Aerosol composition and size versus altitude measured from the C-130 during ACE-Asia



[1] During the Asian Pacific Regional Aerosol Characterization Experiment (ACE-Asia) intensive experiment in the spring of 2001 we used a total aerosol sampler (TAS) and a micro-orifice impactor (MOI) to collect dust and pollution aerosols for ion chromatographic analysis. An aerodynamic particle sizer (APS) was used to estimate the total coarse-mode volume. We conducted postexperiment passing efficiency measurements on the APS, the MOI, and their delivery tubing to constrain the inevitable (and sometimes large) artifacts associated with sampling supermicron particles from an aircraft. We have combined TAS and corrected MOI data to estimate ambient coarse and fine sulfate, ammonium, nitrate, calcium, sodium, chloride, potassium, magnesium, and oxalate. We found significant differences between aerosol composition in the free troposphere (FT) and boundary layer (BL). The molar ratio of nitrate to soluble calcium averaged 1.8 in the BL, but only 0.2 in the FT. Nitrate and calcium frequently had identical coarse size distributions, while sulfate and ammonium often had identical fine distributions. Dust clearly directs NOy toward coarse-mode nitrate. Sulfate in the FT was closest to ammonium bisulfate (half neutralized), while non-sea-salt sulfate (NSS) in the BL was usually completely neutralized to ammonium sulfate. In the presence of dust, up to half the NSS was found in the coarse mode, probably the result of SO2 uptake by CaCO3 in the dust. Soluble calcium averaged 5–8% of the coarse dust mass inferred from the APS. BL aerosol chemistry was seldom a good indicator of ionic composition in the FT.

1. Introduction

[2] Atmospheric aerosols play a significant role in the Earth's radiation budget, and thus affect the climate [Charlson et al., 1992; National Research Council Panel on Aerosol Radiative Forcing and Climate Change (NRC), 1996]. In fact, aerosols represent the largest uncertainty in modeling the radiative forcing of climate [Intergovernmental Panel on Climate Change (IPCC), 2001]. Because of its widespread spatial distribution and large optical depth, the radiative effects of mineral dust are important compared to other types of aerosols [Sokolik and Toon, 1996]. Dust particles modify both short- and long-wave radiation by scattering in the visible and absorbing in the infrared part of the spectrum [Andreae, 1996]. Most mineral dust particles originate from desert regions such as those in Africa and central Asia. Some ascend rapidly to high altitudes where they are then transported over long distances [Prospero et al., 1989]. It is estimated that between 500 and 5000 Tg of mineral aerosol enters the atmosphere each year [Duce, 1994; Sokolik and Toon, 1996]. This will increase if desertification continues. Mineral dust can influence global atmospheric chemistry, cloud properties, and precipitation development [Levin et al., 1996]. Dust has also been found to be a major source of iron and other nutrients controlling ocean productivity [Martin et al., 1994; Bergametti, 1998]. Modeling studies have indicated that the chemistry of the troposphere can be significantly impacted by mineral aerosol reactions with O3, SO2, HNO3, and NOx [Dentener et al., 1996; Tang et al., 2004].

[3] The development of remote sensing instruments for aerosol measurements has improved our understanding of mineral dust processes. Instruments such as the Aeronet CIMEL sky radiometers, the EOS Multiangle Imaging Spectroradiometer (MISR), and the LITE space lidar can constrain certain aerosol optical properties [Sokolik et al., 2001]. Remote sensing techniques, however, are limited in that they cannot measure the dust chemistry and morphology that surface and aircraft measurements can. Coupling remote sensing data with in situ measured chemical properties can therefore improve the accuracy of model calculations [NRC, 1996].

[4] Calcium carbonate in dust influences the gas-aerosol partitioning of semivolatile inorganic components such as SO2, ammonia, and nitric acid [Dentener et al., 1996; Sokolik et al., 2001]. As coal burning and industrial processes release increasing amounts of SO2 and NOx into the atmosphere, it is essential that we quantify the processes that control their transport, evolution, and eventual deposition. Potentially saturated pathways are of particular interest. If the alkalinity in dust is titrated by NOy and SO2 almost to its end point, any additional NOx or SO2 might have a very different fate (submicron particles). Such nonlinearities could be common in dusts of various compositions and concentrations.

[5] Previous chemical and physical studies of Asian aerosols have largely relied on surface measurements [Choi et al., 2001; Kim et al., 1998; Zhang et al., 2000; Falkovich et al., 2001]. However, surface measurements may not be representative of aerosol chemistry in elevated layers: There is likely to be less modification by surface pollutants in isolated aerosol layers aloft. How neutralized is sulfate by ammonium or calcium by sulfate in the FT? Surface measurements are poorly posed to study reactions during transit above the BL. Dust probably evolves chemically as it moves, causing changes in factors such as its hygroscopicity, the angular dependence of its light scattering, and thus its climate forcing.

[6] Aerosol characterization of layers in the boundary layer (BL) and free troposphere (FT) can be used to improve models of aerosol processes and effects at various altitudes. However, poor large-particle inlet and plumbing efficiencies have complicated previous airborne aerosol experiments. Unquantifiable inlet losses or enhancements can skew size distributions measured from aircraft [Huebert et al., 1990b; Baumgardner and Huebert, 1993; Sheridan and Norton, 1998; Blomquist et al., 2001]. Recently, a new type of inlet has been developed at the University of Denver that uses a porous diffuser to minimize turbulence and turbulent losses of large particles [Huebert et al., 2004; Wilson et al., 2004]. The low-turbulence inlet (LTI) has made it possible to compute the relationship between ambient and sampled (apparent) coarse aerosol distributions. Though this inlet tends to enhance large particle concentrations, reliable corrections can be applied. In order for measurements to accurately represent ambient aerosol concentrations and chemistry, both the inlet efficiency and the passing efficiency of the tubing connecting the inlet to the sampler must be known, since the size-dependent inertial impaction and settling of large aerosols in splitters and tubing also skew size distributions [Huebert et al., 2004]. We report here on laboratory efficiency tests of the Asian Pacific Regional Aerosol Characterization Experiment (ACE-Asia) tubing configurations and cascade impactor, which we use to correct our airborne measurements.

[7] During ACE-Asia [Huebert et al., 2003], LTIs were used as sampling inlets aboard the NCAR/NSF C-130 aircraft. Among the instruments downstream of the LTIs were an aerodynamic particle sizer (APS) and a micro-orifice impactor (MOI). With these instruments, we were able to sample both submicron and supermicron aerosols versus altitude. In this paper, we use the results from these instruments to describe aerosol distribution and chemistry over Asia. We combine computations of LTI enhancement of large particles with laboratory measurements of tubing passing efficiency to derive submicron and supermicron ambient aerosol concentrations. We compare these results to measurements made by our total aerosol sampler (TAS), to constrain these corrections.

[8] We also compare our ion chromatography results to other data collected during ACE-Asia. This includes airborne aerosol mass spectrometer data from the Center for Interdisciplinary Remotely Piloted Aircraft Studies (CIRPAS) Twin Otter [Bahreini et al., 2003], whose shorter range kept it generally closer to the Iwakuni airfield. We did not, therefore, sample in quite the same locations except on RF3 and RF15. We also compare with particle-into-liquid-sampler (PILS) ion chromatography data [Ma et al., 2004] collected from the C-130 alongside ours. It had a much higher time resolution (∼5 min) but could not quantify the supermicron particles that our instruments did.

[9] Since we did not make gravimetric measurements of dust mass from the C-130, we consider soluble calcium as a dust proxy. Concentrations of soluble calcium have been shown to substantially increase during dust storm periods [Park et al., 2003; Choi et al., 2001]. One difficulty with this proxy is that the fraction of total calcium that is soluble may vary with source region or processing. During the measurements reported here, Arimoto et al. [2004] found that nearly all calcium was water soluble at Zhenbeitai, near the Chinese Loess Plateau, but only about a third of total calcium was water soluble at Gosan, Korea. Among the possible explanations is that insoluble fly ash and industrial aerosols may contribute more to total calcium near the coast. If that were so, the ratio of soluble calcium to dust mass might not be changed much from the source area. At Zhenbeitai soluble calcium was approximately equal to aluminum, about 6% of the dust mass.

