Journal of Geophysical Research: Atmospheres

Measurements of organic and elemental carbon in Asian outflow during ACE-Asia from the NSF/NCAR C-130

Authors


Abstract

[1] During the spring 2001 Asian Pacific Regional Aerosol Characterization Experiment (ACE-Asia) intensive observation period, we measured elemental and organic carbon (EC and OC) from the National Science Foundation/National Center for Atmospheric Research (NSF/NCAR) C-130 aircraft. Samples were collected on quartz filters using a Particle Concentrator–Brigham Young University Organic Sampling System (PC-BOSS) to minimize positive and negative artifacts and analyzed using a Sunset Labs thermal/optical analyzer. Although short aircraft sampling times limited our sensitivity, total carbon (EC plus OC) was quantifiable on all but one level flight leg. We could not confidently apportion total carbon (TC) between EC and OC on about a quarter of the legs, mostly in the free troposphere (FT). About a third of the samples showed significant evaporation of collected aerosols from their quartz filter (a median of 20% on these samples), while a similar number had too little evaporation to detect. We found that TC and OC concentrations were generally higher in the boundary layer (medians of 7.6 μg TC/m3 and 5.8 μg OC/m3) than in the free troposphere (medians of 3.1 μg TC/m3 and 3.9 μg OC/m3, the latter from only a few samples). The same appears to be true for EC, but the free troposphere (FT) statistics for EC are also poor. Average concentrations were somewhat higher than the medians, reflecting the impact of a few more polluted samples. OC was much more variable than was EC. In the FT the TC to non-sea-salt sulfate ratio ranged from 3.2 to 6.0, so carbonaceous aerosols were considerably more concentrated than non-sea-salt sulfate in the upper troposphere. When dust mass was included, organic matter ranged from about 0.1 to 0.7 of the fine particle mass. TC/EC ratios ranged from 2 to 15, with medians of ∼4–5. Air in the Yellow Sea, the Korea Strait, and the Sea of Japan had generally higher concentrations of OC than that in the East China Sea or the Pacific south and east of Japan.

1. Introduction

[2] There are very few measurements against which to test global climate model predictions of total particulate organic and elemental carbon abundances and their vertical/spatial distribution. In part, this is due to the difficulty of making the measurements: The uncertainties and biases of current measurement techniques are poorly constrained [Huebert and Charlson, 2000; Lim et al., 2003; Chow et al., 2001]. Positive and negative sampling artifacts are common results of vapor adsorption and semivolatile compound evaporation during sampling. Analytical methods necessarily depend on operational definitions of OC and EC, which can produce results that relate poorly to the physical and optical properties of the carbonaceous material. Additional difficulties result from the peculiarities of airborne sampling, among which was our desire to keep sampling times to an hour or less to avoid averaging over multiple air masses and altitudes.

[3] These carbonaceous aerosols can have a large impact on the radiative forcing of climate. EC is among the most strongly absorbing substances in the atmosphere [Clarke et al., 1984; Horvath, 1993], which can cause the heating of elevated aerosol layers and the inhibition of rainfall [Ramanathan et al., 2001]. OC has the potential to scatter light directly and to reduce the hygroscopicity of aerosols, thus changing the way their light scattering varies with RH (f(RH)) [Carrico et al., 2003]. It has been suggested that hydrophobic organic films could slow or stop the activation of aerosols to form cloud droplets, thus changing the indirect effect of aerosols on climate [Facchini et al., 1999]. These impacts depend strongly on the internal or external mixing state of OC with sulfate and other aerosol constituents.

[4] Early models considered the FT aerosol to be almost entirely sulfates. However, Novakov et al. [1997] reported that off the east coast of the United States during the TARFOX campaign there was as much organic aerosol mass (OM) as non-sea-salt sulfate (NSS). Putaud et al. [2000] came to a similar conclusion looking at European outflow from the Canary Islands during the Aerosol Characterization Experiment 2 (ACE-2) experiment. While both of these conclusions were weakened by the potential for sampling and analytical artifacts [Huebert and Charlson, 2000], it is clear that (1) the composition of FT aerosols includes more than just sulfate and (2) defendable measurements of OC above the boundary layer are sorely needed. Among the reasons for the paucity of free tropospheric OC and EC observations are that the concentrations there are usually small and airborne sampling times are of necessity a few hours or less. Half a day or longer sampling times have typically been required to measure small concentrations of OC and EC.

[5] There is unfortunately no standard technique for sampling or analyzing EC and OC [Hering et al., 1990; Huntzicker et al., 1982; Jacobson et al., 2000; Lim et al., 2003; Huebert and Charlson, 2000; Turpin et al., 2000]. The R&P analyzer used by M. Uematsu [Rupprecht et al., 1995] collects particles with a heated impactor, which may undersample semivolatile OC and small-diameter OC and EC [Anderson et al., 2002]. Many other ACE-Asia groups used quartz filters to collect carbonaceous particles [Mader et al., 2002, 2003; Quinn et al., 2004; Chuang et al., 2003], some with and some without devices to limit gas adsorption (positive, McDow and Huntzicker [1990]) and evaporative (negative, Lewtas et al. [2001]) artifacts.

[6] The analysis of the collected material also can be done in several ways, most of which involve heating to set temperatures in the presence of air or other gases [Schauer et al., 2003; Chow et al., 2001; Rupprecht et al., 1995]. Both OC and EC are evolved in this way (and then converted to CH4 or CO2 for quantification), but since they lose their identity upon volatilization, there is considerable ambiguity about how much of the evolved gas was originally OC and how much was EC. The R&P analyzers assume that all carbon evolved below 340°C was OC and that evolved above was EC [Rupprecht et al., 1995], while the Sunset Labs analyzers use light transmission to decide when the OC is gone and EC is finally evolving [Huntzicker et al., 1982]. The net effect of these sampling and analysis issues is that the various EC and OC data sets may not be comparable.

[7] The ACE-Asia intensive field campaign was conducted during the spring of 2001, as part of the IGAC series of Aerosol Characterization Experiments [Huebert et al., 2003]. The objectives of ACE-Asia included the characterization of aerosols in Asian outflow, both at the surface and higher altitudes. We report here on our measurements of ambient carbonaceous aerosol particles from the National Science Foundation/National Center for Atmospheric Research (NSF/NCAR) C-130 aircraft, which was based at the Marine Corps Air Station (MCAS) in Iwakuni, Japan. We discuss the impact of sampling and analytical issues on the values we report.

2. Experimental Methods

2.1. Sample Collection

[8] Quartz filters are commonly used to collect OC and EC for analysis, because they can be baked before use at high temperatures to remove any carbon added during manufacture or handling. This is necessary because the evolved gas analysis methods are nonselective: Upon heating, it is not possible to distinguish carbon that was originally an ambient aerosol from contamination during handling and adsorbed carbon-containing gases. We exposed our quartz filters in a Particle Concentrator–Brigham Young University Organic Sampling System (PC-BOSS) (Ding et al. [2002]; Figure 1) because (1) it preconcentrated the aerosols, allowing us to get analyzable amounts of sample on 20 to 30 min aircraft sampling legs; (2) it employed a diffusion denuder to strip out organic vapors that might otherwise cause a positive artifact, and (3) it included a side filter that could be used to support quality assurance tests on the samples.

Figure 1.

Schematic of the PC-BOSS sampling system flown on the C-130 during the ACE-Asia field campaign.

[9] Samples were collected on each of 19 research flights. Integrating a sample over 20 min in the C-130 implies averaging over at least 120 km horizontally, so there can be considerable variability in concentrations during one of our samples. This is illustrated in Figure 2, which depicts the sample integration times during RF15. Even though we limited our samples to nominally constant-altitude legs, there was significant variation in RH, aerosol number concentration, and PSAP light absorption during many legs. Typically, eight samples were exposed during each research flight. We simultaneously measured anions and cations on many of these legs [Kline et al., 2004], enabling us to compute OC/sulfate ratios and to estimate the mass of mineral dust from soluble Ca.

