Environ. Sci. Technol. 2009, 43, 3448–3454
A Comparison of Summertime Secondary Organic Aerosol Source Contributions at Contrasting Urban Locations E L I Z A B E T H A . S T O N E , † J I A B I N Z H O U , †,§ DAVID C. SNYDER,† ANDREW P. RUTTER,† MARK MIERITZ,‡ A N D J A M E S J . S C H A U E R * ,†,‡ Environmental Chemistry and Technology Program, University of Wisconsin, 660 N. Park St, Madison, Wisconsin 53706, Wisconsin State Laboratory of Hygiene, Madison, Wisconsin 53718, and School of Resources and Environmental Engineering, Wuhan University of Technology, Wuhan, P.R. China 430070
Received September 5, 2008. Revised manuscript received December 17, 2008. Accepted March 24, 2009.
Primary and secondary sources contributing to atmospheric organic aerosol during the months of July and August were quantitatively assessed in three North American urban areas: Cleveland, Ohio, and Detroit, Michigan, in the Midwest region and Riverside, California, in the Los Angeles Air Basin. Organic molecular marker species unique to primary aerosol sources and secondary tracers derived from isoprene, R-pinene, β-caryophyllene, and toluene were measured using gas chromatography-mass spectrometry. Source contributions from motor vehicles, biomass burning, vegetative detritus, and secondary organic aerosol (SOA) were estimated using chemical mass balance (CMB) modeling. In Cleveland, primary sources accounted for 37 ( 2% of ambient organic carbon, measured biogenic and anthropogenic secondary sources contributed 46 ( 6%, and other unknown sources contributed 17 ( 4%. Similarly, Detroit aerosol was determined to be 44 ( 5% primary and 37 ( 3% secondary, while 19 ( 7% was unaccounted for by measured sources. In Riverside, 21 ( 3% of organic carbon came from primary sources, 26 ( 5% was attributed to measured secondary sources, and 53 ( 3% came from other sources that were expected to be secondary in nature. The comparison of samples across these two regions demonstrated that summertime SOA in the Midwestern United States was substantially different from the summertime SOA in the Los Angeles Air Basin and indicated the need to exert caution when generalizing about the sources and nature of SOA across different urban areas. Furthermore, the results of this study suggest that the contemporary understanding of SOA sources and formation mechanisms is satisfactory to explain the majority of SOA in the Midwest. Additional SOA sources and mechanisms of formation are needed to explain the majority of SOA in the Los Angeles Air Basin.
* Corresponding author e-mail:
[email protected]; fax: 608-2620454. † University of Wisconsin-Madison. § Wuhan University of Technology. ‡ Wisconsin State Laboratory of Hygiene. 3448
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 10, 2009
Introduction Particles in the atmosphere can negatively impact human health and cause degradation in visibility and perturbations to climate. Roughly half of fine particles, those with aerodynamic diameter less than 2.5 µm, consist of organic aerosols that come from primary or secondary sources. Primary organic aerosols are emitted directly from motor vehicles, biomass burning, fossil fuel combustion, and other sources. Secondary organic aerosols (SOA) are formed in the atmosphere from the photoxidation of gas-phase species to form less volatile compounds that can condense to form new particles or coagulate with existing particles. Contemporary research has targeted methods of identifying both primary and secondary sources and quantifying their relative strengths. Source apportionment of organic aerosol can be useful in assessing sources in polluted locations where ambient aerosol concentrations exceed regulatory levels designed to protect human health. Molecular marker-chemical mass balance (CMB) modeling is a technique successfully applied to various ambient atmospheres worldwide for the apportionment of primary organic aerosol (1-3). This technique draws upon highly specific and atmospherically stable organic species which are consistently produced by primary sources (3, 4). While primary source emissions are relatively well understood, the comprehension of secondary aerosol sources has proven more difficult. Gas-phase precursors to SOA come from a mixture of biogenic and anthropogenic sources. The ability of a precursor to form SOA depends on its atmospheric abundance, chemical reactivity, and the gas-particle