2. Methods

[10] A variety of chemical and physical aerosol properties were measured by a series of instruments aboard the NCAR/NSF C-130 aircraft as a part of ACE-Asia [Huebert et al., 2003, Technical Appendix]. The aircraft measurements were conducted from late March to early May of 2001 over nineteen research flights and three ferry flights. Our sampling was usually confined to level legs. We sampled air from the Sea of Japan, the East China Sea, Gosan, the Yellow Sea, and the Pacific east and south of Japan. Conditions of the flight legs ranged from extremely dusty to heavily polluted to relatively clean air. Meteorological forecasting and chemical transport models were used to locate interesting features for sampling. As a result, our extensive-variable (concentration-dependent) statistics are no doubt skewed toward higher concentrations (relative to the seasonal averages) but our intensive-variable statistics (such as ion ratios) benefit from the best possible signal-to-noise ratio.

[11] We have chosen to include a substantial amount of detail about our inlets and measurement systems in sections 2 and 4, to help readers understand how much confidence they can place in these measurements. Even though we used a low-turbulence inlet, transmission efficiencies are rarely unity for large particles. While some readers might prefer less technical detail, the technology for sampling and conveying large particles is not yet well enough developed to simply refer to other published discussions of the critical measurement issues. Users who understand the caveats on this data will hopefully become more critical consumers of other aerosol data sets.

2.1. Total Aerosol Sampler (TAS)

[12] A total aerosol sampler (Figure 1 and Huebert et al. [2004]) was used for bulk aerosol sampling. For each sample we inserted a coupled diffuser cone liner and Gelman Zefluor Teflon filter, both of which were removed and sealed immediately after the sampling was terminated. A different cone liner/filter pair was used for each sample. After the flight, all the filters and the cone liners were separately extracted in order to analyze everything that entered the TAS tip during each sampling interval with unit efficiency, regardless of size. This allowed TAS to serve as an ambient reference for other aerosol sampling systems, which employ inlets, splitters, and delivery tubing.

Figure 1.

Total aerosol sampler, TAS. A cone liner/filter pair is used for each sample and then replaced. The cone liner collects any particles that fail to reach the filter, so they can be extracted and analyzed. In this way, every particle that enters the TAS tip is retained, extracted, and analyzed, with no inlet or tubing losses.

[13] For analysis, each TAS filter sample was placed in a microclean polyethylene bag and extracted using 1 mL of ethanol and 9 mL of weak acid solution (10−5 M trifluoroacetic acid) in an ultrasonic bath. Each cone liner was extracted with 30 mL of weak acid solution and manual agitation for 2 min. Cl, NO3, SO42−, oxalate, Na+, NH4+, K+, Mg2+, and Ca2+ were analyzed by suppressed ion chromatography on two Dionex ICs using procedures identical to those described by Huebert et al. [1998]. The amount of nonseasalt (NSS) sulfate was calculated from the measured sodium and total sulfate concentrations by assuming the mass ratio of sea-salt sulfate and sodium in the aerosol to be the same as that in seawater. This might cause us to slightly underestimate NSS in high-altitude dust, which contains some Na. A little more than half the TAS samples were coincident with MOI samples. (There were 68 TAS samples, 60 MOI samples, with 39 in common.)

2.2. Aerodynamic Particle Sizer (APS)

[14] An aerodynamic particle sizer (TSI Model 3320) was used to count and size particles according to their aerodynamic diameter. The APS ran in correlated mode from flight 9 through the end of the experiment with a beta version (2.08) of a PROM designed to avoid the artifacts noted by Armendariz and Leith [2002]. Particle concentration data was logged into 52 aerodynamic size bins between 0.5 and 20 μm and sixteen light side-scattering channels. The APS sample air entered the NCAR/NSF C-130 via a starboard-side LTI with a flow rate of 136 L/min. Since we evaluated transmission losses between the LTI and several instruments (including the APS), we detail this plumbing in Appendix A.

[15] Passing efficiency from the LTI to the APS and other rear-rack instruments was measured by simulating the aircraft plumbing in our lab and sending monodisperse oleic acid particles generated by a vibrating orifice aerosol generator (VOAG, TSI model 3450) through the plumbing. The particles were tagged with the fluorescent tracer Rhodamine-B and transmitted at the ACE-Asia flow rates.

[16] Filters were placed behind each length of tubing to collect the particles that penetrated the plumbing. Upon completion of a run, each piece of the system and the filters were extracted with 2 mL of isopropyl alcohol and 30 mL of distilled water and the concentration of the tracer in the extract was measured with a fluorometer. Table 1 summarizes the results of the passing efficiency measurements to each of the instruments in the rear rack. Since deposits of Rhodamine-B are bright pink, loss patterns could be seen on many pieces. Gravitational settling was seen for both 10 and 7.5 μm particles in the first splitter. Rhodamine-B could also be seen along the outside of the curved bellows section for large particle runs. By extracting each piece of tubing separately, we were able to get a clear picture of where particles were lost. For example, the percent transmission of 10 and 3 μm particles through each major plumbing piece to the APS are shown in Figure 2. Here, the most 10 μm particles were lost in the bellows and the straight stainless pieces.

Figure 2.

Cumulative percent transmission of 10 (black) and 3 (grey) μm particles to the APS through each major plumbing piece on the C-130 aircraft during ACE-Asia. Most 10 μm particles were lost in the bellows and the straight stainless pieces, with losses of almost 70% for 10 μm particles. Caution should be taken in using these results in that they represent the passing efficiency of sticky particles, such as sea salt.

Table 1. Cumulative Tubing Transmission Efficiencies to Rear-Rack Instrumentsa
 Aerodynamic Diameter
3 μm5 μm7.5 μm10 μm
  • a

    Values are given in percent.

Aerodynamic particle sizer (5 L/min)95825327
University of Washington nephelometers (30 L/min)95875734
Radiance Research nephelometers (12 L/min)95845226

[17] A model that predicts losses in plumbing was constructed using the TSI Aerosol Calculator (based on equations given by Willeke and Baron [1993]). This model's losses were consistent with what was found experimentally for 3 μm particles, in that there was a 93% transmission to the UW nephelometers. However, the modeled transmission for 10 μm particles was much larger than the actual transmission: The predicted efficiency was 57% while our measured efficiency was 35%. The model did not predict as much loss in the bellows (we did not have a model for this geometry, so used a curved tube model instead) or the long stainless steel section as was seen, and it overestimated losses in the conductive tubing. These results represent the passing efficiency of sticky particles, which would be similar to sea salt. The passing efficiency of dust particles could be larger, since under some conditions dust particles may bounce upon wall contact and be re-entrained into the flow. We could not confirm that in the laboratory; however, tests done with dry, 3 μm KCl particles produced results very similar to those of the sticky oleic acid particles.

2.3. Micro-Orifice Impactor (MOI)

[18] A micro-orifice impactor (MSP Corp.) was used to measure size-segregated aerosol chemistry. The MOI operated at a flow rate of 100 L/min. Air to be sampled by the MOI entered the aircraft by a port-side LTI. The entire LTI flow either passed to the MOI or to an auxiliary (AUX) 90 mm Teflon filter that was used when no MOI sample was being taken. The MOI consisted of a 10 μm scalper followed by two sharp right angle turns into its distribution plenum: a square tube connected to eight impactor stacks. Each stack contained five sample stages and a backup filter, with the nominal 50% cut points of the stages being 5, 1.4, 0.77, 0.44, and 0.25 μm. Aluminum foil substrates were used in each impactor stage with a 90 mm, 1 μm pore-size Gelman Zefluor Teflon filter as the backup stage. The Al substrates were punched from foil in the lab and washed, to reduce blank variability. The Gelman Teflon filters were also washed to reduce the high nitrate variability in their blanks: They were first wetted with alcohol and then rinsed 3 times with DI water prior to drying on a laminar flow clean bench. The stacks were loaded and unloaded in a glove box to minimize contamination. Each filter was extracted using the same procedure as for the TAS filters, while each foil substrate was extracted with 10 mL of weak acid solution and manual agitation. All samples were analyzed for the ionic species listed above by the same procedure as for the TAS. Samples were analyzed as soon as possible, usually within 24 hours.

[19] The MOI was not ideal in several respects. During flight, ram pressure caused the sampling stacks to expand, changing the jet-to-plate distance. Once this problem was identified, we wedged the stacks in place to prevent expansion. There is little or no MOI data from flights 12–14, since we found a leak: The o-ring seals between the stacks and the plenum could roll out of their retaining slot, which could not be detected visually. From research flight (RF) 15 on, we leak-tested every part of the MOI system before each flight to detect and remediate any displaced o-rings.

