Source signatures of carbon monoxide and organic functional groups in Asian Pacific Regional Aerosol Characterization Experiment (ACE-Asia) submicron aerosol types



[1] Atmospheric submicron particles were collected on Teflon filters downstream of a three-stage concentrator aboard the National Center for Atmospheric Research C-130 aircraft near Japan during the Asian Pacific Regional Aerosol Characterization Experiment (ACE-Asia). Particle-phase organic carbon (OC) was quantified using Fourier transform infrared (FTIR) transmission spectroscopy. Silicate, carbonate, alkane, alkene, aromatic, alcohol, carbonyl, amine, and organosulfate functional groups were identified and separated with a four-solvent rinsing procedure. X-ray fluorescence identified elemental composition. Total OC constructed from FTIR measurements agreed with simultaneous thermal-optical OC measurements with a slope of 0.91 and an R2 value of 0.93. OC varied from 0.4 to 14.2 μg m−3, and organic mass varied from 0.6 to 19.6 μg m−3, representing on average 36% of the identified submicron aerosol mass. Measured carbon monoxide (CO) to OC slopes illustrate 10 groups of air from regions described by an Asian emissions inventory. The CO/OC slope is used to compare sources and their influence on organic composition. Fifty-two percent of ACE-Asia samples have CO/OC slopes indicative of biomass combustion. Unitless CO/OC slopes above 15 are associated with increased fractions of alcohol groups, unsaturated C-H groups, and inorganic nitrate. Increased carbonyl carbon fractions in air originating over northern Asia are consistent with secondary OC formation. Case studies in the boundary layer demonstrate that aerosol compositions downwind of large Asian aerosol sources show clear regional composition signatures.

1. Introduction

[2] Asia is one of the largest aerosol source regions on Earth, and plumes of Asian aerosol have been observed crossing the Pacific to reach the West Coast of the United States [Husar et al., 2001; Jaffe et al., 1999]. The aerosol downwind of the Asian continent is a complex mixture of anthropogenic and natural sources that has been measured in several campaigns but that is still poorly understood [Kinne and Pueschel, 2001; Parungo et al., 1994]. Organic compounds in particles may enhance or inhibit the ability of an aerosol particle to absorb water and form a cloud droplet [Jacobson et al., 2000; Saxena et al., 1995], and in this respect can affect the probability that the Asian plume will be removed by wet deposition before it reaches the West Coast of the United States.

[3] Understanding the behavior and atmospheric effects of the Asian plume requires knowledge of the amount of organic mass present in the aerosol and the solubility properties of the aerosol components. However, complete characterization of the organic aerosol fraction is extremely difficult because the organic fraction covers a wide range of chemical and thermodynamic properties. For this reason, organic speciation studies often identify only 10–20% of the organic compounds present, while thermal methods that can identify all of the carbon present give no information about its chemical and physical properties [Turpin et al., 2000]. Organic compounds are usually the second most abundant component of fine aerosols after sulfates [Heintzenberg, 1989], and are therefore a large component of the atmospheric aerosol that has not been completely characterized. Our understanding of Asian aerosols in particular, and atmospheric aerosols in general, is currently limited by the lack of information about the composition and hygroscopicity of this fraction.

[4] Several studies have used thermal-optical measurements of the organic carbon (OC) to elemental carbon (EC) slope as an indicator of aerosol source types and as a means of finding evidence of secondary organic aerosol formation [Kadowaki, 1990; Kim et al., 2000; Lim and Turpin, 2002]. These analyses utilize the notion that particulate EC is produced only in combustion processes as a primary pollutant while particulate OC is both emitted from sources and produced by atmospheric reactions from gaseous precursors. However, differences between OC-EC analysis methods can lead to factor of two differences in reported EC [Lim et al., 2003]. CO, like EC, is generated predominantly by primary combustion sources [de Laat et al., 2001]. In fact, EC and CO are often found to co-vary [Chen et al., 2002; Lim and Turpin, 2002]. The CO/OC slope, like the EC/OC slope, is therefore an indicator of organic aerosol source types and secondary organic aerosol formation. Sulfate, ammonium, and nitrate are associated with many of these same combustion sources, so that the CO/OC slope is also a source indicator for inorganic species. Dust from sources that are likely to be spatially and temporally uncorrelated with CO sources is not expected to show a trend with the CO/OC slope.

[5] This work presents quantitative measurements of submicron aerosol organic and inorganic composition. OC and organic mass (OM) concentrations are calculated from Fourier transform infrared (FTIR) functional group measurements, and the FTIR-derived OC is compared to OC measured with a thermal-optical method. An air mass classification scheme based on back trajectories is used to separate the Asian Pacific Regional Aerosol Characterization Experiment (ACE-Asia) data into groups with similar CO/OC ratios and organic functional group composition. The CO/OC slope is used along with the OM/OC ratio, functional group information, and an emissions inventory to examine the atmospheric processes influencing the submicron particulate matter in the ACE-Asia region.

2. Experimental and Analysis Procedure

2.1. Particle Concentration and Collection

[6] During ACE-Asia in April and May of 2001, submicron aerosol samples were collected near Japan from the NCAR C-130 aircraft. Sample air was brought into the C-130 at 180 l min−1 through a solid diffuser inlet followed by a 1.06 μm cutoff impactor and three virtual impactors in series. The virtual impactors concentrated the aerosol into a 2 l min−1 submicron sample stream prior to collection on two Teflon filters in series (64 cm s−1 face velocity). Except for the differences noted here, the sampling method is identical to that of Maria et al. [2002].

