thermodynamic equilibrium between sulfate, ammonium, nitrate, and water. Precursors of aqueous phase species such as H2SO4, HNO3, and NH3, were calculated self-consistently according to the gas phase chemistry in the model. In addition to inorganic species, organic carbon fraction of aerosols produced photochemically from xylene, terpene, cresol, olifin, toluene and higher VOCs were also considered in the model.
2.5. Cloud Processes
 A one-dimensional diagnostic cloud module is used to calculate subgrid-scale convective redistribution and wet removal of chemicals. The module assumes that convective clouds lift air parcels from below cloud base into convectively unstable layers, and compensating subsidence pushes cloud free air down throughout unstable layers. Aqueous phase SO2 reactions and wet removal of chemical species are computed using a box aqueous chemical module. Rain at the surface is assumed to be the sum of water from each level of clouds, thus trace species concentration in the precipitation is equal to the vertically integrated mean cloud water composition. The only wet removal process considered for clouds is subcloud precipitation. In computing cloud water composition, ions balance is continuously maintained between positive and negative ions, and concentrations of the ions were determined from Henry's law and dissociation equilibrium [Walcek and Taylor, 1986].
2.6. Dry Deposition
 The removal of trace species by dry deposition is calculated through dry deposition velocity, which is derived from the aerodynamic, sublayer and canopy resistance. The dry deposition fluxes are calculated from deposition velocities and species concentrations in the model surface layer. In addition to the deposition in the surface layer, gravitational deposition for aerosol particles above the surface layer is also considered.
2.7. Model Setup and Meteorological Input
 The model domain (Figure 1) covered east Asia from about 5°N to 55°N, and 70°E to 155°E, with horizontal resolution of 81 km × 81 km. Vertically, the model was divided into 15 layers, with approximately 40 m near the surface and 1–2 km near the tropopause. For trace gases and aerosols with longer lifetime than a few days, the inflow from the boundary can affect the results within the model domain depending on the timescale of horizontal transport [Liu et al., 2001]. At altitudes of 2 and 14 km, it took about 12 and 5 days, respectively, for the westerly winds to transport trace species across the model domain. Below 2 km the wind speed was smaller, and the influence of boundary conditions on the results was reduced. To avoid influences of the boundary conditions on the calculated results, the boundaries of the domain were chosen over the remote areas. Both clean boundary conditions were assumed for trace gases and aerosols other than dust. Most of the Asian dust was assumed to occur within the model domain, because the inflow of the dust from outside the model domain is small and can be neglected [Wang et al., 2000].
 During dust events, dust particles can have a residence time of more than a week in the model domain. Therefore a longer initial run may not be necessary for chemistry simulation, but is important for dust simulation in this study. A 12-day of spin-up run of the model, with both chemistry and aerosol mechanisms included was performed before the beginning of the formal simulation in order to minimize the influence of the initial conditions on the results.
 The numerical simulation was performed from 31 March to 4 May 2001, when 19 research aircraft experiments were conducted. Hourly meteorological data used to quantify transport and diffusion of trace species were calculated from the NCAR/Penn State Mesoscale Meteorological Model (MM5) version 3 using four-dimensional data assimilation. Initial and boundary conditions for MM5 were obtained from outputs of a global model, namely, the National Centers for Environmental Prediction (NCEP) Aviation model (AVN), with a spatial resolution of 1° × 1° at 6-hour intervals.
 In comparing the model results with airborne measurements, the model calculation of the area surrounded by each flight path were averaged for the entire flight period, excluding only the first and last 1 or 2 hours before or after the aircraft arrived or left the interested area. Table 1 lists the flight number, flight time, and the model domain selected for comparing model results with observations. The spatial area surrounded by the flight path 13 was large, and the averaged model results from the area may not representative of the characteristics measured along the entire flight path. Therefore, instead of one large model domain, two smaller but separate model domains covering most of the flight paths were chosen in order to average the results. The results from these two small model domains were averaged again before the comparison with observations.