Figure 2.

(a–c) Carbonaceous aerosol filter sampling during research flight 15 (27 April 2001); integrated samples were taken during time frames indicated with bold lines. Relative humidity and particle absorption (σa0) indicated the variability of the air mass during each integrated sampling period.

[10] Air was conveyed into the aircraft by a solid diffuser inlet. The inlet tip is characterized by a double elliptical leading edge, designed to reduce flow separation. The tip inside diameter was 5.38 mm. The curved tube that conveyed the flow into the fuselage had an ID of 31.8 mm and a radius of curvature of 61.0 cm, and entered the fuselage at a 65-degree angle. The exact passing efficiency of this specific solid diffuser inlet has not been determined. However, it is similar to the solid diffuser inlet tested during the PELTI experiment [Huebert et al., 2004], so we expect the inlet to pass particles below 2 microns with high efficiency. One PELTI conclusion is that under some conditions a fraction of large dust particles may not be retained when they impact on inlet and tubing walls, so that they might also reach the PC-BOSS.

[11] Once inside the fuselage, air was conveyed by a 2.54 cm ID tube to a flow splitter where a minor flow (12 Lpm) was directed to a side filter, while the major flow (135 Lpm) passed into a particle preconcentrator (Figure 1; Ding et al. [2002]) designed to achieve a fourfold enhancement in concentration for particles larger than 0.1 μm. Following concentration, the enriched flow passed through a parallel plate diffusion denuder to remove gaseous organics (VOC). The diffusion denuder consisted of 15 parallel 4.5 × 58 cm strips of carbon-impregnated glass fiber (CIG) filters separated by 2 mm [Eatough et al., 1999]. Ding et al. [2002] estimate the denuder breakthrough to be less than 1% for organic gases, while EC and sulfate passing efficiencies were greater than 96%. This does not completely eliminate the potential for artifacts, but we configured our data analysis to exclude a low-temperature peak that should include the bulk of any VOC not removed by the denuder.

[12] Particles were collected on baked quartz filters (Pall Gelman 2500QAO, 47 mm), prebaked 16 hours at 550°C in accordance to the handling method described by L. Salmon, on 6 October 1994 (R. Flagan, personal communication, 2000). Because of the temperature increase as OC moved from ambient air to the C-130 cabin, some semivolatile species may evaporate from the sample filter. (Our short sampling times made it unlikely that heavy loading would cause breakthrough.) Therefore sample filters were backed by CIG filters (carbon-impregnated glass fiber, S&S GF3649, 47 mm, prebaked 10 hours at 325°C under N2) to collect evaporated organic carbon volatilized from the condensed phase. Filter samples were masked to expose a 20 mm × 28.5 mm region, to further concentrate the analyte on the filter.

[13] During each research flight we exposed a Teflon filter in the “side-all” position (Figure 1) during one sampling leg to compare the sulfate concentrations before the particle-concentrator (PC) with those on the sample filter behind the PC. Since SO2 is removed with unit efficiency by the CIG denuder, the sulfate concentration derived from the Teflon side filter should equal that derived from the quartz sample filter, when corrected for the PC concentration factor. During the remaining sampling legs we exposed a single quartz filter in the side filter position to compare its EC value with EC from the sum of all the remaining sample filters. These allowed us to directly determine the actual PC concentration factor for each flight. Unfortunately, comparison of side filter analyte with sample filter analyte (for both EC and NSS) showed that the PC concentration factor was closer to 2 than the design value of 4.

[14] To understand why, we made monodisperse particles in the laboratory and measured the PC efficiency versus size. Submicron particles were generated by mixing dry sheath air with NaCl particles generated by bubbling a dissolved NaCl solution and then selecting sizes (0.7, 0.5, 0.3 and 0.1 μm in diameter) with a Differential Mobility Analyzer (DMA). Supermicron oleic acid particles tagged with Rhodamine-B (1.2 and 3 μm) were generated using a vibrating orifice aerosol generator (VOAG). The results (Figure 3) confirm that our PC efficiency was far from ideal. Visual inspection also revealed that most of the pink rhodamine particles were being deposited on one edge of the receiving slot. A misalignment (too small to detect visually) caused this virtual impactor to behave as a regular impactor for part of its flow. We have used the measured efficiency from the side filters to correct for this nonideal behavior.

Figure 3.

Concentration factors for the PC-BOSS aerosol preconcentrator as a function of aerodynamic particle diameter (Dp). The Ding et al. [2002] data are the ideal behavior, whereas the crosses are our laboratory-measured efficiencies.

2.2. Sample Analysis

[15] Collected samples were returned to the laboratory (in Iwakuni) and analyzed using a Sunset Labs thermal/optical transmittance (TOT) analyzer. The temperature program was slightly different from the NIOSH routine used in other Sunset Labs analyzers during ACE-Asia [Schauer et al., 2003], but intercomparisons and our own tests suggest this did not significantly change our results. A 1.45 cm2 punch of each quartz filter was step-ramp heated to 870°C in a quartz oven under helium to volatilize the OC. The resulting gas phase material was then oxidized to CO2 over a manganese dioxide catalyst, reduced to CH4 over a hydrogen enriched metal catalyst, and quantified by a flame ionization detector (FID). Following the initial heating, the sample was cooled to 550°C before being heated to 900°C in a He:O2 environment. During this phase any pyrolized OC was oxidized and volatilized along with the EC.

[16] During the course of the run, filter light transmission was monitored using a HeNe laser, to measure the charring of OC and then define the split point between organic and elemental carbon. Figure 4a shows both the FID signal and laser transmission alongside oven temperature for a typical ambient sample and its associated field blank. The split point is defined as the time at which the laser attenuation returns to its initial value. Carbon evolved prior to the split point is presumed to equal pyrolized OC (char), while carbon evolved after the split point is assumed to represent native EC. Since the laser transmission varied strongly with oven temperature even for a blank filter (top dashed line in Figure 4a), we used the difference between the sample and blank laser signals (Figure 4b) to minimize the impact of temperature on the determination of the split point. Additionally, if the absorbance by EC was enhanced by NSS or OC coatings that burn off in the early heating steps, that could cause an overestimate of EC by making the apparent split point earlier than the real one.

Figure 4.

(a) Thermal analysis of a quartz filter sample, taken during research flight 15, using the Sunset Labs thermal-optical analyzer. Note that the blank and signal laser signals vary because of temperature changes, which contributes uncertainty to the split point determination. (b) Same as Figure 4a, but split point from difference between sample and blank laser signals. Note the flattening of the initial laser signal.

[17] Carbonates generally decompose at the highest He-only temperature step. When we saw significant C evolution at that temperature, we exposed an aliquot of the filter to HCl vapor overnight (to drive off the carbonate) and then reanalyzed it. We did not find a useable relationship between this measure of carbonate and soluble Ca, although the extremes were logical (the maxima and minima were coincident).

[18] Field blanks were handled in the same fashion as sample filters, including placement in the sampler and the passage of air: prebaked, loaded into filter holders, brought aboard the aircraft, exposed for just a few seconds, and analyzed exactly like the samples. Campaign averages (standard deviations) for total carbon (TC), OC, and EC field blanks were 0.23 (±0.10), 0.13 (±0.07), and 0.09 (±0.07) μg C cm−2, respectively. Typical ambient sample loadings ranged from 0.1 to 10 μg C cm−2, with a campaign average of 2.8 μg C cm−2 for TC, 1.8 μg C cm−2 for OC, and 0.8 μg C cm−2 for EC. The smallest amount of evolved C that we could quantify above instrumental noise in the field (our Sunset Labs instrument-only DL) was about 0.045 ugC/cm2.