partitioning of its oxidation products (5). A recently introduced paradigm suggested that semivolatile organic compounds emitted from primary aerosol sources rapidly photoxidize in the gas phase and then strongly partition to the aerosol phase and account for the majority of SOA (6). However, other studies have shown that individual volatile organic compounds of biogenic and anthropogenic origin are important to SOA formation in chamber and field experiments. Edney and co-workers have identified particlephase derivatives from precursors like isoprene, R-pinene, and β-caryophyllene and toluene in smog chamber experiments, where precursors were individually subjected to irradiation, NOx, and seed aerosol, and SOA products were identified using derivitization-GCMS (7-10). The organic species identified in the particle-phase were some of the first known tracers of SOA and have been demonstrated to be effective in quantifying reasonable SOA yields and contributions (1, 11). This study uses CMB source apportionment based on both primary molecular markers and SOA tracers in two distinct regions in the United States: the Midwest and the Los Angeles (LA) Air Basin. Samples discussed in this paper were collected during the summertime, which is associated with atmospheric conditions that are favorable for SOA formation (1, 11-13). The results of this study are analyzed in the context of previous studies of organic aerosols in these locations. Parallel chemical analysis of organic aerosols from contrasting locations provides a basis for comprehending local characteristics versus the regional extent of SOA while examining the state-of-the-science understanding of SOA sources and formation mechanisms in geographically distinct locations. 10.1021/es8025209 CCC: $40.75
2009 American Chemical Society
Published on Web 04/21/2009
Methods Sample Collection. Fine particulate matter (PM2.5) was collected using a medium-volume sampler (URG Corp., Chapel Hill, NC) with a Teflon-coated aluminum cyclone inlet that selected particles with an aerodynamic diameter of 2.5 µm or smaller. Air flowed at a rate of 92 L per minute, was controlled by a needle valve, and was measured with a calibrated rotameter. Particles were collected on 90 mm quartz fiber filters (QFF) that were prebaked at 550 °C for a minimum of 15 h. Cleveland, Ohio. PM2.5 was collected every 24 h from 14 to 22 July 2007 beginning at approximately 0900 at three locations in and around Cleveland, OH. Site 1 (41°27′47.02′′N, 82°06′52.05′′W) was located in a rural area with a heavy industrial corridor; site 2 (41°29′38.26′′N, 81°40′42.80′′W) was located near major roadways in a heavily industrialized area; and site 3 (41°28′36.99′′N, 81°40′55.01′′W) was in a residential area with heavy industry. During this study, the average OC contribution to PM2.5 mass was 28%. The Cleveland sites were within the 24 h average PM2.5 and 1 h ozone National Ambient Air Quality Standards (NAAQS) set by United States Environmental Protection Agency. Table S1 in the Supporting Information summarizes other PM2.5 parameters (mass, sulfate, OC, EC) and meteorological information (temperature, relative humidity, ozone, and nitrous oxides concentrations). Detroit, Michigan. Samples were collected at three locations in and around Detroit, MI Site 1 (42°13′43.03′′N, 83°12′30.00′′W) was in a residential area; site 2 (42°18′24.05′′N, 83°08′56.05′′W) was located in Dearborn, MI in a residential area and close to a large industrial corridor; and site 3 (42°20′06.04′′N, 83°06′34.97′′W) was located in a residential area 2 miles west of downtown and major roadways. The timing of the samples collected in Detroit was parallel to that which occurred in Cleveland. In Detroit, the average OC contribution to PM2.5 mass was 40%, and these sites were also within attainment of the EPA NAAQS for PM2.5 and ozone. Riverside, California. Sample collection occurred at the Air Pollution Research Center on the campus of the University of California at Riverside (33°58′18.40′′N, 117°19′21.41′′W) during the Study of Organic Aerosols Riverside (SOAR) field campaign in an area with heavy motor vehicle traffic. Samples were collected from 26 July through 7 August 2005 during the following times: 0500-1000, 1000-1500, 1500-2000, and 2000-0500 and are presented in this study as 24 h averages. Weekend samples included Saturday and Sunday, whereas weekday samples were limited to