Table 1. Summary of the 19 Research Missions of the ACE-Asia C-130 Aircraft Measurements
The upper and lower rows of each mission represents the values above and below 2 km, respectively.
03/30 22:46–03/31 06:34
04/01 23:35–04/02 08:13
04/03 23:13–04/04 09:00
04/05 23:34–04/06 09:48
04/08 03:04–04/08 10:39
04/10 23:15–04/11 07:21
04/11 23:42–04/12 09:14
04/12 23:38–04/13 08:47
04/16 23:39–04/17 09:28
04/17 23:36–04/18 09:03
04/19 23:33–04/20 09:20
39°N–42°N 139°E–144°E/31°N–35°N 136°E–141°E
04/22 22:45–04/23 08:07
04/23 23:30–04/24 09:13
04/24 23:26–04/25 09:19
04/27 00:03–04/27 10:11
04/29 23:31–04/30 08:45
04/30 23:23–05/01 09:21
05/02 00:14–05/02 09:17
05/03 22:56–05/04 07:37
3. Aircraft Samples of Aerosol Particles and VOCs
 Aerosol measurements used to compare with our model results in this study was acquired on board the NCAR C-130. The research aircraft NCAR C-130 was conducted from 30 March to 4 May during the 2001 ACE-Asia intensive field operations. The airfield was located at Iwakuni MCAS, near Hiroshima, Japan, 34.08.6 N, 132.14.2 E. During the intensive operation period, the research aircraft conducted 19 research missions with spatial coverage from 10 to 50 N and 100 to 170 E (Figure 2), and at altitude up to about 7 km.
 The University of Washington light scattering measurements aboard the research aircraft were used in this study to derive the mass of the aerosol particles, including supermicron and submicron particles. The light scatterings of the total aerosol and of only the submicron aerosol were simultaneously measured from integrating nephelometers. The instrument measured integrated total scatter and hemispheric backscatter at 450, 550, and 700 nm wavelengths [Anderson et al., 1996; Anderson and Ogren, 1998]. Nephelometer A measured total aerosol particles; nephelometer B usually measured only aerosol of dry aerodynamic diameter less than 1micron. Periodically nephelometer B was switched to measuring the total aerosol as a check that the two nephelometers were in agreement and to provide field data on instrument precision. Data were collected between 0.2 and 2 sec resolution, and were processed at 1-min resolution. More detail descriptions about the data can be found from Masonis et al.  and Anderson et al. .
 The supermicron and submicron particles masses are calculated from the following formulas: submicron mass = NSUB/mass scattering efficiency of submicron particles; supermicron mass = ((1 − Fscat)*NTOT)/mass scattering efficiency of supermicron particles; and total mass = submicron mass + supermicron mass, where NSUB is the light scattering of submicron aerosol; NTOT is the scattering by the total aerosol; and Fscat is the submicron fraction of scattering (i.e., scattering of submicron aerosol/scattering of the total aerosol).
 Both light scattering of submicron (NSUB) and total aerosol (NTOT) particles were measured at 550nm and low relative humidity. The mass of the supermicron particle is derived from the total masses of the dry particles and the submicron particles. The cut-off point of the aerosol impactor is in aerodynamic diameter of 1 μm at low relative humidity, which is equivalent to about 0.8 μm in geometric size, and hence this study used the 0.8 μm as the cutoff size to calculate the supermicron and submicron particle masses. The mass scattering efficiencies and their uncertainties used for supermicron and submicron particles were 0.8 ± 0.4 and 3 ± 1 m2/g, respectively, and were derived from the concurrent measurements of the aerosol light-scattering coefficient and some estimates of particle mass [Andreae et al., 2002]. The mass of supermicron and submicron particles derived from the above lightning scattering formulas has 30 to 50% of uncertainties due to uncertainties in the mass scattering efficiencies. Since aerosol mass concentration on the NCAR C-130 was not in released data, the aerosol mass derived from the light scattering method were used to compare with the model calculations.