[19] A laboratory comparison of identical filter samples was run among eight groups operating Sunset Labs EC/OC instruments in the field. The results indicate interlaboratory precision of 4–13% for OC loadings of 1.0–25 μg C cm−2 and 6–21% for EC loadings of 0.7–8.4 μg C cm−2 [Schauer et al., 2003].

[20] One major source of uncertainty in the TOT analysis is the assignment of the time at which the laser transmission through the filter has returned to the precharring value, so that any carbon released after that time can be assumed to come from EC. We conducted laboratory tests on quartz filter samples collected at Amami Oshima in April of 2003 to assess this uncertainty. We looked carefully at the laser transmission signal and its relationship to temperature of the oven, which introduced a ±4% uncertainty in the transmission value. When we applied this transmission uncertainty to a selection of ACE-Asia thermograms, we found that the EC portion of TC could be changed by up to 15%.

[21] We also extracted some of the 2003 Amami Oshima samples with water to minimize charring and found that the char (the difference between the extracted and unextracted samples) burned off over the same time period as did native EC. Since the TOT correction scheme implicitly assumes the char burns off before native EC does, the TOT results are sensitive to the implicit assumption that the specific absorption of native EC and char are similar [Yang and Yu, 2002]. We do not know how to put limits on this assumption, but it is a potentially significant source of error.

[22] Since our unmasked area was only large enough to take two punches from each sample, we were unable to do multiple analyses on each sample. Mader et al. [2002] demonstrated that the precision of the TOT analyses was poorer than our blank analysis had suggested. They found that for multiple punches from the same filter, relative standard deviations (RSD) were greater than 0.5 for samples with filter loadings less than 0.2 μg C cm−2. We therefore use as the uncertainty of each individual blank or sample analysis the larger of either (1) the program-wide standard deviation of our blank values (which is very conservative, since it includes all the flight-to-flight variability in how each batch of filters was handled, how long exposed samples had to wait for analysis, etc., that would not affect any one flight's sample-minus-blank computation) or (2) the apparent analyte times the RSD derived from Mader et al.'s Figure 3.

[23] The absolute accuracy of the OC measurements was determined by calibration with a diluted gravimetric standard of sucrose. Unfortunately, no EC standard was available to constrain the EC values and subsequently verify proper differentiation between EC and OC. Consequently, our derived uncertainty is not a bound of the absolute accuracy of elemental carbon, but simply a reflection of the precision of our method.

[24] Two distinct sets of uncertainties need to be considered. The first group includes only those factors that affect the detectability of EC or OC above the blank value. This includes the uncertainty in the μg C cm−2 of analyte on the sample filter and the blank filter, as well as the limit of detection of the analysis method. Since many of our samples were lightly loaded because of the short sampling time on their flight legs, it is the uncertainty in these three factors (relative to the difference between sample and blank analyte) that determines whether we collected enough sample to derive a concentration. The uncertainty of the above-blank analyte includes the blank uncertainty (the larger of the analytes times Mader et al.'s [2002] RSD or the standard deviation of all blanks), the sample uncertainty (done in the same way), and twice the LOD of the instrumental method. The error bars in Figure 5 are these values, to emphasize the certainty of above-blank values.

Figure 5.

Altitude variations of (a) OCq, (b) EC, (c) TCq, and (d) the TCq/EC ratio.

[25] The remaining multiplicative uncertainties, due to flow meter calibrations (±5%), instrument span calibration (±5%), correction for our particle concentrator efficiency (±20%), and the OC-EC split in the analyses (±15%), are propagated into an additional relative uncertainty of 26%. Essentially, this span uncertainty means the whole value and uncertainty could move up or down by 26%, without affecting the issue of whether the sample was significantly above the blank. In the case of ratios like TC/EC (derived from the same analysis of one filter), only the ±15% split-point uncertainty applies, since the same calibrations and flow apply to both numerator and denominator. The TC value does not depend on a split point; we have FT samples for which the TC value is significant even though there was not enough charring to define an OC/EC split point.

[26] CIG filters were also analyzed using the Sunset Labs instrument, employing a 20°C min−1 constant temperature ramp in high-purity helium to a maximum oven temperature of 300°C. The CIG filters were prebaked at 325°C in nitrogen the night before each flight to avoid contamination and stored in a cooler before and after exposure. CIG field blanks ranged from 0.3 to 1.5 μg C cm−2 averaging 0.8 μg C cm−2, while sample CIG filters ranged from 0.3 to 3.9 μg C cm−2, averaging 1.6 μg C cm−2. The volatilized OC ranged from 0 to 2.5 times the OC collected on the quartz filter. Only 11 of 78 CIG samples had a S/N > 2, while 34 had a S/N > 1.

[27] If we assume unit collection of gas phase organics by the diffusion denuder, any collected semivolatile organic aerosols would no longer be in equilibrium with their gas phase so that some would evaporate in an attempt to re-establish the gas/aerosol equilibrium. This would result in a negative artifact on the quartz filter, but the evaporate would be collected by the backup CIG filter. However, if denuder breakthrough occurred, the collection of those residual gas phase organics by the CIG would be misinterpreted as volatilized aerosol OC. Mader et al. [2003] concluded that volatile OC was less than 20% during ACE-Asia, but in a reanalysis of our CIG data since the Mader et al. publication, we have found that 15% of our samples volatilized more than 20% of their OC.

[28] To avoid reducing the number of OC values to the (much smaller) number of CIG detects, we often plot and tabulate “OCq,” the nonvolatile OC from the quartz only. Likewise TCq and EC are solely from the quartz filters. The volatile OC from the CIG is denoted “OCv,” and the OCv + OCq sum is “OCs.” On average, OCs was only 10% larger than OCq, so the OCq statistics should not be misleading.

3. Results

[29] All results are reported as micrograms of analyte per standard cubic meter of air, to remove the effect of altitude changes. We define 101.3 kPa and 298 K as standard conditions, and use sm3 and m3 interchangeably. Ninety-two useable samples were collected on 19 flights on the NCAR C-130 as a part of the ACE-Asia field campaign. Flights lasted 8 to 9 hours, with adequate range to sample a variety of pollution, dust and upper tropospheric aerosol [Huebert et al., 2003]. Most flights were not designed to assess regional average values, but were instead targeted at dust events from the northern Chinese deserts and pollution plumes from China, Korea, and Japan.

[30] We were able to detect TC above its detection limit in all but one of the 92 samples. Twenty-four EC and 16 OC samples were either nondetects (value smaller than its uncertainty) or we could not confidently assign a split point to apportion the TC between them. Most samples were 30 to 60 min duration, at a nominally constant altitude. Denuder shedding was visually evident during parts of the first 4 flights; all contaminated samples were removed from the data set.

3.1. Variations With Altitude

[31] All valid OCq, EC, TCq, and TCq/EC results (detects) are plotted against their mean altitudes in Figure 5. The highest concentrations were all in the BL, where polluted air masses were generally found. The BL-FT difference is most evident in the TC plot, since all but one TCq samples were detects. The BL median of 8 μg TCq/sm3 dropped to 3 μg TCq/sm3 in the FT.

[32] The statistics for three altitude ranges are shown in Table 1. The lowest altitude range is all MBL, the highest range is all FT, while the middle range includes some MBL, some FT, and some isolated layers whose identity is ambiguous. In this table we have included statistics only for the detects (in which the S/N was > 1) unless specified otherwise. Since our analytical uncertainty was greater than the bias introduced by excluding the nondetects in the lower altitude ranges, we did not use nondetects in computations of derived quantities such as OM. The medians and means in the lower two levels changed by 5% or less when the nondetects were excluded. The differences were larger in the FT, however: We could not assign EC or OC values to most of the samples because they had ambiguous OC/EC splits. (In the discussion below we use TC to estimate EC and OC for these FT samples.)

