Environ. Sci. Technol. 1993, 27,2045-2055
Mathematical Modeling of Urban Organic Aerosol: Properties Measured by High-Resolution Gas Chromatography Lynn M. Hlldemann,t Glen R. Cass,' and Monlca A. Mazureks Environmental Engineering Science Department and Environmental Quality Laboratory, California Institute of Technology, Pasadena, California 9 1125
Bernd R. T. Slmonelt College of Oceanic and Atmospheric Sciences, Oregon State University, Corvallis, Oregon 9733 1
t Present address: Department of Civil Engineering, Stanford University, Stanford, CA 94305-4020. 8 Present address: Environmental Chemistry Division, Building 426, Brookhaven National Laboratory, Upton, NY 11973.
make construction of the comprehensive emission inventories needed for model applications extremely difficult. (ii) Analytical Techniques. Hundreds of organic compounds typically are present in fine aerosol samples taken from both urban atmospheres and various emission sources. The few experimental programs which have involved detailed organicanalysesof such samplestypically have identified only a few specific compounds, like the polynuclear aromatic hydrocarbons (PAHs), which constitute at most a few percent of the total organic aerosol mass. In addition, enormous differences again exist in the way that various researchers have extracted and analyzed such samples (21,and the sensitivity and accuracy of the different organic analytical methods used in past studies are known to vary considerably (e.g., ref 5). (iii) Modeling Approaches. Typically, source-receptor relationships are sought using one of two approaches. In receptor-oriented modeling, stable compounds characteristic of particular source types are used as tracers to quantify the presence of a particular source's effluent in an ambient sample. In source-oriented modeling, atmospheric transport and dilution calculations are used to determine the combination of primary emissions that should be present at a particular ambient monitoring site. Use of PAHs in such models to determine source-receptor relationships has been suggested (6),but PAHs alone appear to be insufficient to distinguish between certain sources (e.g., diesel vs spark-ignition engine emissions) (7), while other sources contain very little PAH in their organic emissions (e.g., brake wear, tire wear, paved road dust) (8). Source reconciliation studies have been further complicated by the production of secondary organic aerosols from gas-phase precursors via atmospheric chemical reactions (9-11). Advanced models can be constructed that account for the processesthat form secondary organics (12).However, the predictions of these models are difficult to verify, because the detailed chemical composition of the reaction products that appear as secondary organic aerosols cannot yet be predicted. Hence, the extent to which such processes contribute to the ambient organic aerosol concentrations can be estimated, but the accuracy of such estimates is not currently known. In the present study, a systematic effort is undertaken to examine source/receptor relationships for organic aerosols based on source samples and ambient samples specifically collected for this purpose. A transport model previously developed for the Los Angeles Basin (13-15) is used to compute the mixture of organic aerosol that would be contributed at Pasadena, downtown Los Angeles, and West Los Angeles if primary organic aerosol emissions were transported from source to receptor without further chemical reactions. Those characteristics of the sources
0 1993 American Chemical Society
Envlron. Scl. Technol., Vol. 27, No. 10, 1993 2045
Primary fine aerosol emissions from a variety of urban sources have been quantitatively characterized via highresolution gas chromatography to obtain organic mass distribution fingerprints. To assess the degree of secondary organic aerosol formation in urban areas, a transport model is used to predict the distribution of ambient organic aerosol characteristics that would exist at various sites in the Los Angeles Basin if the primary organic emissions were transported without chemical reaction. Comparisonsbetween the model predictions and ambient measurements show substantial agreement for the nonpolar organics, suggesting that ambient concentrations of this organicfraction result directly from primary emissions. In contrast, ambient concentrations of fine acidic organic aerosols are significantly underpredicted by the model, indicating that secondary formation is important for acidic organics. On the basis of the observed differences between model predictions and measured properties of acidic organics, it is estimated,using monthly averages, that up to 18-27 % of the elutable organic aerosol present in the Los Angeles atmosphere may be secondary in origin.