[7] Mass flow controllers on all flow streams (Teledyne-Hastings HFC-202 and HFC-203) were electronically adjusted in real time to provide constant volumetric flow rates as the inlet temperature and pressure changed, and a relative humidity probe (Vaisala Humitter-50U) immediately upstream of the impactor monitored the relative humidity of the sample stream. In the filter holders, one Teflon gasket was used upstream of the sample filter during research flights 1–6. For research flights 7–19, two gaskets were used, one immediately upstream of the sample filter and the other immediately downstream of the backup filter that was used as a blank. Sampling with a single gasket upstream of the sample filter resulted in an aerosol deposit that was nearly uniform across the entire 31 mm diameter filter. The use of two gaskets confined the sample deposit to the filter's center 10 mm diameter circular area, and increased the amount of sample within the FTIR analysis area by a factor of 2.75. This difference was corrected in the reported concentrations, all of which are at standard temperature and pressure (298°K and 1 atm).

[8] The mass-average concentration factor, defined as the submicron mass concentration in the filter flow divided by that in the ambient air, varied from 10 to 16 during ACE-Asia. A concentration factor was calculated for each filter sample using a laboratory-measured size-dependent concentration curve [Maria et al., 2002] and ambient size distributions measured using an optical particle counter (OPC) (S. Howell et al., manuscript in preparation, 2003). Uncertainties in the solid diffuser inlet efficiency and OPC size distributions add ∼7% to the error of all calculated mass-average concentration factors.

[9] During ACE-Asia, 172 filter samples and 172 downstream blanks (backup Teflon filters) were collected during research flights in the marine boundary layer and in free tropospheric air masses of Asian origin. Of the 172 samples, 69 were collected in high-altitude free tropospheric air, 75 were collected within the marine boundary layer, and 28 were collected during vertical profiles. The one-gasket sampling method was used for the first 42 samples, and the two-gasket method for all others. Sample collection times ranged in length from 9 to 100 min, and allowed for the characterization of Asian aerosol downwind of urban, desert, and volcanic sources. Throughout the project, no detectable peaks were observed on downstream blank filters, consistent with previous studies that found low organic artifacts on Teflon filters [Turpin et al., 1994]. Downstream blanks were placed behind the sample filters to determine the detection limits for each quantified functional group.

[10] After sampling, filters were placed in polystyrene petri dishes (Pall Corporation) sealed with Teflon tape and stored in a freezer until analysis. Powder-free vinyl gloves and Teflon-coated tweezers were used when handling filters.

2.2. Solvent-Rinsed FTIR Spectroscopy

[11] Field samples were analyzed for functional group composition and solubility behavior at the Environmental and Occupational Health Sciences Institute at Rutgers University using the method described by Maria et al. [2002]. The identified functional groups, along with their corresponding absorption frequencies and detection limits, are shown in Table 1. Functional group identification was expanded from previous work to include alkene and aromatic C-H groups as well as alcohol, amine, organosulfur, nitrate and carbonate groups, as illustrated in Figure 1. All of these groups were positively identified in ACE-Asia sample spectra using absorption frequencies determined from previously published aerosol FTIR spectra [Allen et al., 1994; Blando et al., 1998], laboratory-generated standards, and spectral libraries. Particles-into-liquid-sampler (PILS) ion chromatography measurements [Weber et al., 2001] are reported in this paper for characterization of sulfate, ammonium, and nitrate.

Figure 1.

(a) FTIR spectra of a typical ACE-Asia sample, collected south of Korea at 150 m from 02:06 to 02:34 on 27 April 2001 (research flight 15). The original sample is shown (bold solid line) with spectra after rinsing with hexane (thick dotted line), DCM (thin solid line), acetone (thin dashed line), and water (thin dot-dashed line). The 1150–1300 cm−1 Teflon and 2300–2380 cm−1 CO2 interference regions have been removed from the raw data (gray areas). In Figure 1a, vertical lines indicate the functional group absorbance regions of (from right to left) sulfate (618 cm−1), organosulfur (876 cm−1), silicate (1035 cm−1), nitrate (1345 cm−1), ammonium (1437 cm−1), amine (1621 cm−1), carbonyl (1720 cm−1), alkane (2850–2920 cm−1), alkene (2980 cm−1), aromatic (3065 cm−1), and alcohol (3297 cm−1). Carbonate is the residual peak at 1437 cm−1. All functional groups except for organosulfur were above detection limit in this sample. Enlargements show details of (b) the 2600–3400 cm−1 region, (c) the 1500–1850 cm−1 region, and (d) the 1300–1500 cm−1 region.

Table 1. Peaks Used in FTIR Quantification of Aerosol Functional Groups
Functional GroupAbsorption Frequencies, cm−1Quantified Peaks, cm−1Absorptivity abs−1aDetection Limit, μg cm−2Rinsing Stage,b
  • a

    Here, abs (absorptivity) is the peak area (in absorbance units) per micromole of functional group. The calibration of absorptivity is instrument-specific.

  • b

    Here, h, hexane; d, dichloromethane; a, acetone; w, water; and r, residual.

  • c

    All organic absorptivities were determined from n-nonadecane, 1-docosanol, camphor, anthracene, perinaphthenone, citric acid, adipic acid, oxalic acid, EDTA, alanine, and methane sulfonic acid standards.