Table 1. Altitude-Stratified Statistics for Carbonaceous Species and Three Ratiosa
 OCqOCvOCsECTCqTCq/ECOCv/OCsOMq/NSS (TCq/NSS)
  • a

    Statistics are given in μg/sm3. OCv and the OCv/OCs ratios are biased high by our rejection of all nondetected OCv values. The concentrations and ratios in parentheses assume zero for below-DL OCv, producing lower bounds. The EC values in brackets were derived from TC, assuming a TC/EC ratio of 4.5. SEM is the standard error of the mean: the standard deviation divided by the square root of the number of values.

Altitude > 2500 m
Median3.9 [2.5]4.7 (0.0)11.81.2 [0.7]3.14.50.66 (0.00)5.8 (4.7)
Mean4.5 [2.7]4.8 (1.5)11.81.2 [0.8]3.24.50.66 (0.22)5.8 (4.8)
SEM0.8 [0.5]1.4 (0.7)3.40.7 [0.1]0.51.80.05 (0.14)na (0.4)
Min2.5 [0.2]1.8 (0.0)8.50.5 [0.1]0.22.70.61 (0.00)5.8 (3.2)
Max7.6 [7.6]9.3 (9.3)15.21.9 [1.9]8.36.40.71 (0.71)5.8 (6.0)
N6 [18]5 (16)22 [18]1822 (6)1 (6)
 
2500 m > Altitude > 500 m
Median4.11.3 (0.0)4.91.15.24.00.24 (0.00)1.5 (1.2)
Mean4.81.3 (0.4)4.91.26.05.20.25 (0.08)2.2 (2.0)
SEM0.70.1 (0.1)0.40.10.70.70.04 (0.03)0.5 (0.5)
Min0.71.0 (0.0)4.10.51.22.60.18 (0.00)0.3 (0.3)
Max14.21.8 (1.8)5.62.215.414.90.33 (0.33)6.7 (7.2)
N215 (18)42024204 (16)14 (14)
 
Altitude < 500 m
Median5.81.3 (1.0)6.81.87.64.10.22 (0.15)1.2 (1.1)
Mean6.41.8 (1.3)7.82.28.44.50.23 (0.17)2.7 (2.6)
SEM0.50.2 (0.2)0.70.50.70.30.02 (0.00)1.2 (1.0)
Min2.30.4 (0.0)2.80.50.72.50.02 (0.00)0.4 (0.4)
Max18.15.7 (5.7)17.47.721.811.90.51 (0.51)37.6 (34.7)
N4932 (44)3146504631 (43)32 (33)

[33] The medians for OCq and TCq (TCq includes 12–16 more valid FT samples than either OCq or EC) decreased with altitude, but by considerably smaller factors (1.5–2.5×) than inorganic anions and cations measured on the same legs (4–10× [Kline et al., 2004]). EC decreased from the lower BL to the middle layer, but the statistics in the FT were inadequate (just two values) to detect a gradient by direct observation. There is not an obvious altitude gradient in OCv, for which the detects increase into the FT while the lower bound decreases into the FT. The averages for most species were larger than the medians (especially in the BL), because of the influence of a few very polluted samples.

[34] The frequency with which concentrations fell into various ranges is shown in Figure 6. It is evident that the EC concentrations were generally lower than OCq, but also that there was a much wider range of variability in OCq. The vast majority of EC values were below 4 μg EC/sm3, with a handful of values as high as 8 μg EC/sm3. By contrast, all OC ranges between 2 and 7 μg OC/sm3 were about equally probable, with values as large as 18 μg OC/sm3. TCq also varied over a wide range. The TCq/EC ratios were mostly confined to the 3–6 range, with means and medians of 4.5 (±10%) in every altitude interval.

Figure 6.

Frequency plots showing the number of samples in various concentration ranges for (a) EC and OCq and (b) the TC/EC ratio and TCq. The number beneath each pair of bars is the upper limit of that range. All 92 valid samples are included, but the number of EC and OCQ values is less since the char/EC split was ambiguous for lightly loaded samples.

3.2. Carbonaceous Aerosols Relative to Sulfate and Total Mass

[35] To compare the amounts of organic aerosol and sulfate, we need to compare measurements in similar size ranges. However, since the PC-BOSS used a solid diffuser (SD) inlet, its OC cutoff is unknown and potentially variable. This SD geometry was tested in the PELTI experiment, where concentrations of particles (smaller than 2–3 μm diameter, in the MBL) using a SD were reduced by 10–20% relative to ambient concentrations [Huebert et al., 2004]. In dry dusty conditions the SD efficiency sometimes approached 100% for hard minerals that could rebound from walls, but was much lower for sticky materials like sea salt.

[36] We compare our OM with NSS derived from micro-orifice impactor (MOI) analyses on the C-130 [Kline et al., 2004], for which we have 54 valid simultaneous PC-BOSS and MOI samples. We added together NSS on all MOI stages up to 1.4 μm (excluding the first two coarse stages). Usually less than 20% of the NSS was on larger sizes. We assumed a value of 1.4 for the OM/OC ratio, even though Russell [2003] has shown it can be considerably larger in a few heavily oxidized samples. She found the population of OM/OC ratios to be bimodal, occasionally quite large (2.4) in aged air but usually much smaller (1.3) in fresher emissions.

[37] Unfortunately, only one of our FT OCq values corresponded to a valid fine NSS sample. Six fine NSS samples were coincident with valid TCq measurements, though, so we plotted (Figure 7a) and tabulated (Table 1) the TCq/NSS ratio to give better statistics on the relative amounts of carbonaceous and sulfate aerosols in the FT. Figure 7 shows that there is indeed more carbonaceous aerosol than sulfate in the FT east of Asia. However, the Asian FT often contained dust in addition to NSS and carbon. When we assumed that soluble Ca was 7% of insoluble mineral dust [Zhang et al., 2003] and added up all the carbonaceous, ionic, and mineral mass, we found that in the FT OM ranged from 16% to 42% of the fine mass (Figure 8a). In most samples there was significantly more mineral mass than NSS. So there was more carbonaceous aerosol than NSS in the fine mode, but carbon was not a majority of the mass except in a few MBL cases.

Figure 7.

(a) Ratio of carbonaceous to sulfate aerosol. TCq was used rather than OCq or OM, because it has the most valid FT samples. If we assume the MBL TCq/EC ratio of 4.5 and an OM/OC ratio of 1.4, the resulting OM/NSS ratio is just 9% larger than the plotted ratio. (b) Relationship between TCq and total soluble calcium.

Figure 8.

Fraction of total fine aerosol that is OM (a) plotted versus altitude and (b) plotted for six air mass light-scattering related categories: 10, coarse (dust) dominated; 15, coarse dominated with high scattering (>60 Mm−1); 20, similar coarse and fine; 25, similar coarse and fine with high scattering; 30, fine (pollution) dominated; 35, fine dominated with high scattering.

[38] The dominance of dust has a profound effect on the fraction of fine aerosol mass that was OM. In Figure 8b the OM as a fraction of fine aerosol mass from the MOI is plotted versus air mass category as defined by Anderson et al. [2003] and discussed further below. It ranged from 14 to 31% when coarse-mode scattering dominated, but rose to 25–57% when fine-mode scattering dominated. OM can dominate fine pollution aerosols, but it is a minor player when the dust mass dominates. When our OM was compared to total aerosol mass derived from the total aerosol sampler (TAS) [Kline et al., 2004], it was only 3–11% in the dust-dominated cases and 6–40% in the pollution-dominated cases.

3.3. Elemental Carbon

[39] EC was not strongly correlated with either OCq or SO2 (Figure 9), which suggests different sources for the three substances. There are no doubt some sources (like diesel engines) that emit both EC and OC, but in different proportions from open biomass burning or residential heating and cooking sources. VOCs (from cooking, petroleum use, or vegetation) can be oxidized to secondary OM without producing any accompanying EC. The major source of SO2 is probably coal burning, which may not be accompanied by much EC in power station stacks but would be in residential coal use. The relationships in Figure 9 will also be affected by conversion rates of SO2 to NSS and VOC to OC, as well as differential removal rates.