[20] We measured the MOI efficiency in the lab in the same manner as described for the APS: We extracted the foil substrates, backup filter, and each instrument section. Table 2 shows the measured efficiencies of monodisperse aerosols both to substrates as a whole and to their intended substrate. For large particles, the majority of losses were in the scalper and the plenum. During lab tests we found that some of the first-stage jets were positioned above the substrate's hold-down ring rather than the substrate itself, so that some particles were deposited onto this ring. This ring was not extracted and analyzed during the field experiment, but that fractional loss has now been included in our efficiency corrections.

Table 2. Passing Efficiencies to the MOI Stagesa
 Aerodynamic Diameter
1.3 μm2.5 μm5.0 μm7.5 μm10 μm
  • a

    Values are given in percent.

Efficiency to any foil92702360
Efficiency to intended foil(s)64591340

[21] Just as with the APS efficiency tests, caution should be used in interpreting these MOI results: They represent the passing efficiency of sticky particles. Dust particle efficiencies may at times be higher, under dry conditions where rigid particles might bounce.

3. Data Processing

3.1. LTI

[22] The efficiency of the LTI inlet is a function of many factors, including airspeed and particle size. The LTI, in general, enhances particles in the 0.7–1.4 μm range by 3%, in the 1.4 to 5 μm range by roughly 26%, and in the 5–10 μm range by about 50% [Huebert et al., 2004; Wilson et al., 2004]. Numerical modeling using the Fluent program at Denver University was used to construct general relationships that were then fitted to the conditions of each leg to compute the LTI efficiencies.

3.2. APS

[23] The APS data were converted from instrument conditions to standard temperature and pressure conditions, and the following corrections were applied. First, when the APS ran in correlated mode it was apparent that some aerosol particles scattered less light than they should based on their apparent aerodynamic diameter. These “ghost particles” were caused by the slow recirculation of particles through the sensor path. Figure 3 shows a typical plot of fractional number of particles in each scattering channel versus aerodynamic channel. The scattering of particles generally increases with aerodynamic size, so the few low-scattering particles having apparently large aerodynamic sizes (lower right of the figure) are anomalous. These either recirculated around the edges of the jet or were not fully accelerated; their low speed made them seem larger than their actual size. We eliminated them from each APS size distribution. Since we have found in the past that a “loose correction” (removing only the most extreme lower right particles) gives better agreement with OPC and nephelometer data than a strict correction, we have used the loose ghost particle correction. Second, a Stokes correction assuming a density of 2 g/mL was applied to the APS data as described in the TSI Aerosol Instrument Manager Software for APS Sensors, 2002. Finally, LTI and plumbing passing efficiencies were calculated and applied for each APS size bin. The effect of each of these corrections on a volume size distribution for a level leg of Flight 18 is shown in Figure 4. To convert APS aerodynamic to geometric diameter, we assumed a density of 2 g/mL.

Figure 3.

Fractional number of particles in each scattering channel versus aerodynamic channel for a typical C-130 APS sample during ACE-Asia. “Ghost particles” are low-scattering particles with large apparent aerodynamic sizes probably caused by recirculation through the APS sensor path. We correct the APS data to not include these particles using the loose correction, excluding particles below the dashed line.

Figure 4.

Effects of each APS correction on the measured size distribution of samples taken on the C-130 aircraft during ACE-Asia flight 18 starting at 0306 coordinated universal time (CUT). These corrections include conversion to standard temperature and pressure conditions, ghost particle correction, Stokes correction assuming a density of 2 g/mL, a correction for LTI enhancement of large particles, and passing efficiency corrections.

3.3. MOI

[24] Because of nonunity efficiencies for both the LTI inlet and the MOI itself, significant size-dependent corrections (Tables 2, 3, and 4) were applied to the apparent MOI concentrations to estimate ambient concentrations. We corrected for bounce from impactor stages, LTI enhancement, and tubing losses.

Table 3. Statistics on Percent Mass of Each Chemical Species Affected by Bounce Correctiona
  • a

    SEM is the standard error of the mean: the standard deviation divided by the square root of N.

Table 4. Factors Used to Correct MOI Data for LTI and Tubing Efficiencies
 Filter0.26 μm0.44 μm0.77 μm1.40 μm5.00 μm
Tubing Efficiency1. (0.1–0.35)

[25] Counterintuitive calcium size distributions were seen for many samples: Coarse-mode calcium masses (MOI stages 1 through 3) were often high, with little or no calcium on the smaller stages but much higher calcium on the backup filter. We assume that this large backup filter calcium had bounced from an earlier stage where it belonged. Since the jet velocity increases down the stack, anything that bounced from its intended stage would be even more likely to bounce from later stages, until it was collected on the backup filter. We corrected for this effect by assuming that large-particle ions (calcium and sodium) should only be half as concentrated on the backup filter as on the stage above the filter, and that any excess above this was a result of bounce. This excess calcium and sodium was subtracted from the filter and added to the first MOI stage in order to conserve mass. The decision to place it all on the first stage (rather than distributing it somehow among the coarse stages) was arbitrary, so the shape of the resulting coarse size distribution is also arbitrary. Sulfate and nitrate were also bounce corrected, by assuming that the ratios of these ions to calcium on the uncorrected first stage represented their actual ambient coarse-mode ratios. The amount of these species moved from the backup filter to the first stage was derived from these ratios and the amount of calcium that was moved. Table 3 shows statistics relating to these bounce corrections.

[26] The LTI enhanced the concentration of large particles relative to ambient concentrations as described above. By applying the modeled enhancements to the nominal cut points of the MOI, efficiencies were derived for each stage during each sample. The average enhancement factors used for free troposphere (FT) and marine boundary layer (MBL) MOI samples are shown in Table 4, along with tubing loss factors from our lab efficiency tests. However, other complications remain. For example, the first stage of the MOI should contain aerosols between 10 and 5 μm in size. In the lab we found that only 4% of 7.5 μm particles and 13% of 5 μm particles could reach their intended substrate. The correction factor is highly variable with size, even within the design range of a single stage. Corrections for stage 1 would involve multiplying apparent concentrations by a factor of 8 for 5 μm and 25 for 7.5 μm. Slight differences in the ambient size distribution imply large differences in correction factors.

[27] We responded to this uncertainty in the correction factors by making the factor for the largest stage into a variable parameter (within the physically appropriate range for that stage) that we used to force the total ion concentrations on all MOI stages to approximate that of TAS ions for simultaneous samples. Since we wanted all the ion concentrations to match up (but could not justify applying different efficiency factors for each ion), no individual ion was in exact agreement for simultaneous TAS and MOI samples. For most samples we used a stage 1 efficiency of 0.15, but it was sometimes as small as 0.1 or as large as 0.35. It is important to note that this is an additional assumption, which affects the apparent MOI coarse size distributions. Table 4 shows the plumbing efficiencies used to correct the ionic concentrations for each MOI stage. The effect of each of these corrections on a calcium size distribution from Flight 7 is shown in Figure 5.

Figure 5.

Effects of each MOI correction on the measured mass distribution of calcium taken on the C-130 aircraft during ACE-Asia. These corrections include bounce, LTI enhancement of large particles, and passing efficiency.

[28] However, even though the coarse shape contains arbitrary assumptions, the coarse-mode and fine-mode concentrations are more robust. Since the submicron stages required virtually no correction (except for subtracting bounced material) their sum is a defendable submicron concentration. Since TAS constrained the total concentrations, the difference between TAS and the submicron concentrations is a reasonable approximation of coarse-mode concentration for each analyte. Thus the submicron and supermicron concentrations are defendable, even though the coarse distribution's shape is not. The ionic ratios in the coarse and fine modes are also well constrained from uncorrected stages that collected only coarse or fine particles, permitting defendable statements about the NO3/Ca and NH4/SO4 ratios in each mode, for instance. This information can be used to assess chemical impacts on optical properties of the coarse versus fine modes. These ratios and the modal masses should be robust, to the extent that the particles that bounce are chemically similar to those that do not.

4. Results

[29] Since TAS sampling eliminates the loss of large particles and the data processing involves only blank subtraction, it is the most accurate measure of bulk concentrations. Table 5 summarizes the TAS statistics as a function of altitude, while Figure 6 contains altitude plots for nitrate, NSS, soluble calcium, and ammonium. Not surprisingly, the average concentration of most chemical species is higher in the BL than the FT. The exception to this statement is calcium, an indicator for dust, whose FT mean was statistically identical to that in the BL. (Without the highest value, the FT mean drops to 3.3 μg/sm3.)