SO42− sulfate ions612-5, 1103–356180.410.37h, d, a, w
HSO4 bisulfate ions580–90, 867, 1029, 11806180.410.37h, d, a, w
SiO44− silicate ions772–812, 103510350.0110.03h, d, a, w, r
NH4+ ammonium ions1410–35, 3030–52, 3170–320014370.140.056d, a, w
CO32− carbonate ions860–80, 1410–9014370.110.08r
NO3 nitrate ions815–40, 1350–8013450.0190.35h, d, a
H2O liquid water1623, 3350–3450not used  w, r
C-H aliphatic carbon1452–5, 2800–30002850–29201.06c0.57h, d, a, w, r
C = C-H alkene carbon2900–31002980–30050.30c0.64h, d, a, r
C = C-H aromatic carbon3000–310030650.17c0.47h, d, a, r
C = O carbonyl carbon1640–185017200.061c0.11h, d, a, w, r
C-OH alcohols3100–350032970.063c0.08h, d, a
C-O-S organosulfates 8760.031c0.02h, d, a
C-NH2 amines163016210.19c0.26h, d, a

[12] Absorbance peaks were identified both by location and by solubility behavior as shown in Table 1. For example, the 1437 cm−1 peak was separated into ammonium and carbonate absorbances on the basis of rinsing fraction (Table 1), resulting in an improved linear correlation between the integrated areas of the 1437 and 3238 cm−1 ammonium peaks (R2 = 0.8 using only the hexane, dichloromethane, acetone, and water fractions of the 1437 cm−1 peak; R2 = 0.6 using the entire 1437 cm−1 sample peak). Similarly, organosulfur was identified using the fractions of the 876 cm−1 peak that were removed in organic solvents, distinguishing it from bisulfate which is removed in water [Blando et al., 1998]. Alcohol, alkane, and aromatic groups removed in water could not be identified because of the large interference from ammonium, which is removed in the water rinse.

[13] An absorbance peak at 1620 cm−1 was observed in 39% of the ACE-Asia samples. Absorbances at 1620 cm−1 can be caused by organonitrate, aromatic, amide, and amine organic functional groups, as well as by aerosol water [Nakanishi and Solomon, 1977]. The lack of corresponding peaks at 1280 cm−1 (organonitrate), 1500 cm−1 (aromatic), and 1650–1690 cm−1 (amide) eliminates all possible organic functionalities except for amine. Aerosol water is removed with water-soluble aerosol components in the water rinse or is left behind in the residual fraction, so all solvent-removed fractions of the 1620 cm−1 absorbance are quantified as organic amines.

[14] The absorbance calibrations of anthracene, docosanol, methane sulfonic acid, ammonium nitrate, calcium carbonate, ethylenediaminetetraacetic acid, alanine, and sodium silicate varied linearly with the number of moles of bonds and mass loading as shown in Figure 2. These standards, in addition to the calibration of Maria et al. [2002], provide a basis for quantifying the mole and mass quantities detected by FTIR in ambient aerosol samples. For each standard a linear fit above the detection limit had a correlation above R2 = 0.88, and the slope of each fit is reported as the functional group absorbance per micromole of bonds in Table 1. Detection limits for each species were set to a visually-determined minimum identifiable peak size, taking into consideration signal noise from downstream blanks.

Figure 2.

FTIR response for laboratory-generated aerosol of species in varying sample amounts. Circles are CO3 (1437 cm−1) from calcium carbonate aerosol, squares are NO32− (1345 cm−1) from ammonium nitrate aerosol, upward pointing triangles are C-H (aromatic: 3065 cm−1) from anthracene aerosol, diamonds are C-H (alkene: 2980 cm−1) from 1-decene aerosol, downward pointing triangles are C-OH (3297 cm−1) from 1-docosanol aerosol, sideways hourglasses are C-N (amine: 1621 cm−1) from EDTA aerosol, vertical hourglasses are C-S (876 cm−1) from methane sulfonic acid aerosol, plus signs are SiO44− (1035 cm−1) from sodium silicate aerosol, and crosses are C-N (amine: 1621 cm−1) from alanine aerosol. Lines indicate best fit linear correlations to the data.

2.3. X-Ray Fluorescence (XRF) Analysis

[15] After FTIR analysis, XRF was performed on the same filters to quantify all elements heavier than sodium, including iron, silicon, sulfur, calcium, chlorine, and vanadium (Chester LabNet, Tigard, Oregon). XRF detection limits for most elements are on the order of 1–10 ng cm−2, a notable exception being Na (which has a detection limit of 150 ng cm−2, too high to be quantified in these samples). XRF detection limits reflect the signal-to-noise ratio for each element, and reported errors include the detection limit plus a 5% calibration uncertainty as well as spectral overlap uncertainties when a secondary line from one element overlaps the primary line from another element. XRF errors do not reflect uncertainties in field blanks, which were less than 4% of the corresponding sample values.

2.4. CO Measurement

[16] The NCAR CO instrument operates on the principle of vacuum UV resonance fluorescence (Aero-Laser 5002), as published by Gerbig et al. [1999]. It has a detection limit of 3 ppbv with an accuracy of 5 ppbv + 2% for a 1-second sampling rate.

2.5. Organic Mass Estimates

[17] OC and OM were estimated according to the following formulas:

equation image
equation image

[18] As defined, OC includes carbon mass only and OM includes carbon, hydrogen, oxygen, nitrogen, and sulfur. This definition is equivalent to a carbon mass to organic compound mass conversion factor of 1.1 for alkene and aromatic groups, 1.2 for alkane groups, 2.3 for carbonyl and amine groups, 2.4 for alcohol groups, and 4.5 for organosulfur groups. These conversion factors more than span the range given by Turpin and Lim [2001], who suggested that 1.6 ± 0.2 is a reasonable factor for an urban aerosol and 2.1 ± 0.2 is more appropriate for an aged (non-urban) aerosol. Different combinations of FTIR-identified functional groups produce OM/OC ratios that are reasonable both for urban and non-urban conditions. FTIR characterization of OM is the first technique to utilize nearly 100% of the atmospheric OC in its calculation. Other methods have depended on extrapolation from chromatography measurements that typically identify less than 20% of the total OC [Schauer et al., 1999; Turpin and Lim, 2001; White and Roberts, 1977]. The estimated errors in our reported OM/OC ratios are on average 24% [Russell, 2003].