Figure 9.

Relationship between EC concentrations and (a) OC and (b) SO2.

3.4. Spatial Variations

[40] Although any location could have polluted or clean air at a given time, we did not find large differences in the mean and median concentrations between the several regional seas (Table 2). For most substances the medians varied by less than a factor of two among the seas. The exception was OCv, for which the lowest median (0.8 μg C/sm3 in the Yellow Sea) was just 1/4 of that in the East China Sea near Shanghai (3.3 μg C/sm3). The highest median OCq (5.9 μg C/sm3) and EC (1.7 μg C/sm3) values were found both near Gosan and in the Yellow Sea, with the same EC median in the Sea of Japan. The highest individual concentrations were also found near Gosan, in air that had recently been over Korea. The lowest median concentrations (other than for OCv) were found over the East China Sea or to the south and east of Japan.

Table 2. Statistics by Regional Seaa
 OCqOCvOCsECTCqTCq/ECOCv/OCsOMq/NSS
  • a

    Excluding nondetects made virtually no difference for concentrations except for OCv and OCs, which exclusion biases high. The OCv/OCs ratio is therefore also biased high: For example, the Gosan median went from 0.06 to 0.24 when nondetect OCv samples were excluded. Concentrations are in μg/sm3.

Yellow Sea, ∼124°E, >33.5°N
Median5.90.86.81.78.04.10.131.1
Mean6.71.08.91.78.45.20.121.7
SEM1.00.31.40.21.00.80.021.1
N1999172017915
 
Gosan and Korea Strait, 125°–130°E, 33°–34°N
Median5.81.68.01.75.64.40.241.5
Mean6.92.29.12.37.34.50.222.0
SEM0.90.51.80.41.00.30.030.4
N1997182718614
 
East China Sea, <130°E, <33°N
Median3.23.37.61.24.23.40.420.54
Mean3.83.28.01.54.83.90.370.57
SEM0.71.13.10.41.00.40.090.08
N83389834
 
Sea of Japan
Median4.51.55.51.76.23.30.241.3
Mean5.52.67.02.06.94.50.312.0
SEM0.70.91.10.30.80.70.070.5
N2010101722171013
 
Pacific South and East of Japan
Median4.11.35.21.24.64.50.273.2
Mean4.32.06.11.14.85.00.318.7
SEM0.30.40.70.20.61.00.035.9
N101091012896

3.5. Air Mass Types

[41] Anderson et al. [2003] defined a classification scheme based on light scattering for quantifying the properties of pollution-dominated versus dust-dominated air masses. They used one TSI nephelometer to measure total light scattering, while a second measured submicron scattering. The fine mode fraction of scattering was used to identify coarse-dominated (dusty, fine-mode fraction of light scattering [FMF] < 0.3), mixed (0.3 < FMF < 0.6), and fine-dominated (polluted, FMF > 0.6) air masses. These were further split into low and high scattering: below or above 60 Mm−1 total scattering. Unfortunately, the light dust category (mostly FT samples) had too few samples to provide meaningful statistics. The lowest EC median (Table 3) was in the low-scattering, mixed category, while the highest median EC was in the high-scattering coarse (very dusty) category. The lowest TCq median was in the low-scattering, coarse category, in which we had only four valid samples. The lowest OCq median was in the mixed low-scattering air masses, but surprisingly the highest TCq and OCq medians were in the high-dust (rather than high-pollution) category. Perhaps this is because some of our most concentrated dust samples were immediately downwind of Beijing and Qingdao, so there was a lot of pollution even though dust dominated the scattering. Statistics derived from three or fewer samples should be used with caution.

Table 3. Statistics by UW Air Mass Categorya
 OCqOCvOCsECTCqTCq/ECOCv/OCsOMq/NSS
  • a

    Excluding nondetects made virtually no difference except for the EC in category 30: The median with all samples was 1.1. The category numbers are used in Figure 8. Concentrations are given in μg/sm3. Here, na, not available.

Category 10: Low Scattering, Mostly Coarse (Light Dust)
Medianna1.8nana1.5nanana
Meanna1.8nana1.7nanana
SEMnananana0.3nanana
Nna1nana4nanana
 
Category 15: High Scattering, Mostly Coarse (Heavy Dust)
Median6.53.413.41.97.65.00.261.8
Mean7.33.712.12.38.55.50.312.5
SEM0.91.21.70.41.10.70.100.5
N2177172417712
 
Category 20: Low Scattering, Comparable Coarse and Fine (Dust and Pollution)
Median3.11.35.21.04.23.00.292.4
Mean3.12.14.91.14.04.50.293.3
SEM0.60.90.40.20.61.20.021.0
N743710735
 
Category 25: High Scattering, Comparable Coarse and Fine (Heavy Dust and Pollution)
Median3.81.36.81.54.83.90.250.8
Mean5.61.610.01.77.24.30.231.0
SEM0.80.42.20.20.90.30.030.2
N2213132024201210
 
Category 30: Low Scattering, Mostly Fine (Pollution)
Median6.11.84.81.76.54.1.311.5
Mean6.11.84.83.17.24.4.311.5
SEM2.10.3na1.70.51.1nana
N421318311
 
Category 35: High Scattering, Mostly Fine (Heavy Pollution)
Median4.91.07.21.66.54.10.171.0
Mean4.51.36.91.77.24.60.201.3
SEM0.50.30.40.20.50.70.030.2
N1812121718171215

4. Discussion

[42] Thermal measurements of OC and EC are controversial, both because of the potential for sampling artifacts and because the analytes are operationally defined by analytical methods. The Sunset Labs NIOSH 5040 TOT method intercomparison by Schauer et al. [2003] helped with the analytical issue, but it did not resolve it. It essentially showed that the OC and EC data of eight groups can be made comparable within reasonable limits, so that time series data at surface sites and measurements from the aircraft can be compared meaningfully: We all measured the same thing.

[43] Exactly what we measured is less certain, because of the operational definitions inherent in the method. What the NIOSH 5040 method defines as EC and OC may, for instance, be quite different from the EC and OC of the Variation of Marine Aerosol Properties (VMAP) Network [Matsumoto et al., 2003]. VMAP used R&P analyzers that collected carbonaceous aerosols on an impactor rather than a quartz filter, heated the sample in air rather than He and He/O2, and used a fixed temperature (340°C) to apportion TC between EC and OC. Schauer et al. [2003] showed that a single method can be applied uniformly by many groups, but not that the NIOSH 5040 results bear any better relationship to the climatic properties of interest (absorption by EC, scattering by EC and OC, and the hygroscopicity of OC) than the results from any other method.

[44] The analyses reported here were often on lightly loaded samples, because of the short (30–60 min) duration of most level flight legs. This presented us with a different set of obstacles from groups that sampled for many hours. Our median sample OC filter had 1.7 μg C/cm2, while our median EC sample was 0.8 μg C/cm2. Both were near the low end of the concentrations used in the Schauer et al. [2003] comparison. We were able to quantify TC for all but one sample, but the most lightly loaded ones (including most of the ones collected above 2500 m) did not produce enough charring to make a clear distinction between OC and EC. Attempts to apportion the evolved C based on our experience with other thermograms produced EC values that were inconsistent with reasonable ranges of EC specific absorbance. Over the last 3 years as we have learned more about analytical variability, PC efficiency, and minimizing temperature effects on the laser transmission signal, we have applied different corrections to our data; the data we report here constitute the fifth set we have posted since collecting the samples in 2001.