Figure 6.

Total (a) nitrate, (b) NSS, (c) soluble calcium, and (d) ammonium from TAS altitude profiles.

Table 5. Statistics on TAS Samples From Various Altitudesa
  • a

    Values are given in μg/sm3. SEM is the standard error of the mean: the standard deviation divided by the square root of N.

Altitude > 2500 m, N = 10
2500 m > Altitude > 500 m, N = 18
Altitude < 500 m, N = 40

[30] Table 6 summarizes the molar ratios of several ions in TAS samples for the same three altitude ranges. The ratios of pollution-derived nitrate and NSS to soluble calcium were highest in the BL. The neutralization of NSS by NH4+ was highest in the upper BL and lowest in the FT.

Table 6. TAS Molar Ion Ratios From Various Altitudes
Altitude > 2500 m, N = 10
2500 m ≥ Altitude500 m, N = 18
Altitude < 500 m, N = 40

[31] Statistics for corrected MOI data are summarized for the same altitude intervals in Table 7. Here the interesting point is the relative amounts in the coarse and fine modes for each ion. For ammonium, roughly 90% is in the fine mode, regardless of altitude. A little less NSS is fine, averaging 70–80% on the fine mode. The balance (20–30%) is probably associated with calcium on dust particles, suggesting that about a quarter of NSS was diverted to the coarse mode by alkaline dust. Nitrate was the opposite: Most nitrate was coarse in the places we sampled, with only 10–20% fine in the BL and 20–30% fine in the FT.

Table 7. Statistics on C-130 MOI Samples From Various Altitudes
Altitude > 2500 m, N = 9
Concentration, μg/sm3Median0.
Percent Fine ModeMedian3322674539183351365
2500 ≥ Altitude ≥ 500 m, N = 18
Concentration, μg/sm3Median0.
Percent Fine ModeMedian1517815410977334689
Altitude < 500 m, N = 33
Concentration, μg/sm3Median1.
Percent Fine ModeMedian61182575947011790

[32] Figure 7 shows that the MOI altitude profiles were similar to those of TAS (even though just over half their samples were simultaneous). It emphasizes the difference between nitrate (which was mostly in the coarse mode) and NSS (mostly in the fine mode). The coarse fraction of NSS was greatest in the FT, where it was least likely to be neutralized by ammonia.

Figure 7.

Altitude profiles of the sum over all MOI stages of (a) nitrate and (b) nss sulfate with coarse fraction (>0.77 μm aerodynamic diameter) as the density of grey shading.

[33] How comparable are data from TAS and the MOI? The overall passing efficiencies for the MOI can be calculated by comparing the sum of raw concentrations on all MOI stages to the TAS (Figure 8). As expected, the MOI efficiency for presumably large particles was very low, with losses as high as 70% for sodium. Even nss sulfate had reduced passing efficiencies in the MOI (69%), in part because up to half of NSS was in the coarse mode. While the submicron MOI data is only affected by the positive bounce artifact, the supermicron stages should not be used without multiple corrections.

Figure 8.

Comparison of simultaneous samples of (a) nss sulfate, (b) nitrate, and (c) sodium measured by the TAS and the MOI on the C-130 aircraft during ACE-Asia. MOI data represent the uncorrected sum of all substrates and the backup filter. Sodium, which is abundant in the supermicron mode, is undersampled by the MOI with losses of almost 70%. The mostly submicron nss sulfate particles are better sampled by the MOI but still show losses.

[34] The corrections to the MOI data (Figure 5) involved a number of assumptions, some of which were untestable. Our monodisperse laboratory aerosols could not exactly represent the tendency of ambient particles to bounce, for instance. It is difficult, therefore, to assign meaningful uncertainties to our corrected data. However, we can look at the agreement between TAS and the corrected MOI data. From Table 8 it is evident that the corrections improve the agreement between TAS and MOI data, bringing the major MOI analytes to within 20% of their TAS values. The exceptions are sodium, whose bounce correction is probably not reasonable because of the stickiness of sea-salt particles, and oxalate and K+, for which concentrations were often near our detection limit. Since the bounce by Na is probably far less than that of rigid dust particles, our decision to use the same correction factor for all species on each MOI stage no doubt caused the high corrected Na values. Figure 9 shows raw and corrected MOI concentrations of soluble calcium (summed over all stages) versus the TAS concentrations. The uncorrected MOI calcium concentration was just 30% of that for TAS. Once corrections for the LTI, passing efficiency, and bounce have been applied, the MOI results for calcium are within 20% of TAS. It is encouraging that plots of corrected MOI sums versus TAS have slopes close to one and R2 values greater than 0.8 for calcium, nss sulfate, and ammonium.

Figure 9.

Comparison of simultaneous samples of calcium measured by the TAS and MOI (uncorrected and corrected) on the C-130 aircraft during ACE-Asia.

Table 8. Ratio of Simultaneous MOI (Raw and Corrected) to TAS Valuesa
  • a

    Values are given in percent.


[35] While the corrected MOI mass size distributions (Figure 10) for soluble calcium, NSS, nitrate, and ammonium do not faithfully represent the ambient distributions, they do contain useful information. The majority of the MOI mass distributions (42 of the 60 samples) resemble Figure 10a: The NSS and NH4+ distributions have similarly shaped submicron peaks, while soluble Ca and nitrate show nearly identical supermicron distributions. A few samples, however, had NSS associated with soluble calcium in the coarse mode as in Figure 10b. Interestingly, the correlation between calcium and nitrate in the mass size distributions seems to be a function of altitude. For example, in Flight 9, all samples below 1500 m exhibited a mass distribution similar to Figure 10a, with nitrate and calcium being correlated. The one sample that was taken above this altitude had significant NSS mass in the coarse mode. This trend was seen in flights 11, 15, and 17 as well, where high-altitude mass distributions differed from those at low altitude. Of the MOI samples taken, 46 were below 1500 m, 38 of which exhibit mass distributions with nitrate being the dominant anion in the coarse mode, usually strongly correlated with calcium. Regardless of how we assigned the bounced material from the MOI backup filter, calcium and nitrate varied together from stage to stage in the coarse mode.

Figure 10.

Typical aerosol size distributions measured from the MOI on the C-130 aircraft during ACE-Asia. (a) Most common distribution (from the 50 m RF09 leg starting at 0102 CUT) with the majority of nss sulfate and ammonium in the fine mode and soluble calcium and nitrate dominating the coarse mode and (b) distribution typical of higher-altitude samples where nss sulfate was found in the coarse mode (from the 2400 m RF09 leg starting at 0808 CUT).

[36] Figure 11 shows corrected mass size distributions as inferred from the APS for three level flight legs. As discussed above, there are a number of assumptions made in correcting from apparent to ambient size distributions. These three flight legs represent three of the extreme conditions encountered during ACE-Asia. Flight 16 passed though the Shanghai plume and is a pollution size distribution. Flight 10 passed over the Yellow Sea into a large plume of dust, which is reflected in the high concentrations of 3–10 μm particles. The Flight 18 sample, which was in a relatively clean air mass, has almost no coarse particle mass.

Figure 11.

Corrected APS size distributions showing the variety of air masses encountered by the C-130 aircraft during ACE-Asia.

[37] Figure 12 demonstrates the changes in APS-derived size distributions with altitude on Flight 16, near Korea. There is an obvious change in mass versus size with altitude, with lower layers having more mass. However, there is also a change in the shape of the size distributions. The small mass during the high-altitude leg (over Korea) was in large particles, due to a dust layer. The midaltitude and low-altitude samples were collected downwind of Korea in a very polluted MBL that had high scattering and unusually high concentrations of OC, EC, and sulfate.

Figure 12.

Corrected APS distributions from flight 15 at three altitudes.

[38] A profile taken on this flight using instruments that respond far more rapidly than our collection devices (APIMS for SO2 and TSI nephelometers for scattering; Figure 13) demonstrates an important feature of aerosols in Asian outflow: Rarely does the concentration vary smoothly with altitude, contrary to what a 1-D model would imply. The FT in particular was highly layered, with relatively clean air (often with a small concentration of dust) sandwiched between layers with high aerosol concentrations. Although SO2 is a gas, its layering implies that sulfate aerosols would be similarly layered. Unfortunately we had no sulfate measurement with a fast enough response to measure sulfate profiles. We frequently found layering in the MBL as well (Figure 13), since the Yellow Sea water was colder than the air above it. On 64 nearly full profiles the median number of elevated-concentration layers was 4.5 (minimum of 3 and a maximum of 8), although they were not always as distinct and evenly spaced as those in Figure 13.