Tuesday through Thursday. On average, OC contributed to 40% of PM2.5 mass. The sampling location was within the EPA NAAQS for 24-h average PM2.5 and 1 h average ozone but was out of attainment for 8 h average ozone standard for several days in this study. Chemical Analysis. Elemental carbon (EC) and organic carbon (OC) in ambient aerosol were measured by a thermaloptical analyzer (Sunset Laboratory, Forest Grove, Oregon) following the ACE-Asia base case method (14). Solventextractable organic compounds were isolated from ambient aerosol and measured using gas chromatography mass spectrometry (GC-6980, quadrupole MS-5973, Agilent Technologies). This analysis targeted molecular marker and SOA tracer species. QFF containing PM2.5 were spiked with internal recovery standards (13) prior to extraction that included benz(a)anthracene-D12, triacontane-D62, dotriacontane-D66, cholestane-D4, levoglucosan-13C, and ketopinic acid. All of the composite samples analyzed as part of this study are summarized in Table S2 in the Supporting Information. Organic species were extracted from filters by sonication for quantification of levoglucosan and SOA tracer compounds, and by Soxhlet for quantification of other molecular markers. The sonication extraction consisted of extracting QFF using two 20 mL aliquots of dichloromethane (99.9%, Fisher
Scientific) and two 20 mL aliquots of methanol (99.9%, Fisher Scientific) for ten minutes each in a sonicator. The Soxhlet method consisted of extracting QFF with 250 mL of dichloromethane followed by 250 mL of methanol for 24 h each. Extracts were reduced by rotary-evaporation and evaporation under high-purity nitrogen. The sonicate-extracts were divided in two parts. One set of aliquots directly underwent silylation derivatization using N,O-bis(trimethylsilyl)trifluoroacetamide with 1% trimethylchlorosilane (Fluka, Switzerland), which exchanged hydroxyl for trimethylsilyl groups (15). These samples were analyzed by electron-impact (EI) GCMS in the negative mode, and levoglucosan was identified using a pure standard normalized to isotopically labeled levoglucosan-13C6. The second aliquot of sonicate-extracts were blown to dryness, reconstituted in pyridine, and then silylated with the reagent as described above (11). These samples were analyzed by positive chemical-ionization (PCI) GCMS for SOA tracer compounds using the instrument conditions and key ions described by Kleindienst et al. (11). A labeled chromatogram resulting from the PCI-GCMS analysis is shown in Figure S1 in the Supporting Information. At the time of this study, SOA tracers included pinonic acid, pinic acid, 3-methyl-1,2,3-butanetricarboxylic acid (16), and compounds tentative tentatively identified as 2-methylthreitol, 2-methylerythritol, 3-hydroxyglutaric acid, 3-acetyl hexanedioic acid, β-caryophyllinic acid, and 2,3-dihydroxy4-oxopentanoic acid. Nine SOA tracer species are reported in this study, compared to thirteen tracers reported in a previous study in the Midwest (1) and fifteen in the Southeastern United States (11). These differences in the number of SOA tracers observed may have stemmed from differences in analytical detection between this study and pioneering studies. For most SOA tracers, pure standards were not available at the time of this study and quantification was based on pinonic acid, which was quantified using an external calibration curve normalized to ketopinic acid, an approach similar to previous studies of molecular markers (17), but different from Kleindienst et al. (11) who used ketopinic acid for an internal standard and a surrogate for quantification. The Soxhlet-extract was analyzed by EI-GCMS to measure benzo(b)fluoranthene, benzo(k)fluoranthene, benzo(e)pyrene, nonacosane, triacontane, hentriacontane, dotriacontane,tritriacontane,17R(H)-22,29,30-trisnorhopane,17β(H)21R(H)-30-norhopane, and 17R(H)-21β(H)-hopane. These compounds were quantified using pure standards and normalized to isotopically labeled internal standards. The uncertainty of the measurement of all organic species was a propagation of the standard deviation of the field blank and twenty percent of the analytical measurement, based on recoveries of analytes in standard samples. Source Apportionment. Primary and secondary source contributions to organic aerosol were calculated using chemical mass balance (CMB) software available from the United States Environmental Protection