 The VOCs used as indicators/tracers of the pollution were measured using gas chromatographic techniques obtained from 1322 whole air samplers collected aboard the NCAR C-130 aircraft. Some of the measured VOCs and halocarbon were validated and available for use. In this study, the observed ethyne (C2H2), tetrachloro-ethylene (C2Cl4,), and toluene were used to indicate the pollutant fractions in the submicron particles because of their good concentration correlations with the submicron particles.
4. Result of Dust Fraction in the Supermicron Particles
Figure 3 RF(1)-(19) show the vertical profiles of the calculated masses of supermicron particles and dust compared with the observed mass of supermicron particles obtained from research missions 1 to 19, except for Mission 16, which will be discussed later. The observed masses of supermicron particles obtained during each research mission were presented in Figure 3 in 1-km average. Figure 3 shows that the observed supermicron particle concentrations ranged from trace to more than 400 μg/m3 in the troposphere. These missions can be categorized into 3 groups according to their profiles. The first group, group A, as measured during Missions 6, 7, 8, 10, 13, 14 and 15, showed high concentrations of supermicron particles below 2 km and decreasing concentrations above. The second group, group B, as measured during Mission 5, 9, 17 and 19 showed high concentrations of supermicron particles between 2 to 5 km. The rest of the missions, group C, had smaller concentrations of supermicron particles, sometimes with concentrations as low as 20 μg/m3 near the surface.
 Since the airborne measurement was conducted during the period of extensive Asian dust outbreaks, the extremely high concentrations of aerosol concentrations measured during Missions of group A and B could be mainly attributed to dust. The model results in Figure 4 show that the high concentrations of supermicron particles near the surface in group A was due to dust outbreaks that occurred in China mainly during 7–10 April (Figure 4a) and 22–23 April (Figure 4b). These dust storms followed frontal passages and moved out to downwind areas, contributing to the high particle concentrations in the boundary layer. During the dust events, high concentrations of supermicron particles (more than 300 μg/m3) were first observed below 2 km in Missions 6 and 13. The particle concentrations decreased in the next one or two days, but still remained high during Missions 7, 8, and 14. Missions 10 and 15 in group A also observed high concentrations of supermicron particles, probably from smaller dust storms that occurred during 17 to 19 and 27 April.
 In addition to frontal movement, dust can also be uplifted from surface to higher altitudes through rapid vertical transport [Chang, 1996; Chian et al., 1996; Uno et al., 2001]. The uplift probably occurred during Missions 5, 9, 17, and 19 of group B, of which concentrations of supermicron particles greater than 75 μg/m3 were observed in the free troposphere between 2 to 5 km.
 Missions of group C had concentrations of supermicron particles less than 75 μg/m3. Concentration of supermicron particle as low as 20 μg/m3 throughout the troposphere was observed during Mission 3 in group C. Although some missions of group C observed low concentrations of supermicron particles, most missions still observed relatively high concentrations in the surface layer or the free troposphere, indicating the presence of continental aerosol sources.
Figure 3 also shows that the vertical profiles of the calculated supermicron particles and their standard deviation resulted from averaging over the entire flight areas and time periods. Figure 3 shows that matched reasonably well with the observations. The vertical profile of the calculated supermicron particle concentrations mostly fell within the standard deviation of the observations. Although some results could slightly deviate from the observation at some altitudes, Figure 3 shows that the model reproduced most of the concentrations of supermicron particles well. The slight deviations were partly due to the coarse model resolution above the mixed layer. The model resolution of 1–2 km above the boundary layer was too coarse to resolve the detail profile of supermicron particles.