Figure 13.

Variation with altitude of SO2 and both submicron and total light scattering on RF15 near Pyongyang over the Yellow Sea.

5. Discussion

5.1. Comparisons With Published Values

[39] A number of other studies have reported aerosol chemical concentrations in eastern Asia, some of which are summarized in Table 9. In general, our boundary layer TAS results fall within the range of previously reported ground-based values, except for Na+, whose average in the TAS of 5.8 μg/sm3 is nearly double most other reported averages. This clearly depends strongly on recent winds and the times and places we chose to sample. Our average NO3 value exceeds most reported values, except for those found by Choi et al. [2001], in Seoul, Korea, during dust events. The average SO42− concentration from the TAS was relatively high, with comparable values being seen by Choi et al. [2001], again in Korea during dust events, and Kim et al. [1998], during summer time measurements on Cheju Island, Korea. TAS soluble calcium was quite high compared to listed values, except in Beijing, China, where resuspension was noted [Huebert et al., 1988]. These could be real differences due to vertical gradients (other than the Twin Otter all were surface measurements; there were sometimes low concentrations at the surface: Figure 13) or other conditions during the ACE-Asia time frame; they could be the result of our bias toward sampling aerosol plumes; or it could be that other inlet systems discriminated against some supermicron aerosols.

Table 9. Average Aerosol Composition Reported From a Number of Experiments in East Asiaa
LocationExperiment Durationand TypeSizeClNO3SO42−Na+NH4+K+Mg+Ca+NSSNotesReference
  • a

    Values are given in μg/sm3.

  • b

    Manual resuspension by street sweepers was observed.

Iwakuni, JapanMarch-May 2001, surfacePM1 2.36.1 2.0     Bahreini et al. [2003]
East Asia (ACE-Asia)March-May 2001, aircraft Twin OtterPM1 0.973.0 1.3    <100 mBahreini et al. [2003]
  PM1 1.12.5 1.1    100–1000 mBahreini et al. [2003]
  PM1 0.552.3 1.6    1000–3000 mBahreini et al. [2003]
  PM1 0.411.4 3.4    >3000 mBahreini et al. [2003]
Rishiri Island, JapanMarch-May 2001, surfaceTSP1.60.64 1.20.720.12 0.17b2.5 Matsumoto et al. [2003]
Seoul, KoreaMarch-May 1998, surfacePM10 no dustChoi et al. [2001]
  PM10 13130.95.5 0.131.3 dustChoi et al. [2001]
  PM10 8.2111. heavy dustChoi et al. [2001]
Jeju Island, KoreaJuly–Aug. 1994, surfaceTSP3.21.8124.90.891.30.360.5411allKim et al. [1998]
  PM251.00.898.< et al. [1998]
  PM25 0.1014 4.2   14high pollution 
  PM25 2.2<1.3 0.47    low pollution 
Jeju Island, Koreaspring 1992–1994, surfaceTSP1.  Carmichael et al. [1997]
Beijing, ChinaOct. 1988, surfaceTSP 4.08.9 2.3  13b  Huebert et al. [1988]
Xinglong, ChinaOct. 1988, surfaceTSP 4.06.7 2.1  2.2  Huebert et al. [1988]

[40] Concentrations of NO3, SO42−, and NH4+, as measured by Bahreini et al. [2003] using a mass spectrometer in the CIRPAS Twin Otter aircraft during the same time period, are also listed in Table 9. Our values were consistently higher than those made by the Twin Otter (TO), but one must note that the TAS measured all particles regardless of size, while the TO measured only particles less than 1 μm vacuum aerodynamic diameter, which is about 0.6 μm geometric diameter. We also did not sample in quite the same locations as the TO except on RF3 and RF15: The C-130's greater range allowed us to reach the Yellow Sea on 8 flights, for instance. Interestingly, the TO reported an increased concentration of NH4+ at their highest altitudes (>3000 m), whereas the TAS data did not show this. Ionic concentrations in fine aerosols (<1.3 μm) were also measured by Ma et al. [2004] from the C-130 alongside our samples using a PILS coupled to an ion chromatograph. The agreement between TAS and PILS measurements was best for predominately submicron species such as sulfate and ammonium, but TAS concentrations were many times higher than PILS for supermicron species such as calcium and sodium [Ma et al., 2004].

[41] A number of previous studies have reported sulfate and ammonium as dominant fine-mode aerosols in eastern Asia, with nitrate-containing particles dominating the coarse mode [Chen et al., 1997; Zhang et al., 2000]. More specifically, Zhang et al. [2000] found that in the Qingdao area sulfate mainly exists as fine-mode ammonium sulfate while nitrate was predominantly in coarse particles, as a mixture of nitrate with some sulfate. This was explained by the different volatilities of sulfuric acid and nitric acid, with sulfuric acid condensing according to surface area (we might expect it to go by diameter), while nitric acid preferred particles with the greatest neutralization capacity (mineral dust or sea salt).

[42] An important caveat is that although we have compared our mean concentrations with those of other authors to demonstrate that our data are consistent with earlier measurements, it is important to realize that the “average” atmosphere does not exist in this highly variable region. Mean values of extensive variables (concentrations, Tables 4, 6, and 8)) do not properly represent the outflow from Asia. The atmosphere was much more heterogeneous than an average implies, with layers of high and low aerosol concentrations as shown in Figure 13. While most profiles did not have 14 layers, they all contained several abrupt transitions between air masses of very different compositions. Even over the remote central Pacific, distinct layers of ozone and aerosols are common features, reported more than two decades ago in the GAMETAG program [Routhier and Davis, 1980]. The concentrations we found during any one sampling period were a function of both where we happened to sample and what aerosols were in that region.

[43] Fortunately, intensive variables (ion ratios and percent fine mode) are usually less variable. Compare the SEM of NSS and Ca high-altitude concentrations in Table 5 with the SEM of the NSS/Ca high-altitude ratio in Table 6. The concentration SEMs are about half the mean values, while the FT ratio SEM is 21% of the mean. The NSS high-altitude fine-mode fraction SEM (Table 7) is only 10% of the mean value. Whether you're sampling from the middle or the edge of a layer, intensive variables should be less variable than extensive ones.

[44] There is yet another reason to be wary of mean concentrations: The frequency distributions (Figure 14) of four species demonstrate that their concentrations are not Gaussian. In most panels a few very high concentrations were encountered, far above the rest of the distribution. This can also be seen in the fact that the means and medians in Tables 5 and 7 are very different.

Figure 14.

Number of samples with TAS concentrations in each of several concentration ranges. Each row refers to one altitude range. Note that the lowest size bins of nitrate and calcium are narrower, to avoid making the y-axis scale less sensitive. Dashed vertical lines are average concentrations for that altitude range.

5.2. Nitrate Chemistry

[45] We found high coarse-nitrate fractions that apparently resulted from the reaction of gaseous NOy (NO + NO2 + HNO3 + PAN + HONO + N2O5, etc.) with Ca in dust. The similarity of nitrate and soluble calcium size distributions is striking. In the vast majority of cases, scaling one could make it lie directly on top of the other, which suggests that the soluble calcium controlled the nitrate uptake. This observation is not sensitive to our corrections of the MOI data; it is equally obvious in the uncorrected and the corrected data. When dust is present, it affects the size distribution of nitrate.

[46] A number of studies have attempted to quantify the uptake coefficient of nitrate onto alkaline dust, but the values vary by orders of magnitude [Underwood et al., 2001a, 2001b; Grassian, 2002]. Factors such as dust composition, surface area, relative humidity, and anthropogenic contamination have been identified as influences on nitrate uptake rates [Bian and Zender, 2003; Goodman et al., 2001]. Grassian [2002] showed that mineral dust acts as an important sink for nitric acid through heterogeneous uptake. Our MOI size-segregated chemical data support these findings.