[45] As for potential artifacts during sample collection, Mader et al. [2003] summarized the many sampling configurations that the NIOSH 5040 groups used. It is clear from Mader et al.'s analysis that the use of VOC denuders upstream of quartz filters was effective at minimizing positive OC artifacts. However, they used an earlier version (v.2) of the OC and CIG/OCv data that we report here (as v.5), which explains why we now report significantly larger volatilities for a few samples than the 10–20% maximum of Mader et al. Our latest OCv data are more internally consistent, since we excluded thermogram peaks that appeared to have been affected either by VOCs picked up during storage and handling or by OC already on the CIG material prior to use.

[46] We found volatile fractions of OC as large as 71%, which greatly exceeds the values reported by Mader et al. [2003]. What factors influenced this volatile fraction? One of the first things evident in the four plots of Figure 10 is that the two highest-volatility samples were taken at the highest altitudes (3 and 4 km), in the coldest air for which valid OCv and OCq samples were available. They were collected over the central Sea of Japan (Figure 11) in air that had been in the FT for more than 2 days (Figure 12b). Both nitrate (an indicator of urban pollution, Figure 10c) and the fine mode fraction of scattering (which would approach 1.0 in polluted air, Figure 10d) were small, suggesting that the air was not polluted. However, these two samples were heated by 25°–35°C upon entry into the aircraft cabin, which may have promoted the vaporization of their OC. Colder FT temperatures would of course tend to partition more semivolatile OC into the aerosol phase than in the warmer BL. We have too little information to know whether OC is generally more volatile in the FT, but if scavenging during lifting of BL air removes most primary OC, FT OC may have a larger fraction of secondary photochemical products and therefore be more volatile.

Figure 10.

Variation of the volatile fraction of OC [OCv/(OCv + OCq)] with (a) altitude, (b) temperature, (c) total nitrate from TAS, and (d) the fine-mode fraction of light scattering.

Figure 11.

Locations of OC/EC samples, with the size of the X proportional to the fraction of OC that was volatile. Black circles are valid samples for which the OCv loading was too small to quantify, implying that little of the OC was volatile.

Figure 12.

Back trajectories for (a) samples for which almost none of the OC was volatile and (b) samples in which the volatile fraction of OC was greater than 30%. In both cases the altitude of the trajectory is indicated by its color.

[47] Four of the five samples with volatilities in the 40–50% range (Figure 10b) were from RF 16, the lowest four symbols of Figure 11. This flight included two extremes: some of the cleanest air of the program in terms of NSS, nitrate, light scattering, and aerosol concentrations (the 132°E samples) and some of the most heavily polluted air (near Shanghai, the 124°–125°E samples). NSS for instance, was 0.5–0.8 μg/sm3 to the east and 10–21 μg/sm3 near Shanghai. The Shanghai samples also had a high fine-mode fraction of scattering (Figure 10d) and substantial nitrate (Figure 10c), both indicators of pollution. The back-trajectories (Figure 12b) of the clean samples lead back to the subtropical Pacific, while the Shanghai samples had been over the urban/industrial area less than 24 hours before. A frontal system with heavy rain physically separated the two sets of samples.

[48] Surprisingly, OC and EC did not reflect the contrast evident in the sulfate and light scattering. The highest concentrations of both were in one of the Shanghai samples (3.5 μg EC/m3 and 7.8 μg OCq/m3), but the next highest (2.6 μg EC/m3 and 4.8 μg OCq/m3) were in the “cleaner” air that contained almost no sulfate. All four of these samples contained >40% volatile OC, with an OCv S/N of 4–8. The high OC, EC, and volatile fraction in Shanghai's photochemically active air is easy to understand: Many primary and secondary forms of OC would be expected. However, the substantial OC and EC in the otherwise very clean air is harder to explain. The East China Sea is a very active area for shipping. We flew through numerous ship tracks during these otherwise clean MBL legs, so perhaps some of the carbonaceous aerosol came from them. It may also be that we were simply seeing the preferential washout of inorganic aerosols relative to carbonaceous ones.

[49] The remaining samples with more than 30% of their OC volatile (Figure 12b) had all recently experienced contact with polluted areas. This is in contrast to a selection of samples with little or no volatile fraction of OC (Figure 12a), most of which were either well above the surface or over water for several days. Yet there are exceptions, such as the sample that recently was at the surface just north of Shanghai but had virtually no volatile OC. Unfortunately, Figures 1012 give us no solid basis for predicting why some samples were so much more volatile than others. Higher levels of pollutants (Figures 10c and 10d) did not generally cause greater volatility.

[50] Since we had fewer valid OCv samples than TCq and OCq samples, we did not include OCv in many of the plots and tables above, using OCq alone. Figure 13 shows how two derived quantities, the OM/fine mass ratio and the TC/EC ratio change with the inclusion of OCv. Although some points move a bit, neither plot's form is significantly modified by the inclusion of the OCv, even though the statistics do change. About 40% of the samples for which we have valid measurements of OCv evaporated 10% of OC or more from the quartz filter to the CIG filter. Our estimate of the mean volatile fraction (or potential negative artifact, between 12% and 25%) for ACE-Asia samples is only slightly larger than that of Mader et al. [2003], depending on how one handles the valid CIG analyses for which the OCv was too small to detect.

Figure 13.

Impact of including OCv on (a) the fraction of fine mass that is OM and (b) the TC/EC ratio. In both cases including the volatile OC changes the profile only slightly.

[51] How consistent are our data with those of other investigators, both during and prior to ACE-Asia? The SEAREX bulk quartz filter data from the late 1980s [Cachier et al., 1990] suggest that the remote Pacific has a nearly constant springtime background of 0.04–0.05 μg EC/m3, with occasional excursions as high as 1 μg EC/m3. The authors argued that this material was from distant combustion sources and did not identify shipping as a possible source. TC varied from 1 to 5 μg C/m3 (versus our 0.2 to 22 μg C/m3), and was strongly correlated with EC. They noted that the TC/EC ratio varied from 2 to 10 in the North Pacific. Our TC/EC ratios near the Asian continent were in this same range (median of 4.5, range of 2.6 to 15), but our BL concentrations were much higher since we were closer to Asia. We sacrificed sensitivity for mobility: The MDL for our 30–60 min samples was several tenths of a μg EC/m3 or larger. Their background values in the mid-Pacific MBL would have been nondetects for us.

[52] In view of the significant differences in our operational definitions of TC and EC (they used a temperature ramp without optical detection of charring), it is striking that our TC/EC ratios agree so well with those of Cachier et al. [1990]. They contend that TC/EC ratios in the vicinity of 4–6 are indicative of urban and industrial combustion, while much larger ratios (7 to 20) would be typical of forest and grass fires. We did see a few TC/EC ratios greater than 10, but none of these had the high K+ concentrations one might expect in biomass burning smoke. It may be that we encountered so little biomass smoke during ACE-Asia that we simply could not see that relationship, but the urban industrial ratio of 4–6 agreed very well with our data.

[53] Earlier authors had also found higher concentrations closer to Asian sources. Kim et al. [1999] reported a handful of summertime measurements at Seoul and Cheju (now Jeju), Korea. Their quartz filter samples were analyzed using a MnO2 catalyzed thermal oxidation method. At Seoul they found 5–11 μg EC/m3 and 8–14 μg OC/m3, with <1 μg C/m3 of each in particles larger than 2.5 μm. PM2.5 values at Cheju were considerably lower, averaging 2–5 μg OC/m3 and 0.1–0.4 μg EC/m3. TC/EC ranged from about 2.5 at Seoul to 13–60 at Cheju. Our median springtime OC near Jeju (5.8 μg OC/m3) was at the upper end of Kim's range, but our EC median was somewhat larger, 1.7 μg EC/m3. Both sets of measurements were of such limited duration that disagreement would not be surprising. Our TC/EC ratios were rarely as large as those reported at Jeju, but agreed well with Kim et al.'s Seoul values. Our highest EC and OC values were found in the BL south of Korea, for air masses that had recently crossed the Korean Peninsula.