 Another reason for the slight deviation of the model results compared with the observations was because the model results were averaged for the areas surrounded by the flight track rather than along the flight track. The average model results might not represent conditions along the flight track where the samples were taken. In addition, the C-130 deliberately flew into regions where high aerosol concentrations were anticipated. The samples are likely to be biased toward high values of particle concentrations and can cause discrepancies in comparison with model results. On the other hand, since dust storm events usually affect areas greater than hundreds or thousand of kilometers, the concentration should not be significantly different over small areas taken by aircraft measurements. In a small number of missions, the areas covered by the flight track were large and the discrepancies were substantial. In fact, the only significant discrepancies between the observations and the model calculations were in Missions 6 and 19, when dust emissions over the northern China [Liu et al., 2002] contributed to the high concentrations of the supermicron particles during the missions. Although the model calculations show high concentrations of supermicron particles near the surface and at about 2.5 km, the model significantly underestimated the concentrations.
 Uncertainty in the aerosol measurement could also contribute to the discrepancies between the observed and the calculated concentration of supermicron particles presented in Figure 3. The aerosol samples were derived from light scattering measurements collected from inlets on board the NCAR C-130. The inlet efficiency for various sizes and shapes of aerosol is an ongoing study, and descriptions and evaluations associated with inlet and plumbing issues are presented elsewhere in this issue. The aerosol mass derived from the lightning scattering measurement also has large uncertainties in terms of the mass scattering efficiencies. As discussed in section 3, the derived aerosol mass has uncertainties of 30–50% due to uncertainties in mass scattering efficiency, and can result in the discrepancies between the model calculations and the observations shown in Figure 3.
Figure 3 also shows that the calculated dust mass of the supermicron particle for all the flight missions reached a maximum below about 6 km. The dust concentrations mainly remained below 2 km, although during some flight missions, the dust concentration extended several kilometers beyond 6 km, where aircraft measurements were usually not available. Some observations showed that dust uplifted into the free troposphere was transported across the North Pacific [Duce et al., 1980; Shaw, 1980; Uematsu et al., 1983; Merrill et al., 1989; Bodhaine, 1995; Husar et al., 1997]. Instead of finding the high concentrations of dust in the free troposphere, here it showed that during dust events, the dust impacts were restricted in the lower troposphere, mainly in the boundary layer.
Figure 3 shows that most of the supermicron particles measured during the flight missions were composed of dust. In most of the flight measurements, the predominant of dust in the mass of supermicron particles was obvious throughout the troposphere. However, for some cases near the boundary layer, where the pollutants were encountered, the dust fractions in the supermicron particles were small. Although sea salt aerosols are also a component of supermicron particles in the marine boundary layer, few sea salt aerosols were observed during the missions due to low wind speeds over the ocean surface [Anderson et al., 2003].
Table 1 summarizes the calculated total mass of supermicron particle and its dust fraction above and below 2 km in the troposphere for each flight mission. The 2 km height was selected to distinguish between the aerosol characteristics above and below boundary layer. Table 1 shows that the calculated total mass of supermicron particle ranged from very small to more than 400 mg/m2 in the troposphere. Without dust storms, the total masses of supermicron particles were usually much greater in the boundary layer than above it. During the strong dust storm period as observed during Missions in groups A and B, the total mass of supermicron particles was mostly more than 100 mg/m3 above or below 2 km, depending on where the dust was concentrated.
Table 1 shows that dust was a significant component of the supermicron particles over the downwind areas of dust outflow. Dust fractions in the supermicron particles obtained from research flights were mostly greater than 45% and 60% in and above the boundary layer. During the dust storm events, the percentages were usually larger than 90% at the altitudes where dust was concentrated. Consistent with the previous studies [Zhou et al., 2002; Chun et al., 2001; Cahill et al., 2002], the results here shows that the dust significantly affects the concentration of the supermicron particles over the east China Sea during this time.