[47] Tang et al. [2004] defined an “anion limited” situation, in which CaCO3 in large dust particles takes up sulfate and nitrate and releases CO2. Tang et al.'s model predicts that more than 80% of nitrate should be in the coarse mode under these conditions. Song and Carmichael [2001] concluded that these uptake processes would increase aerosol nitrate concentrations by 10 to 40%, since nitrate aerosol is not dry deposited as rapidly as is nitric acid vapor. (Of course, this would only operate in the BL.) For about half the samples in Figure 15a, coarse nitrate is close to the Ca(NO3)2 line, suggesting that the soluble Ca has taken up as much nitrate as possible. (Given our sampling and correction uncertainties, samples slightly above the line are not significantly different from the line.) This suggests that the amount of soluble Ca may limit the fraction of nitrate that can remain in the coarse mode, as Tang et al. [2004] suggest. An alternative view is that nitric acid can stick to any surface, so that it might contribute to solubilizing Ca. Presumably, if it were only adsorbed, however, it would ultimately sublime to the most basic surface available.

Figure 15.

Relationship between soluble calcium and (a) coarse and fine nitrate and (b) coarse and fine NSS concentrations.

[48] This makes the fine-particle data in Figure 15a rather surprising. The concentration of fine (<0.77 μm) nitrate was not reduced by high levels of soluble calcium. In fact, some of the highest submicron nitrate was found in the presence of rather high Ca concentrations, suggesting that submicron nitrate is relatively insensitive to soluble Ca. This supports the conclusion of Tang et al. [2004] that high ammonia emissions in China will cause submicron ammonium nitrate to form, even in the presence of dust.

[49] Nitrate clearly is most concentrated in the boundary layer (Figure 6), since that is where the NOx sources are. This generalization does not apply far from source areas, however, where free tropospheric total nitrate (nitric acid vapor plus aerosol nitrate) concentrations exceed those in the boundary layer [Huebert and Lazrus, 1980]. Removal processes during transport deplete nitrate faster in the BL than in the FT. However, near source regions, NOx and its products are most concentrated near the surface, where most NOx is released.

[50] Any dust that was mixed into the FT prior to reacting with NOy should have a very low nitrate concentration. Simultaneous measurements at Zhenbeitai in Shaanxi Province of China [Arimoto et al., 2004] found a nitrate/calcium molar ratio of 0.015 in the most concentrated dust (when the ratio of pollution to dust should have been a minimum), suggesting that the native dust contained little or no nitrate. Beyond the Asian coast, one of the lowest coarse nitrate fraction samples was from 6600 m (Figure 7a), in air that must have contained little or no NOy from pollution. Likewise the TAS sample with the highest total soluble calcium in the FT did not contain elevated nitrate (at 4 km in Figures 6a and 6c). Dust can change the size distribution of nitrate, but the amount is ultimately limited by NOx emissions. It is likely that much of the dust transported very long distances in the upper FT (too high for the C-130 to sample) had a very low nitrate/calcium ratio, since it was often lofted near its source region and avoided the large urban NOx source areas along the coast.

[51] Why should one care about the size of nitrate? Fine-mode nitrate is usually in the form of ammonium nitrate, which can evaporate to reform nitric acid and ammonium when temperature or gas concentrations change. While nitric acid is often thought of as a terminal form of NOy, it can photolyze to reform NO2 [Huebert et al., 1990a; Liu et al., 1992], a process that may be important in remote parts of the FT. Thus tying up Asian nitrate as nonvolatile Ca(NO3)2 may reduce NOx concentrations downwind in the Pacific FT.

[52] Our group has been measuring nitric acid vapor and nitrate aerosol nightly at the Mauna Loa Observatory using Teflon/nylon filter packs for more than a decade [Galasyn et al., 1987; Lee et al., 1994; Huebert et al., 2001]. Our data for 2001 (Figure 16a) show that the impact of the dust storms we studied near Asia was evident in the central Pacific several days later. It also shows (Figure 16b) that elevated Ca levels, even though they are orders of magnitude smaller than those measured nearer the source, cause aerosol nitrate to exceed nitric acid vapor. Since HNO3 vapor was almost always in large excess over nitrate aerosol at MLO in the absence of elevated Ca (just 3–4 exceptions out of more than 300 samples), we conclude that dust reduced the potential to reform NO2 from HNO3 in the FT thousands of km downwind of the Asian deserts.

Figure 16.

Mauna Loa Observatory data during 2001. (a) Soluble calcium concentrations early in the year, showing the pulses of high concentration in the late winter and spring. (b) Nitrate aerosol and nitric acid vapor as a function of soluble calcium. Aerosol exceeded vapor nitrate concentrations only for the highest-calcium cases, showing the impact of dust.

[53] The nitrate/soluble calcium ratios in the dusty MLO samples ranged from 0.2 to 0.7. This range is similar to the near-Asia ratios: the mean of our C-130 TAS ratios was 0.23, with a maximum of 0.64. The average MLO ratio was 0.36 for samples with soluble Ca > 0.1 μm/sm3, which is 50% larger than the mean leaving Asia. To the extent that these are real differences (as opposed to sampling or statistical artifacts), this suggests that either additional NOy was taken up or that less modified (larger?) particles were preferentially removed en route. Although the former seems likely because of HNO3 from sources such as lightning and the stratosphere, we cannot rule out the latter possibility because nitrate/calcium ratios sometimes decreased slightly toward the largest sizes (Figure 10a).

[54] Another impact of nitrate's interaction with dust is that the dust becomes more hygroscopic as it adds soluble ions. This aging phenomenon increases f(RH), the increase of aerosol light scattering with humidity, as the aerosol moves farther from its source (S. G. Howell et al., unpublished manuscript, 2004). Still, the coarse f(RH) is unlikely to become as large as f(RH) for the accumulation mode, because much of the dust mass would still be insoluble. Furthermore, if the soluble calcium is unevenly distributed on dust particles, there may well be both low-calcium particles for which the f(RH) is invariant and high-calcium ones for which nitrate formation does increase f(RH).

[55] The nitrate/soluble calcium ratio is a property for which surface observations are clearly not representative of the column above the surface. The median molar NO3/Ca ratio was 1.61 in the lower BL {almost entirely Ca(NO3)2}, 1.03 in the upper BL and lower FT {about half of soluble Ca was Ca(NO3)2}, but only 0.18 in the mid-FT. Thus dust would be considerably more hygroscopic in the BL than the FT, in part because of the higher NO3/Ca ratio in the BL. Models based solely on surface-measured f(RH) for dust would therefore overestimate its direct and indirect radiative impacts in the FT. However, in view of the low RH in these FT dust layers, f(RH) may be largely irrelevant (∼1.0) until the dust either settles into moister layers or subsides into the MBL.

5.3. Sulfate Chemistry

[56] Sulfate distributions with size and altitude are interesting in part because they inform us about the relative importance of several processes that remove SO2. The uptake of SO2 by large dust particles depletes SO2 that might have otherwise have nucleated new sulfate particles after homogeneous oxidation by OH. Since the control of particle number is at the heart of the indirect effect of sulfate aerosols [Charlson et al., 1987], any competition for SO2 by other processes is of great importance. Since the uptake of SO2 by alkaline dust particles produces mostly supermicron NSS, our data can be used to derive limits on those processes.

[57] Usher et al. [2002] comment on the uptake of sulfate by mineral dust particles. In a model study, Dentener et al. [1996] noted that mineral aerosols can have a significant impact on the chemistry of the troposphere. These model calculations suggested that in east Asia over 10% of the sulfate is associated with mineral aerosol surfaces, while at least 40% of total nitrate may be associated with dust. Our measurements confirm this.

[58] It is clear that the coarse fraction of NSS is usually much smaller than that of nitrate (Figure 7). In the BL the NSS coarse fraction median was 10%, but it rose to 35% in the FT (Table 7). This could be because there is more calcium relative to nitrate and NSS in the FT: The bulk nitrate/calcium and NSS/calcium molar ratios in the BL averaged 1.83 and 2.12, respectively (Table 5), which implies that some other cation(s) were present. Hydrogen ions and ammonium are the likely candidates. In the FT, by contrast, molar ratios averaged 0.23 and 0.33, suggesting that calcium alone could have been the sole cation for both the nitrate and NSS. Those anion/calcium profiles are in part due to the altitude profile for calcium, which had less of a gradient than either nitrate or NSS. This could be because the soluble ions and their precursors are more likely than mineral dust to be scavenged during convective lifting. Furthermore, the dust often originates in very dry areas, while in the coastal urban/industrial regions the humidity (and thus the probability of pollutant washout during lifting) is much higher.