[54] Several groups were making similar OC and EC measurements during ACE-Asia (Table 4). From the Twin Otter, Mader et al. [2002] found that PM2.5 OC and EC below 3 km ranged from 0.6 to 30 μg OC/m3 and from 0.2 to 1.8 μg EC/m3, very similar to the ranges reported here. They found considerably higher levels of OC in polluted air, and (like us) saw little difference in EC levels between air mass types. Mader et al. used the same NIOSH method we did, with an XAD denuder upstream of their quartz filters.

Table 4. Comparison With Other (Mostly ACE-Asia) Observationsa
Location and PlatformMean OCMean ECTC/ECReference
  • a

    C-130 OC and TC from quartz filters only. Uncertainties are standard deviations, since that is most readily found in other publications. Concentrations are in μg/m3. N. Am., North America; nr, not reported.

Yellow Sea
   C-130 ACE-Asia6.7 ± 4.31.7 ± 0.75.2 ± 3.1this study
Gosan and Korea Strait
   C-130 ACE-Asia6.9 ± 3.92.2 ± 1.84.5 ± 1.2this study
   TO ACE-Asia8.4 ± 9.11.3 ± 1.37.5 ± 5.0Mader et al. [2002]
   Brown ACE-Asia1.9 ± 0.80.4 ± 0.26.5 ± 1.9Quinn et al. [2004]
   Jeju ACE-Asianr∼0.8nrChuang et al. [2003]
   Jeju ACE-Asia3.2 ± 1.50.9 ± 0.34.8 ± 1.2Schauer et al. [2003]
Sea of Japan
   C-130 ACE-Asia5.5 ± 2.91.7 ± 1.44.5 ± 3.0this study
   TO ACE-Asia2.5 ±1.61.1 ± 1.03.5 ± 3.6Mader et al. [2002]
   Brown ACE-Asia3.4 ± 0.70.7 ± 0.26.1 ± 1.5Quinn et al. [2004]
   Brown ACE-Asia2.40.74.7Lim et al. [2003]
   VMAP Rishirinr0.4 ± 0.3nrMatsumoto et al. [2003]
   VMAP Sadonr0.7 ± 0.4nrMatsumoto et al. [2003]
Pacific south and east of Japan
   C-130 ACE-Asia4.3 ± 1.11.4 ± 0.65.0 ± 2.8this study
   TO ACE-Asia1.7 ± 2.40.9 ± 0.82.8 ± 4.8Mader et al. [2002]
   Brown ACE-Asia1.3 ± 0.40.4 ± 0.14.8 ± 1.1Quinn et al. [2004]
   Brown ACE-Asia0.70.293.5Lim et al. [2003]
   VMAP Hachijonr0.3 ± 0.2nrMatsumoto et al. [2003]
   VMAP Chichijimanr0.2 ± 0.1nrMatsumoto et al. [2003]
Free troposphere
   C-130 ACE-Asia4.5 ± 1.91.2 ± 0.74.5 ± 2.6this study
   TARFOX N. Am.TEC: 4.3nrnrNovakov et al. [1997]
   Teneriffe, background0.2 ± 0.1nrnrPutaud et al. [2000]
Teneriffe, N. Am./dust0.5/0.7nrnrPutaud et al. [2000]

[55] Some groups separated fine and coarse carbonaceous aerosol. On Jeju Island, Chuang et al. [2003] found significant amounts of EC (0.5 to 1.3 μg EC/m3) in the coarse mode during dust events, and modeled the impact of its mixing state. Size-dependent measurements on the R/V Ronald Brown [Bates et al., 2004] during ACE-Asia found significant amounts of OC on the coarse mode in the presence of dust (2–3 μg OC/m3), but <1 μg coarse EC/m3. Their highest supermicron OC and EC values were found when dust was present, suggesting that dust agglomerates with carbonaceous aerosols. In the seven air mass types they encountered, polluted dusty air also had the highest submicron OC (3.5 μg OC/m3, almost identical with our medians of 3.1 μg OC/m3 and 3.8 μg OC/m3 in the dust-and-pollution categories, 20 and 25) and the highest submicron EC (0.8 μg EC/m3, about half of our polluted and mixed EC medians). Remote Pacific values averaged 0.3 μg OC/m3 and 0.08 μg EC/m3, with the majority of both in the submicron fraction, far smaller than what we measured near the continent. Both groups used the NIOSH thermal/optical method as we did.

[56] The R/V Ronald H. Brown did not encounter concentrations as high as those we observed from the C-130 in the same areas. This could be due to layering: Other pollutants were frequently highly layered, with higher concentrations aloft. This is illustrated in Figure 14, a profile of potential temperature and SO2 over the Yellow Sea. Cold water under warmer air created a series of stable atmospheric layers near the surface, which inhibited vertical mixing within the lower MBL. Thus the highest concentrations we encountered in this profile were two pollution plumes that started several hundred meters above the water. This altitude difference is reflected in the averages in Table 4, where the C-130 and TO often found higher concentrations in an area than did the R/V Ronald H. Brown, Gosan, or VMAP. Although the surface concentrations were lower than those aloft, the TC/EC ratios were similar (every one includes 5.0 within its error bars), which suggests that the surface stations probably encountered a similar aerosol, but less of it.

Figure 14.

Profile of potential temperature (Theta) and sulfur dioxide over the Yellow Sea in RF15. The highest concentrations of SO2 are well above the surface because of the inversion caused by colder water below.

[57] The VMAP surface network organized by M. Uematsu of Tokyo University [Matsumoto et al., 2003; Uno et al., 2003] made OC and EC measurements during the spring of 2001 at four sites from 27°N to 45°N, near 140°E. Average EC ranged from 0.21 μg EC/m3 at the southernmost, marine site, to 0.63 μg EC/m3 at Sado, the closest to urban and industrial sources. OC ranged from 0.5 to 1.7 μg C/m3 across the network. However, these data cannot be directly compared to the other ACE-Asia EC and OC data, because they were collected and analyzed using a different technique. Intercomparisons on Amami Oshima subsequent to ACE-Asia (B. J. Huebert et al., unpublished data, 2003) and Pittsburgh [Anderson et al., 2002] suggest that the two methods can produce quite different values when run side by side.

[58] The principal reason for using an aircraft is to measure vertical profiles, but we were not always successful. In our case we failed to measure OC and EC in the FT except in a handful of cases (Figure 5), preventing us from confidently describing FT-MBL differences. Low FT concentrations combined with reduced pumping efficiency and short sampling times made it difficult to collect enough of these analytes to quantify.

[59] However, in all but one case we were able to quantify TCq, even when the OC/EC split point was ambiguous. We can use those TC data to estimate FT EC and OC. If we assume that the TC/EC ratio of 4.5 (which was the median, mean, or both in every altitude range) applied to all the TC samples, we obtain the circles in Figure 15. Above 2500 m this gives us 18 EC values, with FT medians of 0.7 μg EC/m3 (Table 1, bracketed values in the top section) and 2.5 μg OCq/m3. (Bracketed values are not included in Table 1 for the lower altitude ranges, since filling the few gaps using TC made negligible changes in those statistics.) The resulting BL/FT ratios are 2.3 for OC and 2.6 for EC. The TC gradient was 2.4; even if we assumed that none of this was EC in the FT, the OC altitude ratio would not decrease by much.

Figure 15.

(a) OC and (b) EC altitude profiles, derived by assuming that the MBL TC/EC ratio of 4.5 applied to all samples for which TC was measured but the OC/EC split point was ambiguous.