Agency (EPA, CMB version 8.2). The model solved the effective-variance leastsquares solution to the linear combination of sources and their relative contributions to ambient OC for the organic molecular marker compounds observed in ambient aerosol (18). The model assumed that molecular marker species are atmospherically stable during transport for source to receptor and contributing sources were well-characterized (3). These assumptions have proven to be reasonable in previous CMB studies as demonstrated by agreement between CMB results and colocated measurements or source reconciliation efforts (1, 2, 12, 19). The stability of SOA tracer species was not characterized at the time of this study, and for the purpose of source apportionment tracers were assumed to be atmospherically stable. VOL. 43, NO. 10, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3449
TABLE 1. Summary of Source Profiles, Their Chemical Characteristics, and References aerosol source
chemical characteristics
diesel engines gasoline motor vehicles noncatalyzed engines biomass burning
elemental carbon, hopanes, PAH hopanes, PAH elemental carbon, hopanes, PAH levoglucosan, PAH
vegetative detritus
C29-C33 n-alkanes with odd carbon preference 2-methylthreitola, 2-methylerythritola 3-hydroxyglutaric acida, pinic acid, 3-acetyl hexanedioic acida, 3-methyl-1,2,3-butane tricarboxylic acid, pinonic acid β-caryophyllinic acida 2,3-dihydroxy-4-oxopentanoic acida
isoprene-derived SOC R-pinene-derived SOC
β-caryophyllene-derived SOC toluene-derived SOC a
Gasoline-Diesel Split Study (21) Gasoline-Diesel Split Study (21) Gasoline-Diesel Split Study (21) Midwest and West Coast Regional Averages (20, 23) Leaf Abrasion (22) Isoprene Chamber Study (11) R-Pinene Chamber Study (11)
β-Caryophyllene Chamber Study (11) Toluene Chamber Study (11)
Indicates tentative compound identification.
The input source profiles are listed in Table 1 and represent the most relevant and current profiles available at the time of the study and are summarized elsewhere (20). For the apportionment of OC in the Midwest, the primary input sources were mobile sources (21), including diesel engines, gasoline motor vehicles, noncatalyzed engines; vegetative detritus (22); and biomass burning. Noncatalyzed engines refer to gasoline engines for which catalytic converters do not properly function or are not installed. The biomass burning profile was an average of Midwestern United States tree species (20, 23). In the study of OC sources in Riverside, primary source inputs included diesel engines, gasoline motor vehicles, and biomass burning, based on an average of tree species in the West Coast of the United States (20, 23). Secondary source contributions to OC were estimated using fixed tracer-to-OC ratios, which were 0.15 ( 0.4 for isoprene, 0.21 ( 0.11 for R-pinene, 0.023 ( 0.005 for β-caryophyllene, and 0.0079 ( 0.0026 for toluene (11), and secondary tracer species were assumed to be unique to their precursor gas with a high degree of certainty. A boundary analysis was conducted to test possible range of secondary source contributions based on the range of data from Kleindienst et al. (11) and possible contributions range from 70-130% for isoprene, 50-220% for R-pinene, 70-120% for β-caryophyllene, and 60-160% for toluene. The standard deviation of the SOA tracer-to-OC ratios across chamber studies was found to be suitable as a source profile uncertainty since it accounted for the more-atmospherically relevant lower boundary SOA estimates. Model results were considered acceptable if the correlation coefficient (R2) values were greater than 0.75 and if the ratio of calculated to measured concentrations of molecular marker species agreed within 20%. The difference between the sum of the calculated sources and the measured OC is referred to as unapportioned or “other” OC. For two samples from Detroit, the sum of the calculated source contributions exceeded the measured OC; other sources were not statistically different from zero and these source contributions were normalized to the calculated OC. The uncertainties of source contributions at a single site correspond to the standard error calculated by the model that is based upon analytical uncertainty of the ambient measurements and source profiles and their quality of fit. The average source contributions for an urban area are presented with the standard error of the average.