[19] Equations (1) and (2) are similar to equations proposed by Maria et al. [2002], expanding those expressions to include groups other than alkane and carbonyl. To illustrate the importance of this extension, consider an organic aerosol composed of 50% nonadecane, 25% anthracene, and 25% oxalic acid by mass. For this example, assuming that enough sample is collected such that all functional groups are above detection limits, the current equations would overestimate OC by 3% and OM by 2% while the equations of Maria et al. [2002] would underestimate OC by 29% and OM by 32%.

3. Results and Discussion

3.1. OC Comparison

[20] Measurements of OC, EC and carbonate carbon (CC) were performed on ACE-Asia C-130 samples (T. Bertram et al., manuscript in preparation, 2003) using a Sunset Labs Model 3 TOT carbon analyzer [Schauer et al., 2003]. The ACE-Asia TOT sampling system on the C-130 was a PC-BOSS [Lewtas et al., 2001] that incorporated a denuder upstream of a quartz filter to minimize positive sampling artifacts and a carbon-impregnated glass fiber filter downstream to correct for negative artifacts. In the analysis OC is operationally defined as carbon that volatilizes upon heating to 870°C in the absence of oxygen, plus any OC that charred and then oxidized after oxygen addition.

[21] The correlation between TOT OC and FTIR OC is strong, with a coefficient of determination (R2) of 0.93 (Figure 3a) for simultaneous samples shorter than 20 minutes in length. Samples were considered simultaneous if differences in sampling start and stop times for the two collection systems did not exceed 10% of the total sample time. The slope of the correlation (0.91) and the zero intercept imply that both the FTIR and TOT methods measured very similar groups of species as OC. A paired t-test shows no significant difference between the two data sets (p = 0.78). The TOT OC/EC distinction is accurate for many organic compounds, with average measurements of less than 3% EC for sucrose and 1% EC for EDTA [Birch, 1998]. In the more complex ACE-Asia samples, the OC/EC split determined by TOT has larger uncertainties. The less than 10% difference between the TOT and FTIR measurements might be caused by a combination of uncorrected positive artifacts in the TOT collection system (due to possible denuder inefficiency), deviations from assumptions inherent to the correction for pyrolysis of OC to EC [Subramanian et al., 2002], and uncorrected negative artifacts in the FTIR collection system. Such artifacts would increase in magnitude with increased sampling time and increased ambient temperature variations.

Figure 3.

FTIR OC concentrations compared to TOT OC measurements, with the 1:1 line shown as a dotted line. Solid circles indicate samples that were collected simultaneously. Open circles represent all other samples collected. The best fit lines (solid) are shown for (a) samples less than 20 minutes in length (slope = 0.91, R2 = 0.93) and (b) all simultaneous samples (slope = 0.90, R2 = 0.82).

3.2. CO-OC Relationship

[22] CO is produced in Asia through incomplete combustion processes, and is often associated with the emission of particulate EC. General circulation models have predicted a relatively long atmospheric lifetime of CO (10–360 days [Holloway et al., 2000] as compared to an OC lifetime of 3.4 days [Cooke et al., 2002]), resulting in a CO/OC ratio that increases with distance from the emission source. Formation of secondary organic aerosol decreases the CO/OC ratio. The CO/OC ratio varies widely between sources, and despite atmospheric processing is suitable for the identification of dominant source types. On the C-130 during ACE-Asia, the CO/OC ratio was more accurately measured than the OC/EC ratio.

[23] CO concentration as a function of OC concentration is shown in Figure 4 for all ACE-Asia samples. By averaging the smallest 5% of CO measurements, a background CO concentration of ∼100 ppbv can be inferred from the ACE-Asia data both above and within the boundary layer. This value is comparable to the average CO concentration of 92 ppbv for aged marine air downwind of Asia during the spring of 1994 [Talbot et al., 1997]. The unitless CO/OC slope for ACE-Asia samples is calculated using the following formula:

equation image

where x is the CO/OC slope, CO and OC are in μgC m−3, and the 49 μgC m−3 offset represents the background CO concentration. In Figures 4a and 4b, location-specific and source-specific CO/OC slopes were calculated from direct measurements of OC and CO.

Figure 4.

Carbon monoxide concentrations versus FTIR OC concentrations, with ACE-Asia measurements shown as open circles. In Figure 4a, lines correspond to CO/OC slopes from measurements at (1) Atlanta, Georgia [Lim and Turpin, 2002], (2) Fort Meade, Maryland [Chen et al., 2002], (3) Nagoya, Japan [Kadowaki, 1990], and (4) Asia [Streets et al., 2003]. In Figure 4b, lines correspond to CO/OC slopes from the following emission sources: (1) automobiles [U.S. Environmental Protection Agency (USEPA), 2003; Schauer et al., 2002] (the former is available at, (2) diesel trucks [Lloyd and Cackette, 2001], (3) wood burning [USEPA, 2003; Cabada et al., 2002], (4) biomass burning [Andreae and Merlet, 2001], (5) cigarette smoking [Martin et al., 1997; Cabada et al., 2002], (6) natural gas [USEPA, 2003; Cabada et al., 2002], and (7) coal [USEPA, 2003; Cabada et al., 2002].

[24] The CO/OC slope is as high as 110 for some ACE-Asia samples but falls near 6.4 for the majority of the ACE-Asia project. The average measured ACE-Asia CO/OC slope is significantly lower than slopes measured in Japan and the United States, and 56% of ACE-Asia samples have CO/OC slopes lower than that of the average Asian primary CO and OC emissions [Streets et al., 2003] (Figure 4a). The large number of samples with low slopes suggests that emissions from the transportation sector are not as large as emissions from biomass burning and coal combustion in the ACE-Asia region (Figure 4b). A large portion of the total coal and biomass combustion in China is residential, and Klimont et al. [2002] estimate that emissions factors from residential combustors are more than 10 times larger than those from industrial combustors. Asian coal and biomass combustion may therefore significantly affect the CO/OC ratio in some regions. Figure 4b shows that 52% of ACE-Asia samples have CO/OC slopes within 25% of the value for wood and biomass burning sources and 43% have slopes within 25% of the values for automobiles and diesel trucks. The remaining 5% of samples have CO/OC slopes that are between biomass burning and coal sources. These percentages were calculated using only samples with CO concentrations larger than 75 μgC m−3 or OC concentrations larger than 1.0 μg m−3, because of uncertainty in the CO/OC slope for samples with lower concentrations.