Introduction Despite the prevalence of atmospheric organic aerosols in urban areas, progress in determining their exact origin has been slow. Systems used for sampling, analysis, and modeling of organic air pollutants vary greatly between investigators,making it difficult to interpret results based on the synthesis of large numbers of prior studies. In order to compare source and ambient samples via airquality modeling techniques, a variety of barriers must be overcome: (i) Sample Acquisition. Carbon-containing particles are emitted directly from more than 70 different types of air pollution sources (1). Large differences exist in the methodologies that have been used by different research groups to collect and store samples of this directly-emitted organic particulate matter, which is referred to as primary organic aerosol (2). In addition, some of the methods used to sample hot stack exhausts have been found to greatly underestimate the amount of primary organic aerosol present (3) and to bias the type of organic aerosol collected (4). Finally, for certain types of sources, there is a complete lack of information regarding emissions of organic particulate matter. These inconsistencies and inaccuracies
0013-936X193/0927-2045$04.00/0
of primary organics that can be discerned via highresolution gas chromatography (HRGC)analyses of source and ambient aerosol samples are used to examine model performance. Instead of undertaking the expensive and time-consumingprocess of analyzingfor individual organic species, the faster, more economical approach taken here is to subdivide the entire extractable organic aerosol which elutes via HRGC into 48 independent groups of organic compounds, for both the primary source emissions and the ambient samples (4). Each group of organic compounds emitted from a particular source is then quantified and tracked by the air-quality model. Calculations are conducted as if the emissions are nonreactive. Predictions of ambient concentrations based on the source emissions strengths and characteristics are compared to HRGC analyses of actual ambient fine organic aerosol samples,seekingdiscrepancies that could be used to quantify the effect of atmospheric transformations. All source samples (4,161 and ambient pollutant samples (I 7-20] are collected by comparable methods and analyzed by the same methods in the laboratory. As a result, comparisons between model predictions and atmospheric concentrations reflect atmospheric processes and not the differences in sampling or analytical methods from one study to the next.
Modeling Approach To predict the effect of primary organic aerosol emission sources on ambient organic characteristics at sites in Los Angeles, the air-quality model developed by Cass (13,14) and modified by Gray (15)is used. In this model, ambient aerosol concentrations due to primary emission sources are computed by a Lagrangian particle-in-celltechnique. Single mass points marked with the appropriate primary aerosol mass emission rate are released to the atmosphere at measured time intervals from each source in the air basin. Depending on the plume rise characteristicsof each source, the particles may be inserted either above or below the base of an elevated temperature inversion. Exchange of air parcels between the stable layer above the inversion base and the zone below the inversion base occurs as the inversion base height changes on an hourly basis. For air parcels inserted below the inversion base, three different regimes are used to define the vertical distribution of pollutant particles: (1) sufficiently near a point source, the inversion base has no influence on the concentration distribution of emissions from that source, and pollutant concentrations assume a Gaussian distribution in the vertical domain; (2) far from the source, the emitted particles become fully mixed in the vertical dimension beneath the inversion base; and (3) between these two extremes, the inversion base influences the concentration profile of the emitted airborne particles. In the horizontal domain, trajectories of successive particles released from a source form streaklines downwind of that source, computed from the time sequence of the hourly groundlevel wind speed and direction observations plus a simulation of the effect of horizontal diffusion. Particle losses due to dry deposition also are computed. The horizontal displacement of each particle at each hour then is paired with ita probable location in the vertical dimension, and organic aerosol mass concentrations are computed by summing the contributions of all particles residing within the ground-level layer above a matrix of 2046
Envlron. Scl. Technol., Vol. 27, No. IO, 1993
receptor cells that formsa grid over the air basin. Monthly average pollutant concentration increments, ( c ) , are computed for each source type at each receptor site, as follows:
T)
w(t
- r ) dT dt]
s(x’)dx’ (1)
where Q(x,tJx’,t - r ) is the transition probability density function that describes the probability that a pollutant particle will be found in the small air volume (Axl, Axz, Ax31 surrounding location x at present time t , given that it was released from location x’ at time t - T ; w ( t - 7 ) is the normalized diurnal variation in the emission rate for the source type of interest; S(x’)is the spatial distribution of the monthly averagepollutant emissionsfrom the source class of interest; t s is the starting time for the long time period over which concentrations will be averaged; and T is the length of that averaging time; in this case T equals 1 month. By repeating this calculation for each source class in the air basin and then superimposing the results onto an estimate of aerosol carbon background air quality, a multiple source urban air-quality model for long-term average primary aerosol carbon concentrations is obtained. Superposition is permitted because the primary aerosol carbon particle concentrations are modeled in a form that is linear in emissions. In the study by Gray (15),the mathematical model just described was applied to compute monthly average elemental carbon and total carbon aerosol concentrations over the western coastal plain of the Los Angeles Basin. Primary aerosol carbon emissionsfrom 70 classes of mobile and stationary sources were estimated for each of 625 grid cells located within the 80 X 80 km study area shown in Figure 1. The primary aerosol carbon increment due to each source type was computed for each month of the year 1982 based on hourly wind speed and direction data observed over central Los Angeles plus hourly mixing depths estimated from the morning and afternoon mixing depths reported by the South Coast Air Quality Management District. Concentration increments from all source types were summed and then added to the background aerosol carbon content of marine air measured upwind of the city at San Nicolas Island. Model predictions for elemental carbon concentrations were compared to ambient observations at seven monitoring sites in the study area, and good agreement between model predictions and ambient observations was found. The emissions data, modeling procedures, and comparison of model predictions to observations have been thoroughly documented (15),and model results at Pasadena, CA, have been published previously (21). In the present study, characteristics of the ambient primary organic aerosol that are observable by HRGC analysis are predicted by substituting the HRGC properties of the emissions source effluent into the source term S(x’)of the air-quality model just described. Source tests were conducted between 1986 and 1989 by dilution sampling and grab samplingfor the primary organicaerosol sources shown in Table I, providing samples that represent the major primary emission sources of fine (aerodynamic diameter 1 2 pm) organic aerosol to the Los Angeles
-SOUTH
CWST AIR YOIN
SOx80km HEAVILY
URBANIZED REOION
Flgure 1. South Coast Air Basin, showlng the 80 X 80 km modeling domaln centered over downtown Los Angeles.
atmosphere (see refs 3 and 16 for the source test program and mass emissionsrate data). These source samples were sealed in annealed glass jars with Teflon-lined lids and stored in the dark at -25 "C until the end of the source test program. The fine organic particulate emissions from various sources then were characterized by solvent extraction of the organicmaterial, followed by high-resolution gas chromatography (4).Two fractions of the organic extract were analyzed one fraction analyzed by HRGC without derivatization was considered to represent the emissions of neutral (nonpolar) compounds; a second fraction, derivatized to convert acidic components to methyl ester and methoxy analogues that will elute from the GC column used, also was injected to characterize the acid+neutral fraction. The difference between the results obtained from these two injections was considered to represent the acidic organic aerosol alone. The complex pattern of peaks representing hundreds of separate compounds emitted from each source was divided into well-definedgroups of compounds using the elution points of the c12-c36 n-alkane series as dividing lines, and the organicmass contributed by each group within each sample was calculated by summing the compound masses that eluted between each of the adjacent alkanes (4).In this manner, the extractable organics mass emission rate from each source was characterized in terms of the quantities of 48 independent groups of neutral and acidic compounds. Analytical uncertainties associated with source sample extraction and analysis have been