Results Carbonaceous Aerosol. The ambient concentrations of carbonaceous aerosol components OC and EC are presented in Table S3 in the Supporting Information. The observed 3450
reference
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 10, 2009
concentrations of OC in Cleveland ranged from 2.70-3.60 µg C m-3 and EC ranged from 0.23-0.58 µg C m-3. In Detroit, OC concentrations ranged from 2.87-4.39 µg C m-3, while EC ranged from 0.21-0.67 µg m-3. Riverside had higher concentrations of carbonaceous aerosol, with OC concentrations ranging from 6.78-7.75 µg C m-3, while EC levels ranged from 0.70-1.19 µg C m-3. The ratio of OC to EC in Cleveland, Detroit, and Riverside was 8.3, 9.8, and 9.6, respectively. Molecular Markers for Primary Aerosol Sources. Primary aerosol molecular markers are also presented in Table S3 in the Supporting Information. Levoglucosan, a widely used tracer for biomass combustion (4), indicated the presence of biomass burning aerosol in every sample. In Cleveland, levoglucosan concentrations ranged from 21-93n g m-3, in Detroit levels ranged from of 35-202 ng m-3, and in Riverside, concentrations were much lower at levels of 5-10 ng m-3. The observed levoglucosan concentrations in Riverside were low compared to the Midwestern sites, which are consistent with a minimal biomass burning impact on ambient PM2.5, also noted by Docherty et al. (12). A series of n-alkanes ranging from C29-C33 with an oddcarbon preference signifies contributions to ambient organic aerosol from modern vegetative matter whereas alkanes with no odd-carbon preference is consistent with geologically aged plant matter that is present in fossil fuels (24). In Midwestern samples, the odd-carbon signature for vegetative detritus was observed and ambient concentrations of n-alkanes ranged from below detection to 3.0 ng m-3. In the Riverside samples, only low concentrations of n-alkanes were observed, ranging from below detection to 2.3 ng m-3, with no discernible odd-carbon preference; this indicated that vegetative detritus had negligible contributions to organic aerosol. Hopanes, which are part of a homologous series of tricyclic terpenes, are found in petroleum and its combustion emissions (24). They are known to be associated with gasoline motor vehicles and diesel engines through the use of motor oil. At Midwestern sites, hopane concentrations ranged from 0.02-0.38 ng m-3 and in Riverside concentrations were 0.09-0.32 ng m-3. PAH are generally nonspecific tracers for the combustion of carbonaceous material. Benzo(b)fluoranthene, benzo(k)fluoranthene, and benzo(e)pyrene are produced differently by noncatalyzed engines versus catalyzed engines and can be used to quantitatively distinguish these two sources (21). PAH were observed in all of the Midwestern samples in concentrations ranging from 0.02-0.21 ng m-3 but were not quantifiable in Riverside. Molecular markers for natural gas and coal combustion were not observed in any samples in this study and these
FIGURE 1. Ambient concentrations of secondary organic aerosol (SOA) tracer species derived from isoprene (a), r-pinene (b), β-caryophyllene (c), and toluene (d). sources were not considered to be important contributors to organic aerosol. Cholesterol, historically used as a tracer for meat cooking, was not used as tracer in this study because it is believed to have multiple atmospheric sources (13, 19) and the samples discussed in this paper did not contain cholesterol above the analytical detection limit. Consequently, if meat cooking or any other unidentified source were important to ambient OC concentrations, they would be unapportioned by the CMB model and represented by the category other sources. Organic Tracers for Secondary Organic Aerosol. The concentrations of SOA tracer species are presented in Figure 1 and in Table S3 in the Supporting Information. Isoprene is emitted from plants and effectively forms SOA in the presence of acidic aerosol species (10). Two major photoxidation products of isoprene have tentatively been identified as, 2-methylthreitol and 2-methylerythritol, which were observed in all Midwestern samples at concentrations of 3.3-46ng m-3. These values are consistent with ambient observations made in North Carolina in the summertime (11) but are lower than levels observed in Detroit in the summer of 2004 by less than a factor of 2 (1). Only 2-methylthreitol was observed in Riverside, at concentrations of 6.0-6.7ng m-3. The absence of 2-methylerythritol in Riverside was unexpected since previous studies have