3.3. Back Trajectory Classification

[25] Source regions were assigned to each aerosol sample using isentropic 5-day back trajectories from the HYSPLIT4 model [Draxier and Hess, 1998], with the resulting qualitative groups shown in Figure 5 numbered in order of increasing measured CO/OC slopes. An average trajectory was calculated for each back trajectory group, and the total emissions [Streets et al., 2003] from the area within 113 km (1 degree) of the average 5-day back trajectory were summed to calculate a primary CO/OC slope. CO/OC slopes were comparable within 7 out of the 10 back trajectory groups (R2 = 0.52–0.93) and agreed well with the primary emissions inventory for 6 out of the 10 back trajectory groups (see Figure 5).

Figure 5.

ACE-Asia back trajectories classified into 10 groups based on source region. Each trajectory plot contains all of the sample trajectories for the group (thin dotted lines), the calculated average trajectory for the group (thick dashed line), and the group number (upper right hand corner). To the right of each trajectory plot are the corresponding CO and OC measurements (black circles), the best fit linear trends to the data (solid lines) with equations and R2 values, and the calculated emissions from Streets et al. [2003] along the 5-day average back trajectory (thin dashed lines). For group 10, the emissions along only the first 6 hours of the back trajectory are also indicated (thick dot-dashed line).

[26] Group 10, unlike the other back trajectory groups, contains samples collected directly within the plume of a major city (Qingdao, in the Shandong Province). A summation of the emissions from the area within 113 km of the first 6 hours of the back trajectory (instead of the entire 5-day back trajectory) agrees well with the measurements for group 10, demonstrating the importance of the local source in this case. The large CO/OC slope (84) for this group reveals that the Qingdao emissions are dominated by large sources of CO that do not produce appreciable amounts of OC. Automobiles, which are associated with cities such as Qingdao, are an example of such sources. High-OC sources such as biomass, coal, and natural gas combustion are not significant contributors to the Qingdao emissions. The inventory of Streets et al. [2003] shows that the transportation sector is the largest contributor to CO emissions in the Shandong Province.

[27] CO/OC slopes 50–55% smaller than predicted by the emissions inventory were observed in air masses that originated over northern Asia (groups 1–3). For these groups, measured CO/OC slopes were ∼7.4, consistent with biomass burning sources. CO/OC slopes calculated from the emissions inventory were 14–16. This discrepancy suggests secondary OC formation in these air masses, because secondary formation lowers the CO/OC slope of an aerosol after emission. Alternatively, the emissions inventory may be underestimating the amount of biomass burning in the North Asian source regions for groups 1–3. Uncertainties in the modeled paths of the back trajectories may also contribute to the discrepancy.

[28] Tables 2 and 3 show the average fractional ACE-Asia functional group composition for each back trajectory group in order of increasing CO/OC slope. Variations in functional group composition within each back trajectory group are small compared to differences between groups, suggesting that the aerosol source region and its associated CO/OC slope are good indicators of aerosol composition in the ACE-Asia region. Aerosol sulfate, ammonium, and nitrate all show positive correlations with the CO/OC slope. Dust species do not show a correlation because dust does not have sources or regions that are common to CO. The small OM/OC ratios of all categories are consistent with organic compounds emitted from urban and industrial source regions less than 24 hours before sampling.

Table 2. Fractional Submicron Compositiona
 Group 1bGroup 2Group 3Group 4Group 5
  • a

    Averages and standard deviations of the mass fractions of each submicron aerosol component are reported here for each back trajectory group.

  • b

    Group numbers are the same as in Figure 5.

  • c

    The sulfate, ammonium, and nitrate values used here were taken from the PILS C-130 data set [Weber et al., 2001].

OM0.49 ± 0.070.38 ± 0.080.31 ± 0.060.57 ± 0.050.25 ± 0.03
Dust0.22 ± 0.040.26 ± 0.050.49 ± 0.030.25 ± 0.030.31 ± 0.05
SO42−c0.19 ± 0.030.24 ± 0.030.12 ± 0.060.16 ± 0.010.31 ± 0.02
NH4+c0.05 ± 0.010.08 ± 0.030.04 ± 0.010.01 ± 0.010.08 ± 0.02
NO3c0.05 ± 0.010.05 ± 0.000.04 ± 0.010.01 ± 0.010.05 ± 0.00
 Group 6Group 7Group 8Group 9Group 10
OM0.48 ± 0.050.41 ± 0.050.31 ± 0.060.25 ± 0.070.10 ± 0.05
Dust0.09 ± 0.030.27 ± 0.040.17 ± 0.030.07 ± 0.040.20 ± 0.05
SO42−c0.30 ± 0.020.17 ± 0.020.32 ± 0.030.48 ± 0.020.31 ± 0.04
NH4+c0.09 ± 0.020.09 ± 0.030.11 ± 0.020.12 ± 0.010.12 ± 0.01
NO3c0.05 ± 0.010.06 ± 0.010.10 ± 0.020.08 ± 0.010.26 ± 0.05
Table 3. Fractional Organic Compositiona
 Group 1bGroup 2Group 3Group 4Group 5
  • a

    Averages and standard deviations of the fractions of OM are reported here for each back trajectory group.