[59] The relationship between NSS and calcium is much different from that of nitrate and calcium: High soluble calcium does seem to suppress submicron NSS (Figure 15b). Also, with a few exceptions the amount of coarse NSS is well correlated with the amount of soluble calcium, with a slope of about 0.1 mol NSS/mol Ca. The relationship between coarse NSS and soluble calcium is much tighter than that for nitrate. (The most notable exception (low calcium and 12 μg coarse NSS/sm3 in Figure 15b) was in the Shanghai plume in very moist air, a sample in which coarse ammonium was roughly equal to coarse NSS on a molar basis. Here water caused small NSS and ammonium particles to grow larger than 0.77 μm.)

[60] Interestingly, at MLO in the dusty samples (Figure 16) the NSS/Ca molar ratio averaged 0.9, considerably greater than for FT samples near the Asian coast. This suggests that there is substantial aging of dust during long-range transport. How much of this is reaction with new SO2 versus selective removal of some particles remains to be determined. The amount of unreacted SO2 measured during the vast majority of the MOI samples would have been enough to fully neutralize the Ca in those samples with NSS. Even in the relatively unpolluted FT precursor gases are continually being oxidized to form secondary nitrate and sulfate. It is clear that for nitrate (Figure 16b) dust does change the gas/aerosol partitioning at MLO significantly, but without SO2 measurements at MLO we cannot produce a similar plot for sulfate.

[61] The percentage of NSS in the coarse mode also seems to be controlled by soluble calcium (Figure 17). Above about 8 μg coarse Ca/sm3, the NSS coarse fraction is always above 30%, although there are very few data points to support this conclusion. It seems to level out at about 50% for the highest calcium concentrations, although the statistics are limited at those high levels. Bahreini et al. [2003] reported that the peak vacuum aerodynamic diameter of the submicron sulfate mode in BL air was around 400–500 nm (250–300 nm geometric diameter), but our MOI unfortunately cannot supply comparable information about the peak size of coarse-mode sulfate.

Figure 17.

Coarse fraction of NSS versus coarse soluble calcium.

[62] It is possible that some sulfate could be from gypsum, a hydrated calcium sulfate mineral that is sometimes found in dust. If there were significant amounts of this primary sulfate, it could mislead modelers who may try to account for all NSS as a product SO2 oxidation reactions. In bulk filter samples we collected at Zhenbeitai in Shaanxi Province, China [Arimoto et al., 2004], we found that the ratio of sulfate to soluble Ca in the dustiest samples (for which the influence of pollution should be minimized) was 0.09 ± 0.04. Thus, in this western dust the maximum possible moles of primary sulfate was 10% of the moles of soluble calcium. Interestingly, in our C-130 samples the minimum molar ratio of NSS/Ca was also 0.1 (Table 6). This suggests that any coarse sulfate in excess of 10% of calcium was almost certainly secondary, from gas to particle conversion reactions.

[63] NSS does not decrease as rapidly with altitude as does nitrate. The nitrate/NSS molar ratio in the BL averaged about 1.0, while in the FT it dropped to about 0.7. This may reflect the more rapid oxidation of NOx to HNO3 (relative to SO2 oxidation to NSS), thus putting the nitrate into its most soluble form more quickly. The modest solubility of SO2 means that it might be less efficiently scavenged during moist transport into the FT, although rapid in-cloud oxidation by peroxide could reduce this difference. Bahreini et al. [2003] used an Aerodyne AMS to measure vertical profiles of sulfate during ACE-Asia from the CIRPAS Twin Otter. They reported that sulfate concentrations (and organics) often peaked in layers a few hundred meters thick, sandwiched between layers of cleaner air. Thus sulfate does form distinct layers like the light scattering and SO2 shown in Figure 13.

[64] The molar ratio of total NSS to soluble Ca is another property for which surface measurements do not represent the FT. The BL mean (2.12) was almost an order of magnitude larger than that in the FT (0.26). Also, the coarse fraction of NSS was 2–3 times greater in the FT than the BL. Presumably both the smaller FT ratio of NSS to soluble Ca (Table 6) and the lack of FT ammonium (Table 5) direct a larger fraction of FT than BL NSS into the coarse mode. As Figure 13 demonstrates, at times surface measurements would not even represent the MBL adequately.

5.4. Ammonium Chemistry

[65] The C–130 data is valuable in addressing the role of atmospheric bases in the neutralization of sulfuric and nitric acids. Molar ratios of NH4+ to NSS sulfate averaged 1.7 for the TAS data and agreed quite well with that from the MOI samples, which averaged 1.9. Aircraft measurements by Bahreini et al. [2003] during the same experiment estimated that during sample legs influenced by Korean emissions, the molar ratio of NH4+/SO4−2 was 1.9. Although the MOI ratios stayed relatively constant with altitude, the TAS ratios varied slightly being 1.68 below 500 m, 2.12 between 500 and 2500 m, and 0.87 above 2500 m.

[66] A greater neutralization ratio in the MBL than the FT agrees with the hypothesis of Huebert et al. [1998] that since the source of NH3 is at the surface, there should be more acid neutralization in the BL. Although the 1998 study falsified this hypothesis for the Southern Ocean, it appears to be valid in eastern Asia: Neutralization of NSS by NH3 in the FT was about half of that in the BL. Figure 18a shows molar concentrations of total NH4+ versus those of NSS sulfate in the fine mode (less than 0.77 μm aerodynamic diameter) from MOI measurements. The data fits a line with slope 2.1 and an R2 of 0.91, agreeing well with the expected molar ratio of neutralization via formation of ammonium sulfate. The grey scale shows the fraction of NH4+ in the fine mode, which, for the majority of the samples, is very close to one. PILS measurements reported by Lee et al. [2003] also found that NH3 was sufficient to completely neutralize submicron SO42− throughout the experiment.

Figure 18.

Neutralization plots of total ammonium versus (a) fine (<0.77 μm aerodynamic diameter) nss sulfate with grey shades showing the fraction of fine NH4+ and (b) total nss sulfate with grey shades showing the concentration of soluble calcium. The dashed line indicates a slope of 2, or complete neutralization of nss sulfate via formation of ammonium sulfate.

[67] There were only 3 of the 60 MOI samples where the fine fraction of NH4+ was less than 0.69, all in the Shanghai plume. The very high RH during this sampling may have caused submicron ammonium sulfate to grow larger than 0.77 μm. Since it appears as though most of the ammonium was being used to neutralize nss sulfate in the fine mode, it is reasonable to assume that most nitrate was neutralized by calcium, as discussed above. In the few high dust cases where total NSS sulfate is in excess of the neutralizing NH4+ (Figure 18b), the excess NSS is no doubt on the coarse mode, having reacted with calcium (Figure 15b). In Figure 18b the molar concentration of NH4+ is plotted against total NSS, with soluble calcium displayed as the grey scale. Not surprisingly, the samples with the lowest NH4+/NSS ratios mostly have high soluble calcium concentrations.

[68] The NH4+/NSS ratio is yet another intensive variable for which BL measurements do not represent the FT values. The BL ratios were almost always close to 2 (fully neutralized to ammonium sulfate), while those in the FT were closer to 1 (half-neutralized, to ammonium bisulfate). Models need to reproduce this difference to correctly describe the f(RH) and climatic impact of sulfate in the FT.

5.5. Soluble Calcium as a Proxy for Dust

[69] Unfortunately, there was no gravimetric measure of dust mass on the C-130. We did measure soluble calcium, however, which might be the best proxy we had for dust mass. To test this proxy for our C-130 data, we compare soluble calcium to coarse-mode mass estimated from APS data. Figure 19 is a plot of the concentrations of total soluble calcium as measured by the TAS and the coarse soluble calcium measured by the MOI versus the APS super-0.77 μm mass (assuming a density of 2.0 g/mL). We note that the APS mass is probably an underestimate, because of the nonsphericity of dust particles. TAS calcium is linearly related to coarse APS mass with a slope of 0.058 and an R2 value of 0.92. We do not have MOI data during two highest-dust TAS samples. MOI super-0.77 μm calcium has a weaker linear relationship to APS super-0.77 μm mass, with a slope of 0.077 and an R2 value of 0.54. This implies that soluble calcium represents between 5% and 8% of the coarse mass, in line with the ZBT data of Arimoto et al. [2004]. The nonsphericity impact on the APS data makes the lower end of that range more likely (closer to 5%). The insoluble fraction of Ca that Arimoto et al. found at Gosan may well be from fly ash or industrial sources, which would not change the relationship between soluble Ca and dust mass. We conclude that soluble calcium can be cautiously used as a proxy for dust mass, although the uncertainty in this 5% figure is at least a factor of 2.