[60] Alternately, one could assume constancy of the EC specific absorption (ESA, or mass absorption efficiency) and derive EC values in the FT from PSAP absorption data. Clarke et al. [2004] report the project-average absorption measured from the C-130 as about 12 Mm−1 in the BL and 2 Mm−1 in the FT. This factor of 6 BL/FT gradient in absorption is larger than the factor of 2.7 that we inferred for mean EC. One possibility is that the smaller relative amount of NSS in the FT (Figure 7a: material coating EC would contain more OC than NSS in the FT) reduces the absorption efficiency of EC in the FT to about half its BL value, and our TC-derived EC is a good approximation. Another is that our samples do not cover exactly the same times as the PSAP did (we only measured during level legs, and not all of those), so sampling bias could explain some of the difference. A third is that the ESA really is constant, so the mean FT contains only one sixth as much EC as the BL, or about 0.4 μg EC/m3. Although we don't know whether the “constant ESA” or “constant TC/EC” assumption is better, we can conclude that the FT average EC was between 0.4 and 0.8 μg EC/m3.

[61] In Asian outflow TC (and therefore OM) was more concentrated than NSS in every FT sample we were able to measure (Figure 7a). Novakov et al. [1997] were the first to suggest this possibility. Likewise, Putaud et al. [2000] reported that there might be less sulfate than OM in the FT at Teneriffe. Our data support the hypothesis that the FT contains more carbonaceous aerosol mass than sulfate. Direct forcing climate models must therefore include sources, vertical transport, and removal mechanisms that can correctly forecast the total FT carbonaceous aerosol mass. Sulfate alone is clearly not a realistic approximation of FT aerosols. It is important to note, however, that all of these measurements have been made near continents; there are no published observations that indicate whether or not OM dominates over NSS in remote areas.

[62] Why might OM/NSS in the FT be different from that in the BL? From Figure 7a it is evident that much lower values of TCq/NSS can be found in the BL, although the upper end of the range is similar in both regimes. Evidently, a larger fraction of C (either as OM aerosol or precursor VOC) than SO2 and NSS mixes into the FT, perhaps because the carbonaceous species are less susceptible to washout. It is also possible that low solubility confers a slightly longer FT lifetime on OM than NSS, although its solubility should increase with time because of the photochemical oxidation of OM.

[63] The layering that is apparent for rapidly measured parameters like SO2 (Figure 14), ozone, CN, and light scattering must also apply to OC and EC, even though we could not measure these species during vertical profiles. There are probably OC and EC layers with concentrations that differ by up to an order of magnitude over scales of a few hundred meters. During the RF15 profile in Figure 14, the aerosol absorption went through excursions of more than an order of magnitude, so the same must be true of EC. Our slow integrating method was incapable of characterizing the vertical variability of EC and OC, but related data suggest that it must be similar to that of SO2, with layers of polluted BL air sandwiched between cleaner FT air.

[64] One way to determine the fraction of the FT impacted by anthropogenic sources is to use a substance like SO2 that is a clear indicator of pollution and was measured rapidly during vertical soundings. If we look only at profile data (no level legs) above 2000 m for the 19 flights from Iwakuni, 31% of the 10 s average SO2 values are above 50 pptv, a reasonable upper limit for background air. Thus, if we assume that BL volcanoes had little impact on the FT, almost a third of the FT air between 2 and 7 km showed evidence of impact from anthropogenic sources. This high prevalence of polluted layers persists well out into the Pacific, where frequent layers of high ozone, CO, and aerosols have been reported for decades [Routhier and Davis, 1980].

[65] In every altitude regime and region we found a tremendous range of concentrations. The max/min ratios are an order of magnitude or more for EC and OC and approach 2 orders of magnitude for TC. Naturally, the most extreme concentrations were in the BL where the pollution sources are located. The broad range of commonly observed values for OC was striking (Figure 6).

[66] The presence of dust does not seem to change the amount of TC in most air masses (Figure 7b). If anything, high soluble Ca (high dust) was associated with higher TC. This may reflect the fact that our highest dust concentrations were in storms that had passed over the Beijing and Qingdao urban areas. Dust reduces the fraction of the total mass that is carbonaceous, but it does not change by much the absolute amount of TC. Dust probably doesn't include much native OC or EC, since by nature dust source areas are sparsely vegetated and populated. TCq/Ca was virtually identical (0.5) in the northeastern lakebed dust of RF13 and the far western desert dust in RF06. Trajectories suggest the former may have had less contribution from urban and industrial sources like Beijing and Qindao.

[67] It may be, however, that the urban areas are not the largest sources of OC and EC in China. The amount of miscellaneous, largely uncontrolled combustion in China is profound, as it no doubt is in any heavily populated agricultural region. While working in Shaanxi Province and its surrounds in 2000, B. J. Huebert could always see numerous smoke plumes. They came from piles of burning rubbish, coal-fired kitchens, agricultural waste burning, brick kilns and other small industry, large industrial facilities, diesel trucks, aged power plants, and residential heating and cooking. This huge, dispersed source of OC and EC nearly defies quantification by its very diversity. Streets et al. [2003] assume uncertainties of 360% for BC and 450% for OC for their Asian emissions inventory [Huebert et al., 2003]. Virtually the only way to determine whether a source model makes sense is to use it as the basis of a chemical transport model, and compare the results with ambient measurements such as those we describe above. Hopefully these data will prove useful in that regard.

5. Conclusions

[68] Although short aircraft sampling times limited our sensitivity, total carbon (EC plus OC) was quantifiable on all but one level flight leg. We could not confidently apportion TC between EC and OC on about a quarter of the legs, mostly in the free troposphere (FT). We found that TC and OC concentrations were generally higher in the boundary layer (medians of 7.6 μg TC/m3 and 5.8 μg OC/m3), than the free troposphere (medians of 3.1 μg TC/m3 and 3.9 μg OC/m3, the latter from a subset of the former). The same appears to be true for EC, but the FT statistics are poor. Average concentrations were somewhat higher than medians, reflecting the impact of a few more heavily polluted samples. OC was much more variable than was EC.

[69] The VOC denuder in the PC-BOSS sampler should have largely eliminated vapor-adsorption artifacts. About a third of the samples showed significant evaporation of collected aerosols from their quartz filter, while a similar number had too little evaporation to detect. The volatilized fraction of OC, which was captured by backup CIG filters, ranged from negligible to as much as 70%, although this highest value may reflect warming of the aerosol during sampling. The mean volatility was between 12 and 25% of collected OC. We were unable to identify the factors that control this volatility.

[70] We used two methods to estimate EC in the FT, a constant TC/EC assumption that made use of the many valid FT TC samples, and a constant ESA assumption that made use of absorption measurements. The resulting mean EC values probably bracket the true FT mean EC, as between 0.4 and 0.8 μg EC/m3.

[71] In the free troposphere (FT) the TC to non-sea-salt sulfate ratio ranged from 3.2 to 6.0, so carbonaceous aerosols were more concentrated than NSS in the upper troposphere. When dust mass was included, organic matter ranged from about 0.1 to 0.7 of the total fine particle mass. TC/EC ratios ranged from 2 to 15, with medians of about 4–5, suggesting urban/industrial sources rather than agricultural burning. Air in the Yellow Sea, the Korea strait, and the Sea of Japan had generally higher concentrations of OC than that in the East China Sea or the Pacific south and east of Japan.

[72] In every altitude regime and region we found a tremendous range of concentrations. The max/min ratio approaches 2 orders of magnitude for BL OC. Since pollution sources are mostly at the surface, we found the most extreme concentrations in the BL. It is important to recall that the values tabulated here only represent the spring outflow. Our sampling was intentionally biased toward dust and pollution events, with the exception of several survey flights intended to cover large areas and FT samples taken during transits to boundary layer features of interest. The statistics in Tables 14 are therefore useful guides to typical concentrations, but should not be used as figures that regional models must match each spring.

Acknowledgments

[73] 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 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. The UH work was supported by NSF grant ATM00-02698 and amendments thereto, and the BYU work was supported by ATM01-29520. This research is a contribution to the International Global Atmospheric Chemistry core project of the International Geosphere-Biosphere Program. This is SOEST contribution 6346.

Ancillary