  • b

    Group numbers are the same as in Figure 5.

Alkane0.58 ± 0.030.60 ± 0.040.60 ± 0.030.64 ± 0.070.65 ± 0.04
Alkene0.15 ± 0.030.15 ± 0.030.18 ± 0.010.12 ± 0.020.18 ± 0.02
Aromatic0.08 ± 0.010.08 ± 0.010.06 ± 0.020.09 ± 0.010.08 ± 0.01
C-OH0.02 ± 0.010.01 ± 0.010.02 ± 0.010.02 ± 0.010.01 ± 0.01
C = O0.08 ± 0.010.06 ± 0.020.05 ± 0.010.02 ± 0.010.03 ± 0.02
C-NH20.08 ± 0.010.10 ± 0.020.09 ± 0.010.12 ± 0.010.06 ± 0.01
C-O-S0.00 ± 0.000.00 ± 0.000.00 ± 0.000.00 ± 0.000.00 ± 0.00
OM/OC1.35 ± 0.021.36 ± 0.041.31 ± 0.021.35 ± 0.011.27 ± 0.03
 Group 6Group 7Group 8Group 9Group 10
Alkane0.65 ± 0.030.51 ± 0.020.56 ± 0.030.47 ± 0.040.44 ± 0.05
Alkene0.20 ± 0.030.28 ± 0.030.23 ± 0.040.22 ± 0.020.24 ± 0.04
Aromatic0.08 ± 0.020.09 ± 0.010.09 ± 0.030.17 ± 0.040.07 ± 0.02
C-OH0.01 ± 0.010.02 ± 0.010.03 ± 0.020.05 ± 0.010.04 ± 0.01
C = O0.03 ± 0.010.04 ± 0.020.04 ± 0.010.01 ± 0.010.05 ± 0.01
C-NH20.04 ± 0.010.05 ± 0.020.05 ± 0.010.07 ± 0.010.16 ± 0.03
C-O-S0.00 ± 0.000.00 ± 0.000.00 ± 0.000.00 ± 0.000.00 ± 0.00
OM/OC1.29 ± 0.021.27 ± 0.021.29 ± 0.031.31 ± 0.021.49 ± 0.28

[29] The composition of group 5, which includes only back trajectories that spent at least 5 days over the ocean before sampling, is a reference marine background composition with submicron aerosol mass composed of 25% OM, 31% sulfate, 8% ammonium, 5% nitrate and 31% dust elements (Table 2). The OC in group 5 is composed largely of alkane carbon (65%) with an additional 18% alkene carbon, 8% aromatic carbon, and 3% carbonyl carbon (Table 3). This background marine air has a CO/OC slope of 9.9 (R2 = 0.33), which is between the CO/OC slopes of biomass burning and wood burning. Marine CO/OC slopes are also within 15% of the modeled value for total Asian emissions, consistent with an aerosol that reflects the regional background emissions rather than being dominated by local sources.

[30] Compared to this background, larger carbonyl carbon fractions of OC (6–9%) were observed in the North Asian air masses associated with groups 1–3. With increasing CO/OC slopes, the alkane carbon fraction of submicron OC decreases from ∼60% (CO/OC < 13) to ∼50% (CO/OC >15), with alkene and aromatic carbon fractions correspondingly increasing. Alcohol groups are present at elevated concentrations in groups 7–10 (CO/OC >15), along with elevated nitrate and ammonium fractions of submicron mass. These correlations suggest that a less saturated OC (alkene and aromatic as opposed to alkane) and elevated levels of alcohol, ammonium, and nitrate are all associated with combustion sources in the transportation sector that have large CO/OC emission ratios. Consistent with this observation, unsaturated OC comprises more than 40% of the chemically-resolved OC emitted from diesel trucks and noncatalyst-equipped automobiles but less than 5% of the chemically-resolved OC emitted from Asian biomass burning [Schauer et al., 1999, 2002; Sheesley et al., 2003]. Alcohols have not been measured in significant quantities in motor vehicle exhaust, making the identity of the high-CO/OC-ratio alcohol source unclear. The measured alcohol groups may indicate the presence of secondary organic aerosol, or may indicate that alcohols comprise a significant portion of the >95% of OC emitted from motor vehicles that cannot be resolved by chromatographic techniques.

[31] To illustrate the sample compositions and source areas associated with each trajectory group, and to analyze deviations from the general trends, case studies of four individual filter samples are described in detail: Hokkaido emissions over rough seas (group 1), a large Takla Makan dust layer (group 3), Shanghai emissions (group 7), and a dust and pollution mixture over the Yellow Sea (group 10).

3.4. Case Studies

[32] Rough seas and pollution from Hokkaido were observed during a 285m level leg from 3:52 to 4:47 GMT on 20 April 2001 (research flight 11, group 1, Figure 6). The CO/OC slope of 7.2 for this group is similar to that of biomass burning sources. Isentropic back trajectories show that the air mass had been over the ocean for less than 4 hours before sampling, and that the sampled air mass was at an altitude of less than 150m when it was over land. This sample contained low concentrations of OC and OM (1.9 ± 0.3 and 2.5 ± 0.4 μg m−3, respectively, with an OM/OC ratio of 1.35). Carbonyl carbon comprised 4% of OC, lower than the average of 6% for the ACE-Asia project, making this sample an outlier from the other samples in group 1 (see Table 3). The lower carbonyl fraction for this sample may be related to the local volcanic influence that was not present for the other samples of group 1. The average PILS-determined NH4+ to SO42− molar ratio for this sample was 1.9 (with a minimum of 1.4 from 4:12 to 4:15 GMT), smaller than the value of 2 in ammonium sulfate. Cl concentrations were lower than average at 0.01 μg m−3, suggesting that most of the sea salt was in supermicron particles.

Figure 6.