5. Result of Dust Fraction in the Submicron Particles
 In contrast to supermicron particle, a significant amount of submicron particle comes from gas-to-particle conversion of anthropogenic air pollution, and thus model calculation of submicron particle concentration is sensitive to emission inventory of air pollutants. In order to understand the pollution fraction of the submicron particles observed during ACE-Asia aircraft measurement, we examined the correlations between the observed submicron particles and VOCs. Some VOCs observed during aircraft experiments were anthropogenic generated, and are indicator of pollutions. Ethyne (C2H2) was chosen from the VOCs because of its similar lifetime with submicron particles (about 2 weeks) and thus the two concentrations are better correlated than with other observed VOCs.
Figure 5. RF1-15 presents the concentration of submicron particles versus ethyne from the available aircraft measurements observed below 2 km. Some flight measurements are missing in Figure 5 due to lack of VOCs observations. Since most of the aircraft measurements were conducted below 2 km, where more than 700 aerosol samples were obtained from each mission, the samples were sufficient for statistical significance. In Figure 5, the observed concentrations of the submicron particles were best fitted by linear regression, with 95% confidence interval plotted for each mission. Figure 5 shows that observed concentrations of ethyne and submicron particles were mostly well correlated. The correlations coefficients varied from 0.25 to 0.83, but usually greater than 0.5. Because ethyne is usually generated from combustion processes, the good correlation of the concentrations between submicron particles and ethyne suggested that a major amount of the submicron particles obtained during ACE-Asia aircraft experiments came from anthropogenic emissions.
 The correlated submicron particles and ethyne could be either locally produced or long-range transported from upwind areas. Because various sources of aerosol with different pollutant components were collected during different missions, the correlations of the concentrations between the submicron particles and the ethyne also changed accordingly, and no significant difference in the correlation was found between dust and nondust period.
 During dust storm events, both strong and weak correlations between the submicron particle and ethyne were observed. Missions 6, 7, 8 and 9 conducted during strong dust events observed correlation coefficients above 0.7, implying high percentage of pollutants involved in the submicron particles. On the other hand, Missions 10 and 13 only observed correlation coefficients of 0.57 and 0.29 during the dust event, suggesting that some of the submicron particles and ethyne came from different sources.
 Both high and low correlations between submicron particle and ethyne were also observed during small and nondust period. With low dust concentrations, Mission 2 obtained a correlation coefficient of more than 0.8, indicating that significant amount of pollutants could be involved in the submicron particles during the missions, consistent with the relatively lower dust percentage found in the boundary layer as shown in Table 1. On the other hand, a low correlation does not always correspond to less pollution. Missions 1 and 4 conducted under low aerosol concentrations observed correlation coefficients of 0.26 and 0.58, because more than one source of pollutants were measured during the missions.
 With small amount of dust present in the troposphere, Mission 11 also observed correlation coefficients of 0.25 between submicron particle and ethyne (Figure 5). The low correlation observed during Mission 11 resulted from 2 sources, one with high correlation and the other with no correlation, suggesting that the submicron particles were mixtures of pollutant and nonpollutant components. The nonpollutant component reached a value of more than 80 μg/m3. In addition to dust, a volcano could also contribute to the submicron particles, because Mission 11 flew around an active volcano south of Japan.
 To examine the model results, Figure 5 also shows the calculated submicron particles versus ethyne for each mission. Since ethyne was lumped with other alkane species in the model for simplifying the calculations of atmospheric VOCs reactions, the ethyne concentrations were calculated from the lumped alkane concentrations according to their observed concentration ratio [Talbot et al., 1996; Blake et al., 1996]. Figure 5 shows that the model results did not match the observation very well. In all the missions, the discrepancies of the averaged concentrations between the model results and the observations was about a factor of 1 to 3 for submicron particles, but was about a factor of 1 to 5 for ethyne. Other anthropogenic VOCs (such as toluene and tetrachloro-ethylene, i.e., C2Cl4) were also applied to substitute ethyne in this correlation study, but the discrepancies between the model calculations and the observations of the submicron particles as well as the anthropogenic VOCs were still significant, implying that there were uncertainties in the emission inventory of both species.