published previously (4). Emission rate data then were supplied to the air-quality model for each group of compounds released from each major source type. Emission source activity data were drawn from Gray's inventory (15)(e.g.,the spatial distribution of kilometers driven per day by catalyst-equipped automobiles), and the extractable organic mass emission rates were quantified in the present study (e.g.,micrograms of fine-particle neutral organics eluting between the C I ~ and CU n-alkanes that are emitted per kilometer driven by catalyst-equipped automobiles). The model was used to compute the ambient fine-particle concentrations at Pasadena, downtown Los Angeles, and West Los Angeles that would be expected for each month of the year 1982 if primary emissions in each of the 48 independent groups
of organic aerosol compounds were transported without chemical reaction. An estimate of background fine organic aerosol concentrations for each group of organic compounds then was added to the source increments, based on characterization of the fine organic aerosol measured upwind of Los Angeles during 1982 at San Nicolas Island, located about 80 km west of Santa Catalina Island (which is shown in Figure 1). Filter samples of the San Nicolas Island fine aerosol collected by Gray and co-workers (17)were employed for this purpose. Due to the low levels of organic aerosol found there, filters collected over a period of several months were compositedto give adequate organicaerosol concentrations for HRGC analysis. The San Nicolas Island (SNI) composite containing filters collected during the months of July-September, 1982, was used in the model to characterize the background aerosol during April-September, while the composite consisting of filters collected during October-December was used to characterize the colder months of the year 1982. Ambient fine-aerosol samples for comparison to the airquality model predictions were composited using 24-h average filter samples collected at 6-day intervals during 1982 by Gray et al. (17). These samples were stored individually in sealed petri dishes and kept frozen in the dark prior to analysis. Monthly composites of ambient samples taken at Pasadena, downtown Los Angeles, and West Los Angeles plus the two composites from San Nicolas Island were extracted, derivatized, and quantified by HRGC, as described previously (18, 19). Organic compound concentrations falling into the 48 groups of neutral and acidic compounds defined between the elution points of the C12-C36 n-alkane series were computed for comparison to the predictions of the source-oriented airquality model. Results
A large number of attributes of the primary organic aerosol can be used to compare air-quality model results to ambient data. In this study, first the extent to which primary emissions of aerosol organic carbon can account for ambient organic carbon aerosol levels (as measured by combustion techniques) is examined. Next, sourceEnviron. Sci. Technol., Vol. 27, No. 10, 1993 2047
Table I. Major Sources of Fine Aerosol Organic Carbon Emissions within an 80 Surrounding Los Angeles (for 1982) source type meat cooking operations charbroiling frying paved road dust fireplaces soft wood hard wood forest fires cigarettes brake lining wear roofing tar pots tire wear dust structural fires other fugitive sources catalyst-equipped gasoline vehicles automobiles other vehicles noncatalyst gasoline vehicles automobiles other vehicles diesel vehicles heavy-duty trucks other vechicles residual oil-fired ships jet aircraft railroad (diesel oil) other mobile sources residual oil stationary sources distillate oil stationary sources industrial other refinery gas combustion natural gas combustion residential sources other sources coal burning surface coating organic chemical processes misc. petroleum industry processes primary metallurgical processes secondary metallurgical processes mineral industrial processes other organic solvent use asphalt roofing wood processing misc. industrial point sources other stationary sources total
organic carbon emitted (kg/day)a
X
80 km Heavily Urbanized Area
source tested
source used in model
Fugitive Sources 4938b*c*d 1393b*c*d 4728b
X' X
3332b2C 84ObpC 877 802b9c 690b 556b 414b 63 23 Mobile Sources
X X X X X X
780bvC 79 2088b>c 1372 1242b*c 617 66 92 211 52 Stationary Sources 206 13b,c 23 195
X
X
3ObvC 262 76 1433 692 278 228 167 158 106 81 74 393 152 29 822
X
X
a Except where otherwise noted, values are based on previous literature surveys (1, 15). b Fraction of organic carbon in fine emissions estimated during present study (16). Fine maas emission rate estimated duringpresent study (16). Mass emission rates include both commercial and domestic cooking. e Only commercial charboiling (40% of total) was included in the model.