Vertical profiles of aerosol components on 20 April 2001 (research flight 11) and their solubility characteristics. Samples were collected between 34.5° and 42.7°N and 133.2° and 142.4°E. The sample composition at 285 m is associated with the Hokkaido case study. Sulfate, ammonium, and nitrate were identified with PILS, and all other functional groups were identified with FTIR. PILS data show average values and variability during the FTIR sampling times. For FTIR data, from left to right, black areas represent the residual fraction, dark gray areas represent the fraction removed in hexane, medium gray areas represent the fraction removed in dichloromethane, light gray areas represent the fraction removed in acetone, and white areas represent the fraction removed in water. Also shown are total condensation nuclei (CN) concentration and potential temperature.

[33] During a 450m level leg from 3:50 to 4:30 GMT on 11 April 2001 (research flight 6, group 3, Figure 7), the C-130 encountered the highest dust levels of the entire field project. The isentropic back trajectory suggests that the sampled aerosol had been over the ocean for 16 hours before sampling, and that the sampled air mass passed over the Takla Makan desert at 1500m 48 hours prior to sampling. This leg contained 52 ± 0.5 μg m−3 of SiO44−, 6.2 ± 2.9 μg m−3 of carbonate carbon, 3.5 ± 0.4 μg m−3 of Al, 2.3 ± 0.3 μg m−3 of Ca, 1.5 ± 0.2 μg m−3 of K, and 2.5 ± 0.1 μg m−3 of Fe in the submicron mode, indicating large amounts of mineral dust. These large concentrations of dust species from the Takla Makan desert are characteristic of all of the samples in group 3, comprising an average of 49% of submicron mass. As for group 1, the CO/OC slope of 7.5 for this group is similar to that of biomass burning sources. OC and OM, 14.2 ± 3.1 and 19.0 ± 3.5 μg m−3 respectively, were near the upper limit of values observed during ACE-Asia, and the OM/OC ratio was 1.34. The PILS-determined sulfate to ammonium molar ratio was 0.56, consistent with ammonium sulfate. The ratio of Si to Al was 1.7, considerably smaller than the average value of 2.9 for PM10 and PM2.5 reported by He et al. [2001], Winchester et al. [1981], and Zhang et al. [1993] but similar to the value of 1.9 reported by Zhang et al. [2003] for PM2.5 in Zhenbeitai on 11 April 2001. Zhenbeitai is a site near a major dust source region in China that is along the back trajectory of the air sampled during ACE-Asia. Al is the best tracer for clay minerals [Gomes and Gillette, 1993]. A smaller Si to Al ratio represents an enhancement of the clay fraction of the mineral aerosol in the ACE-Asia samples.

Figure 7.

Vertical profiles of aerosol components on 11 April 2001 (research flight 6) and their solubility characteristics. Samples were collected between 33.6° and 36.1°N and 124.3° and 128.5°E. The sample composition at 450 m is associated with the dust case study. The format is the same as in Figure 6.

[34] During a 250m level leg from 5:22 to 5:50 GMT on 30 April 2001 (research flight 16, group 7, Figure 8), Shanghai emissions were sampled by the C-130. This group has an intermediate CO/OC slope of 14.8 that is between the emissions ratios of the transportation sector and wood burning sources. The isentropic back trajectory reveals that the Shanghai plume had been over the ocean for ∼20 hours, with meteorological observations that indicated precipitation during this transit. The sampled air mass was at approximately 500m over Shanghai, within the well-mixed lowest layer of the atmosphere. This leg contained a condensation nuclei (CN) concentration of 5500 cm−1 and the highest submicron scattering during the project (270 Mm−1), as well as the sixth largest measured OM concentration (16.3 ± 4.8 μg m−3). The Shanghai emissions were associated with elevated nitrate concentrations (0.4 μg m−3) and an NH4+ to SO2−4 molar ratio of 0.57, consistent with ammonium sulfate. Particulate nitrate, ammonium, and sulfate can be formed as automobile emission products [Schauer et al., 2002], and the high CO/OC slope is consistent with influence from transportation sources. The OM/OC ratio of 1.24 measured in the Shanghai plume is indicative of organic compounds that have not been oxidized in the atmosphere. There were no significant tracers of sea-salt or dust sources in the Shanghai plume, with SiO44−, Al, Ca, and Cl all being below detection limits.

Figure 8.

Vertical profiles of aerosol components on 30 April 2001 (research flight 16), and their solubility characteristics. Samples were collected between 23.5° and 33.5°N and 124.2° and 131.9°E. The sample composition at 250 m is associated with the Shanghai plume case study. The format is the same as in Figure 6.

[35] Dust and pollution in the Yellow Sea were sampled at 490 m from 5:43 to 7:00 GMT on 12 April 2001 (research flight 7, group 10, Figure 9). Of the group 10 samples, this sample was collected furthest from Qingdao and has nearly a background concentration of CO (103 ppbv, see Figure 5). Back trajectories for this sample show that the air parcel had been near the surface of the Yellow Sea for approximately 28 hours before being sampled by the C-130. This aging under marine conditions is consistent with the larger observed OM to OC ratio of 2.0 for this sample (OC = 1.1 ± 0.2 μg m−3, OM = 2.2 ± 0.2 μg m−3). The large alcohol concentrations characteristic of group 10 were seen in this sample (10% of OC) along with high carbonyl carbon concentrations (8% of OC). Significant amounts of silicate and carbonate were also observed (4.5 ± 0.1 μg m−3 and 0.3 ± 0.1 μg m−3, respectively), indicating some contribution from a dust source. Cl was below detection limit.

Figure 9.

Vertical profiles of aerosol components on 12 April 2001 (research flight 7) and their solubility characteristics. Samples were collected between 33.1° and 34.9°N and 124.3° and 130.0°E. The sample composition at 490 m is associated with the Yellow Sea case study. The format is the same as in Figure 6.