 As mentioned previously, the emission inventory of air pollutants over east Asia is not well understood, and its uncertainty is probably the main reason that results in the concentrations discrepancy between the model calculation and observation of ethyne as well as submicron particle in the boundary layer. Sources of submicron particles include direct emissions and gas to aerosol transformation, and both have large uncertainties. The gas phase chemicals that can transform into submicron particles include sulfur dioxides and VOCs. The VOCs emissions is not well known and have lead to the discrepancies in the model calculation of ethyne concentrations. On the other hand, the discrepancies of the calculated submicron particles are smaller, because a significant portion of submicron particle comes from sulfate dioxides, whose emission is better estimated [Benkovitz et al., 1996]. In addition to emissions, the aerosol module applied in this study can also lead to the underestimate of the submicron particle concentration. For example, in the aerosol module, the gas-to-particle conversion of VOCs only included the processes of condensation, while chemical reactions of aerosol were neglected.
 Same as for the supermicron particles, the vertical profiles of the submicron particle mass and its dust fraction of each mission were calculated and compared with the observations. Despite that the calculated concentrations of submicron particles for all the missions were underestimated in the boundary layer mostly due to the uncertainty of the pollution emissions, their vertical profiles matched reasonably well with the observations above the boundary layer. Because the major components of aerosol particles were dust and pollutants over the Asian dust downwind areas [Anderson et al., 2003], the results indicated that the dust concentrations of the submicron particles were well calculated. Therefore the calculated dust concentrations were used to derive the dust fractions of the submicron particles. The dust fraction was calculated by dividing the calculated dust mass by the observed total mass of submicron particles observed during each mission. Since most of the aircraft measurements were conducted in the boundary layer, and the aircraft samples above the boundary layer were not sufficient, only the dust fractions in the boundary layer were calculated.
Table 2 shows the observed concentrations of submicron particles and their dust fractions calculated for the available missions. The calculated dust fraction varies from 10% to more than 57% depending on the missions. During strong dust events when high dust concentrations were observed near the surface (missions in group A), the dust fractions in the submicron particles is usually greater than 29% (Missions 6, 8, 10, and 13 in Table 2), and can be more than 56% in the boundary layer. Mission 7 is the only case when low dust percentages were observed during dust events. Table 2 shows that the dust percentage as low as 24% was observed during Mission 7, indicating that significant pollutants were also presented during dust events, consistent with the findings obtained from Figure 5. Anderson et al.  found that dust was usually accompanied by pollution. The low dust percentage as well as the high correlation of ethyne and submicron particle observed during Mission 7 also implied that pollutants could come with dust through long-range transport.
Table 2. Observed Submicron Particles and Their Calculated Dust Fractions During ACE-Asia C-130 Research Aircraft Measurements
Observed Submicron Particles, mg/m2
Calculated Dust in Submicron Particles, %
 During weak dust events or when dust was not concentrated in the boundary layer, the dust fraction of the submicron particles was usually less than 28%. During Mission 2, the fraction was even as low as 10% (Mission 2), and according to Figure 5 (r2 = 0.82 for Mission2), this is because high pollutant fraction was involved. In contrast to supermicron particles, Table 2 indicates that pollutants were usually the major component of the submicron particles, even during some dust storm periods.