reported these isomers together (10). A possible explanation for the lower concentrations observed in the LA Basin was that aerosol in the Midwest was more acidic and conducive to isoprene-derived SOA formation, a spatial trend that is consistent with previous studies of sulfate deposition in the United States (5). R-Pinene is a monoterpene emitted from plants that was expected to be important to SOA formation (9). Five R-pinene tracers were observed in every sample analyzed as part of this study: pinic acid, pinonic acid, and three compounds tentatively identified as 3-hydroxyglutaric acid, 3-acetylhexanedioic acid, and 2-hydroxy-4-isopropyladipic acid. The sum of these compounds ranged from 38-60 ng m-3 in Cleveland, 36-57 ng m-3 in Detroit, and 30-61 ng m-3 in Riverside. β-Caryophyllene, a sesquiterpene emitted from biogenic sources, was found to yield particle-phase β-caryophyllinic acid (8) and in this study concentrations of this tracer ranged from 1.1-1.9 ng m-3 in Cleveland, 0.6-1.7 ng m-3 in Detroit, and 1.7-2.0 ng m-3 in Riverside. The concentrations of R-pinene and β-caryophyllene tracers are lower than summertime levels previously observed in Detroit (1) and in North Carolina (11), which may have been associated with differences in atmospheric conditions between this and previous studies, time-averaging (4-5 days in this study, versus a monthly averages), or quantitative analysis of tracers. VOL. 43, NO. 10, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3451
FIGURE 2. Measured primary, secondary, and other (unknown) source contributions to ambient organic carbon (OC) measured by chemical mass balance (CMB) modeling in absolute concentration (a) and by percent (b).
TABLE 2. Summary of Primary and Secondary Sources Included in the CMB Model, and Other Sources (Unknown) That Were Not Included in the Modela primary source contributions
secondary source contributions
other (unknown) source contributions
site
range (µgC m-3)
average (µg C m-3)
average % OC
range (µgC m-3)
average (µg C m-3)
average % OC
range (µgC m-3)
average (µg C m-3)
average % OC
Cleveland, OH Detroit, MI Riverside, CA
0.61–1.67 0.93–3.09 1.43–1.59
1.24 ( 0.15 1.22 ( 0.34 1.51 ( 0.08
37 ( 2% 44 ( 5% 21 ( 3%
0.93–1.94 1.20–1.77 1.45–2.44
1.47 ( 0.15 1.44 ( 0.10 1.94 ( 0.50
46 ( 6% 37 ( 3% 26 ( 5%
0.24–1.08 0.00–1.31 3.74–3.88
0.58 ( 0.15 1.44 ( 0.25 1.94 ( 0.07
17 ( 4% 19 ( 7% 53 ( 3%
a Results are presented as average mass contributions in µg OC m-3 across sites, weekdays, and weekends; and % of total organic carbon (OC). Uncertainties are the standard error of the average.
Toluene is an aromatic hydrocarbon abundant in the atmosphere that has formed SOA in chamber studies that contained a unique compound tentatively identified as 2,3hydroxy-4-oxopentanoic acid (7). Emissions of toluene to the atmosphere are linked to use of motor vehicles and was expected to be highest in urban areas. In this study, 2,3hydroxy-4-oxopentanoic acid was observed in Midwestern samples at levels of 5.1-10.1 ng m-3 and in Riverside at levels of 9.5-16.1 ng m-3. At Research Triangle Park in North Carolina, concentrations of this compound ranged from 2.0-6.9 ng m-3 in the months of July and August (11), and in Detroit, previous levels of 2.9-3.1 ng m-3 were measured in July and August (1). 3452
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 10, 2009
Source Apportionment of Ambient Organic Carbon. The contributions of primary, secondary, and other sources determined by the CMB model are shown in Figure 2. Results are summarized by urban area in Table 2 and listed by site in Table S4 in the Supporting Information. In Cleveland, primary sources contributions accounted for on average 37 ( 2% of ambient carbonaceous aerosol, with motor vehicles accounting for 26 ( 3% (including diesel engines [3 ( 1%], gasoline vehicles [7 ( 2%], and noncatalyzed engines [16 ( 3%]), vegetation producing