Figure 19.

Concentration of calcium as measured by the TAS and MOI (super 0.77 μm) compared to the coarse-mode mass estimated from APS size distributions. As calcium is an indicator of dust, these should be related, with the slope being the percent soluble calcium in the dust sample. There are limitations, however, in that we have total calcium as measured by the TAS, not size-segregated, and the corrections associated with MOI calcium concentrations involve many assumptions. (Note log scaling.)

5.6. Airborne Aerosol Measurements

[70] No doubt some readers will be alarmed to learn how tenuous airborne measurements of supermicron particles can be. Multiplying apparent values on the largest MOI stage by a factor of 3 (or 10) naturally makes one uncomfortable, even though the resulting total ion concentrations are mostly within ten or twenty percent of the TAS totals. While even well-designed transport systems will introduce losses in the tens of percent (Figure 2), our largest corrections were largely from problems with the MOI design, which can be solved.

[71] With submicron particles, one can fairly readily get related measurements to agree to within tens of percent. Small particles can be moved through tubing with only modest losses. However, to study supermicron particles like sea salt or dust, it is a whole different matter: Inertia is significant. Turbulence and bends can throw big particles to walls, from which they may or may not bounce. Enhancement also becomes a real possibility in splitters and inlets. Conveying supermicron aerosols from ambient air into an instrument (especially a rapidly moving one) is a very difficult problem.

[72] We have evidence of progress, though, and reasons for optimism. The TAS reference sampler allows us to test the efficiency of other devices for even the largest particles and to correct some types of data accordingly. The LTI airborne aerosol inlet has a nonunity-but-calculable efficiency versus size. Future programs will no doubt minimize transmission line uncertainties by shortening and simplifying the tubing to critical sensors. MOIs will be rebuilt so that sharp bends and bounce are managed more effectively. Although the corrections are large and the caveats numerous, the ensemble of measurements is beginning to build a consistent picture.

[73] Sun photometry also provides an invaluable constraint. The AOD difference between the top and the bottom of an aerosol layer (a measurement that does not involve inlets) has been compared with scattering and absorption measured on particles that were conveyed from ambient to instrument using inlets, splitters, and tubing. The results are not half bad: Redemann et al. [2003] and Schmid et al. [2003] find agreement at the 12% or better level between in situ and remotely sensed measurements of light extinction. For samples whose fine-mode fraction of scattering exceeded 60% (i.e., mostly small particles), the in situ versus AATS-6 slope was 1.01, with an R2 of 0.81. The submicron measurements are highly consistent and the supermicron measurements are quite respectable.

[74] Although we sometimes apply large correction factors, the bottom line aerosol products seem to be reliable to tens of percent. In situations where more than half of the extinction is due to supermicron particles, that's reason for encouragement.

6. Conclusions

[75] Perhaps our most obvious conclusion is that the technology for sampling and conveying supermicron particles to instruments is far from mature. Enhancements, losses, and varying degrees of bounce from walls are inevitable for particles larger than a few microns. However, careful laboratory calibrations, the use of complimentary instruments and reference samplers, and the modeling of transport efficiencies can together produce defendable data sets. Some earlier airborne aerosol data sets may warrant reexamination, in view of what we now know about sampling large particles.

[76] Even our raw uncorrected MOI data contains valid information about what ions are associated with what other ions in the coarse and fine size ranges. Nitrate, for instance, was usually coupled to calcium rather than ammonium. Sulfate usually was mostly in the fine mode with ammonium, although up to half could be found in the coarse mode in the presence of high dust concentrations. The raw ion ratios on coarse and fine stages provide a defendable picture of the composition of particles in each size range, even though they do not necessarily represent ambient concentrations faithfully. The corrected data, which are constrained by the total aerosol sampler and include LTI and tubing efficiency corrections, provide a reasonable picture of the ambient concentrations of each ion in submicron and supermicron modes, although they probably misrepresent the shape and peak diameter of the coarse mode. Those can be derived from APS and OPC data.

[77] A comparison of soluble calcium concentrations and APS data suggest that the total dust mass during ACE-Asia can be approximated by assuming soluble calcium is 5–8% of the total mass. This is consistent with values measured simultaneously at Zhenbeitai closer to the dust sources, but may not be applicable to all dust source regions and times.

[78] Ion ratios were in general very different in the free troposphere and the boundary layer, which implies that surface measurements cannot represent FT chemistry accurately. Nitrate/soluble calcium ratios varied from a median of 1.6 in the BL to 0.2 in the FT. NSS/soluble calcium had a median of 1.4 in the BL that dropped to 0.3 in the FT. The ammonium/NSS neutralization ratio median was 1.7 in the BL but just 1.2 in the FT. Nitrate and ammonium decreased more rapidly from the BL to the FT than did NSS and soluble calcium.

[79] Fine-mode fractions (percent of total in particles smaller than 0.77 μm) also changed with altitude. Median FT and BL fine-mode percentages were 26/11 for nitrate (mostly coarse), 91/94 for ammonium (all fine), 13/7 for soluble calcium (mostly coarse), and 65/90 for NSS (fine with up to half coarse). The inverse of the latter may be more informative: Only 10% of NSS is coarse in the BL, but 35% is coarse in the FT. This coarse fraction of NSS was found to depend strongly on soluble calcium (no doubt because of the absorption of SO2 by CaCO3). In contrast, while most nitrate was on the coarse mode with a size distribution nearly identical to that of calcium, the fine- and coarse-mode nitrate concentrations were not strongly correlated with soluble calcium concentrations.

[80] Aerosols in the Asian atmosphere were found to be highly variable, with distinct layers of very concentrated dust and pollution aerosols interspersed with much cleaner air. This dramatic variability suggests that average concentrations cannot represent the reality of aerosols in Asian outflow. Likewise surface measurements cannot represent composition in the upper boundary layer or the free troposphere.

Appendix A:: APS Plumbing

[81] Upon entering the fuselage via a 60 degree curved tube behind the starboard LTI, the air was split between front (59 L/min) and rear (77 L/min) instrument racks. The first split angle was approximately 30 degrees with the upper arm going to rear rack and the lower to the front rack. The rear rack air, some of which went to the APS, passed through a stainless steel bellows piece designed to accommodate fuselage expansion during pressurized flight. This piece was approximately 13 cm in length and had a curvature of about 45 degrees. Following the bellows was a 122 cm straight stainless steel section that transitioned into a 23 cm straight stainless steel section. A 23 cm 30 degree curved stainless steel piece connected the straight sections to a manifold, which contained a ball valve and two Ts that allowed for blank sampling. During a sample, air passed straight through the 32-cm-long, 2.2-cm-diameter manifold into a 3-way block splitter. The block splitter inlet had a diameter of 2.2 cm that transitioned into two 1.75 cm and one 1.11 cm diameter exits. The length of each path through the splitter was approximately 13 cm. Thirty L/min of air was pulled through the identical right and left arms of this splitter through 1.9 cm conductive tubing to the University of Washington (UW) nephelometers. Through the third, center, arm of the block splitter, 17 L/min of air was pulled into a small 2-way (vertically orientated) splitter with an inlet of 1.11 cm and exits of 0.79 cm. The path within the splitter was approximately 9.5 cm. On the aircraft, 12 L/min was pulled through the upper arm of the split via 1.27 cm conductive tubing to the Radiance Research (RR) nephelometer and 5 L/min was pulled through the lower arm to the APS via a length of 1.27 cm ID conductive tubing. Of the 5 L/min total APS flow, 4 L/min was used as sheath air and 1 L/min for sample flow.


[82] The authors are particularly grateful to the staff of NCAR's Research Aviation Facility, who frequently went above and beyond to make these measurements successful. We are also grateful to Jack Fox and the NCAR Design and Fabrication Shop for building instruments such as TAS and for helping us integrate all our systems with the C-130. We wish to thank Karyn Sawyer and the staff of UCAR's Joint Office of Science Support for excellent logistical support. We thank Tad Anderson and Sarah Doherty for the use of light-scattering data. This work was supported by NSF grant ATM00-02698 and amendments thereto. This research is a contribution to the International Global Atmospheric Chemistry core project of the International Geosphere-Biosphere Program. This is SOEST contribution 6329.