 To verify the model calculation of dust fractions, the correlation coefficients obtained from Figure 5 and the dust fraction derived from Table 2 were examined. Despite that other component may also be involved in the submicron particles (such as particles from local constructions, volcano and so on), air pollutants and dust are the major component of submicron particles in the dust downwind areas [Anderson et al., 2003]. Therefore, for each mission, dust fractions calculated in Table 2 and the correlation coefficients obtained from Figure 5 should be inversely related. That is, if high dust percentage is involved in submicron particles, the pollution fraction of the particles is low, and the correlation coefficient between the submicron particles and the ethyne is also low. As discussed previously in this section, there is no difference in the dust fractions (or the pollutant fractions) of the submicron particles between dust and nondust events. Therefore both high and low dust fractions were observed during strong (as well as weak) dust events. Figure 5 shows that the highest correlation coefficients (0.82 and 0.83) were obtained during Missions 2 (weak or no dust event) and 7 (strong dust event), indicating that high percentages of pollutants were involved in the submicron particles during the missions. Consistent with the results shown in Figure 5, Table 2 also shows that the calculated dust percentages as low as 10% and 24% were observed respectively during both missions. On the other hand, the lowest correlation coefficients (0.57 and 0.29 in Figure 5) observed during Missions 10 and 13 (strong dust events) were consistent with the highest dust percentages (more than half) calculated (Table 2). For other missions with correlation coefficient in between, the correlation coefficients may not be inversely proportional to the calculated dust fraction, but the correlation coefficients were still good indicators of pollutants and could be used to verify the model calculation of the dust fractions. The only exception was when more than one source of submicron particles was observed, as discussed previously for Missions 1, 3, 4, and 11. Under these circumstances, the correlation coefficients are not very good indicator of pollutants and more verification is necessary to understand the pollutant fraction of the submicron particles.
6. Special Case Discussions
 Mission 16 was conducted at 23 PM, 29 to 8 AM, 30 April in UTC time, and the measurements were taken south of Japan and east of China (Figure 6). Different from other missions, this flight reached the most southerly latitude of all the flight missions, and some of the aircraft samplings were taken downwind of Shanghai. Since the characteristics of the air samples obtained from this mission are expected to be different from the other missions, it is discussed separately in this section. Figure 6 shows the flight path of the mission and the selected domain for averaging the model results for comparison with the observations.
Figure 7 shows the backward trajectories of the air parcels starting from two locations along the flight track. The trajectories of the air parcels were calculated by applying the outputs of the meteorological model (NCAR/Penn State MM5) used in this study. As the flight headed south, Figure 7 (03 UTC, 30 April) shows the flight track was dominated by southerly wind, and the air was clean. As the flight turned north and flew closer to Shanghai (07 UTC, 30 April), the flight track was dominated by the westerly wind and the flight passed through a polluted plume. Therefore both clean and polluted air were measured during the missions.
Figure 8 shows the vertical profiles of calculated supermicron and submicron particle concentrations and their dust masses compared with the observations obtained during Mission 16. Figure 8a shows that the observed concentrations of supermicron particles reached a value of 55 μg/m3 in the boundary layer and the concentration decreased to less than 10 μg/m3 right above the boundary layer. The calculated dust mass of the supermicron particles on the other hand were quite low in the lower troposphere and increased to be the predominant component of supermicron particles above the lower troposphere. Similar to many other missions, Mission 16 (Figure 8a) indicates that dust was not the major component of supermicron particles in the lower troposphere. Since some of the measurements were taken downwind of a pollutant plume, some fractions of supermicron particles were composed by pollutions.
Figure 8b shows that the submicron particle concentration observed during Mission 16 was about 10 μg/m3 in the boundary layer and decreases to little above it. Similar to Figure 8a, Figure 8b shows that little dust was calculated in the boundary layer, suggesting that pollution component in the submicron particles was significant.
 To examine the pollutant fractions of the supermicron and submicron particles for Mission 16, the observed concentrations of supermicron and submicron were plotted against ethyne in Figure 9. Figure 9a shows that most of the observed supermicron concentrations varied from very small to over 280 μg/m3, and some high concentrations of supermicron particles were not correlated with ethyne. Although Figure 9a also shows that some concentrations of supermicron particles seem to correlate with ethyne concentrations, the correlations are difficult to identify.