3.5. Solubility of Measured Components

[36] Inorganic and organic submicron aerosol compositions and solubility characteristics for research flights 6, 7, 11, and 16 are shown in Figures 69. While the majority of inorganic ions were either very soluble in water or completely insoluble, the organic aerosol fraction exhibited a much wider range of solubility behavior including some partially soluble compounds.

[37] Inorganic species behaved as expected from the rinsing of standard compounds in the laboratory, with silicate and carbonate remaining on the filter after all rinses were performed. An exception to this trend was observed during flights 11 and 16 when large amounts of the silicate peaks were removed in hexane, acetone, and water. Less than 0.01 ng of some types of silicate, including sheet silicate, is expected to be removed from each Teflon filter during rinsing with 3 ml water [Nagy, 1995]. Differences in silicate rinsing behavior can be explained by an enhanced surface area which may be associated with small silicate-containing particles in some samples [Brantley et al., 1999; White and Brantley, 1995] or by differences in the physical form of the measured silicate between samples (for example, NaO4SiO4 is very soluble in water).

[38] Sulfate, ammonium, and nitrate show the same profile with altitude, as seen in Figures 6, 7, 8, and 9. Silicate and carbonate were also correlated (R2 = 0.77), suggesting similar sources or source regions of each. Elevated levels of both silicate and carbonate from 400 to 1000 m during flights 6, 7, and 11 demonstrate that elevated dust levels are often found at the top of the boundary layer in the ACE-Asia sampling region.

[39] Despite the variability in solubility characteristics of the organic aerosol fraction, aromatic and carbonyl groups maintained consistent solubility characteristics throughout the ACE-Asia project. On average, 68 ± 26% of the quantified aromatic C-H groups are found in the residual fraction, the fraction remaining on the filter after all rinses (Figures 69). Large organic molecules containing many carbons per functional group (i.e., low OM/OC) are most likely to remain in the residual fraction, suggesting that for the average ACE-Asia aerosol 68% of aromatic C-H bonds are associated with relatively large and insoluble organic molecules. Another feature of the data is that a small fraction of the carbonyl group often remains in the residual fraction. Poorly water-soluble carbonyl groups are present, perhaps associated with the identified residual aromatic compounds.

[40] During the large dust event of 11 April, alkane groups were removed in water while alkene, alcohol, amine, and organosulfur groups were removed in hexane. Carbonyl, aromatic, and alkene groups were all below detection limits. Alkanes require a polar functional group to be water soluble. The rinsing behavior suggests that the water soluble alkane groups are not associated with alcohol groups, but they may be associated with carbonyl groups which were not concentrated enough to be measured. The large alcohol signature is unique to flight 6, and may be characteristic of the dust source because alcohols can be a product of biodegraded vegetal detritus [Alves et al., 2001]. A common source and an internal mixture of dust and alcohols could explain the removal of a fraction of silicate in the hexane rinse. The Shanghai plume contained water-soluble alkanes like the dust layer of 11 April. These Shanghai samples also contained water-soluble carbonyl groups but lacked aromatic, alcohol, or organosulfur signatures.

4. Conclusions

[41] FTIR and TOT measurements produced OC values that were related with a slope of 0.91 and an R2 value of 0.93, suggesting that the two methods quantified very similar groups of species as OC. The CO/OC slope correlated to aerosol source region, with measured slopes comparing well with the source characteristics of an emissions inventory [Streets et al., 2003]. OC composition varied with the CO/OC slope, with CO/OC slopes above 15 associated with increased fractions of alcohol groups, unsaturated C-H groups, and inorganic nitrate. These functional groups are all associated with large CO/OC ratio emissions, most likely from the transportation sector.

[42] These results demonstrate the ability of a primary combustion tracer (CO), OM/OC ratios, functional groups, and back trajectories to distinguish among different source types and to identify secondary organic and inorganic aerosol formation. The types of aerosol categories also indicate consistency with emissions inventories. Three of the back trajectory groupings (groups 1–3), all containing trajectories that originated over northern Asia, had lower measured CO/OC slopes than predicted. The excess OC and moderately increased OM/OC ratio may be evidence of secondary OC formation, although uncertainties in back trajectories or emissions inventories cannot be ruled out.

[43] Examples of the vertical distribution of chemical components illustrated that inorganic composition can be understood by examining source characteristics. Elevated dust from the Takla Makan desert, elevated nitrate and ammonium sulfate downwind of Shanghai, and elevated sulfate downwind of Hokkaido are all explained by local sources. Dust sources are spatially distinct from CO sources and cannot be predicted by the CO/OC slope.

[44] CO/OC slopes, OM/OC ratios, and functional group information together provide a method of classifying atmospheric aerosol samples into source-based categories. The correlation between CO/OC slope and organic composition allows for the approximation of organic aerosol properties, such as hygroscopicity, based on the value of the CO/OC slope. These approximations can be useful for initializing aerosol hygroscopicity models or for achieving aerosol mass closure, because the CO/OC slope is a routine measurement that is often available when measurements of further organic aerosol characteristics are not.


[45] This research is a contribution to the International Global Atmospheric Chemistry (IGAC) Core Project of the International Geosphere Biosphere Program (IGBP) and is part of the IGAC Aerosol Characterization Experiments (ACE). Support was provided by NSF grants ATM-0002035 and ATM-0002698 and by NASA grant NAG5-8676. We are grateful to the NCAR Research Aviation Facility for their help in the field. We appreciate the help of Tim Bertram in collecting and analyzing the TOT samples and the help of Adam Reff in maintaining the aerosol generation system used for collection of reference compounds. We are also grateful to the NIEHS Center of Excellence at EOHSI for use of the weighing facility and FTIR spectrometer.