Figure 9b shows that the concentration of submicron particles varied by up to about 100 μg/m3 and the concentration versus ethyne can be separated mainly into 3 groups–Groups I, II, and III. Group I consisted of the submicron particles concentrations that are not correlated with ethyne concentrations. Groups II and III both included submicron particles whose concentrations varied with ethyne concentrations, and group II had a larger slope (inclination angle) than group III. Figure 9 indicates that the submicron particles of group I mainly came from nonpollutant sources, while the correlated particles of groups II and III came from pollutions. The two slopes in Figure 9 (groups II and III) indicate that the pollutants of two different components were observed during Mission 16, and with larger slope, group II has higher submicron concentrations in the component. The uncorrelated concentrations seen in group I of Figure 9b were also be seen in Figure 9a under almost the same ethyne concentration. The extremely high concentrations of supermicron particles and the uncorrelated relationship with ethyne indicate that the nonpollutant component of the aerosol particles can be dust. Since Mission 16 covered a wide area of the East Sea, including clean and polluted air, the different groups seen in Figure 9 also indicates that the air over the area included a mixture of different sources of air mass.
 The dust fractions of the super and submicron particles in the troposphere were both calculated for Mission 16, and the fraction were 56% and 52%, respectively. The calculated dust fractions were consistent with the correlation between submicron particles and ethyne discussed in Figure 9b, where both correlated and uncorrelated submicron particles with ethyne were seen. The results also indicate that the correlation between submicron particle and anthropogenic VOCs can be used to verify the pollutant or dust fraction of the submicron particles.
 The Taiwan air quality model incorporating a dust module was applied to calculate the concentrations of supermicron and submicron particles and their dust fractions during the ACE-Asia aircraft experiments, and the results were verified with observations. The results show that there was a high percentage of dust in supermicron particles over areas affected by the Asian dust. During dust events, the contribution of dust on supermicron particles was more than 90% at altitudes where dust was concentrated. Without strong dust storms, dust was still greater than 45% and 60% in and above the boundary layer, respectively. Most of the dust was distributed below 6 km, but during some occasions, dust impacts can extend several kilometers above 6 km. During dust events, high concentration of supermicron particles was mainly observed in the boundary layer rather than in the free troposphere.
 By applying the current emission inventory over east Asia, model still underestimated the concentration of the submicron particles by a factor of 1 to 3, and a factor of 1 to 5 for ethyne and some other anthropogenic VOCs. Correlation between observed submicron particles and anthropogenic VOCs can be used to verify the pollutant fraction of the submicron particles. A strong correlation of the concentrations indicates that most of the submicron particles obtained during ACE-Asia consisted of pollution. During strong dust events, the high correlations of the observed submicron particles and the anthropogenic VOCs also implied that pollution could come with dust through long-range transport. On the other hand, a low correlation between the observed submicron particles and the anthropogenic VOCs can result from different sources or collocated sources but under several mixing processes.
 The model calculation indicated that pollution was usually the major component of the submicron particles. The dust percentage in the submicron particles was usually less than 28% and can be as low as 10% when high concentration of pollution were encountered. During dust outbreaks, the dust percentage was usually more than 40%, but can also be lower than 24% when dust was accompanied by pollution. Only during very strong dust events, the dust fraction in the submicron particles can reach as high as 57%. The model results of dust contributions on submicron particles and the correlation study between the observed anthropogenic VOCs and submicron particles are consistent. The results suggest that anthropogenic VOCs can be used as a signature of pollution to verify the pollutant or dust fraction of the submicron particles. Up to now, there is still uncertainties involved in measurements and model quantification of submicron particles in the atmosphere and little information is know about their dust and pollution contributions. Emission inventory and microphysical processes of aerosol particles are still not well known so far, resulting in large uncertainties in the model calculation of submicron particles and their dust or pollution contributions. By applying anthropogenic VOCs as indicators of pollutions, this study provided a different approach to understand the components of the submicron particles.