receptor relationships are developed for that fraction of the primary organic aerosol that is solvent-soluble and elutes from a HRGC column under the conditions of our experiments. Finally, the ability of the model to account for the distributions of organiccompounds found in HRGC analyses of the ambient aerosol is tested. (i) Organic Carbon. In the study by Gray (151, airquality model predictions for ambient elemental carbon and total carbon concentrations were compared to ambient observations at seven sites in the Los Angeles area for each month of the year 1982. Ambient elemental and organic carbon levels were determined by a thermal evolution and combustion technique (22,23). Elemental carbon concentrations were closely reproduced by the model, as would be expected since fine-particle elemental 2048
Environ. Scl. Technol., Vol. 27. No. 10, 1993
carbon concentrations are due entirely to direct primary emissions to the atmosphere. Surprisingly, air-quality model predictions for total aerosol carbon also were in fairly close agreement with observed ambient concentrations. If elemental carbon concentrations are subtracted from total aerosol carbon concentrations, the results of Gray (15) also can be presented as modeled and measured ambient values of primary organic carbon aerosol (OC). Using the fine organic aerosol carbon inventory of Gray (15),Figure 2 shows the results of the model on a monthby-month basis for the three ambient sites of interest here. The vertical bars shown in Figure 2 represent f l u intervals about the monthly mean OC value; the principal contributor to those uncertainties arises not from analytical
25.0
- 22.5
I
4
L3
5 20.0 17.5 15.0
1
(a) MODELRESULT
-
PASADENA
MEASURED VALUE e
1
-
38 12.5 0
8
10.0
'
"E
7.5
-
1 T
b
2.5
-r IL
-
0.01
I
I
I
I
I
(b) MODELRESULT
I
I
-
I
I
I
I
DOWNTOWN L.A.
MEASURED VALUE e
2
1
1
"'
5.0- -
T
20.0
0 . 0 1 ,
I
I
(')
I
MODELRESULT MEASURED VALUE
I
I
I
I
I
I
I
WEST L.A.
8
20.0
4
17.5 15.0 4 15.0 12.5
E5-.
T
4
1
8 80 10.0 -
2 a5 73 5.0
2.5 0.0
.=I:!, 1
TI ? 1
(ii) Elutable Organics. A similar comparison can be presented for the fine organic particulate matter emissions that are extractable and elutable through the HRGC column under the conditions of our experiments. This provides a separate test of model performance because for some sources, like natural gas home appliance and roofing tar pot emissions, essentially all of the organic matter is extractable and elutable, while for other sources, such as brake wear debris, only a small fraction of the total fine organic aerosol extract is detected via HRGC analysis (4). The modeling study conducted by Gray (15) included emissions from more than 70 separate carbon particle source types, with emissionrates determined in many cases by literature review. When modeling the characteristics of the aerosol observed by HRGC techniques, only those sources contained in the model that were sampled during our source testing effort and subsequently analyzed using HRGC can be included in the modeled results; Table I lists such sources. The organic carbon emission inventory of Hildemann and co-workers (16)can be used along with the air-qualitymodel to calculate the fraction of the organic carbon that is represented by these measured sources. According to the inventory and the model, the combination of the background levels measured at San Nicolas Island plus the organic aerosol sources actually tested should account on an annual basis for 76% of the primary organic carbon at downtown Los Angeles, 79% at West Los Angeles, and 83% at Pasadena. If the source tests were applied to wider categories (for example, by using the compound distribution in natural gas home appliance emissions to represent all types of natural gas combustion, or the emissionsfrom heavy-duty diesel trucks to represent all diesel engines), the fraction of the primary organic carbon represented by GC traces within the model would total 88% at downtown Los Angeles, 89% at West Los Angeles, and 91% at Pasadena. Using the first, more conservative approach in which only those sources actually tested are represented within the model, model predictions for HRGC elutable organic aerosol are compared with ambient observations in Figure 3. In Figure 38-12, the neutral elutable organic material is plotted, while in Figure 3d-f the acid+neutral fraction of the organic aerosol is shown. The results for both of these fractions show reasonable agreement between the modeled and measured values, with a seasonal peak in the concentrations seen in the winter months for both the measurements and the model predictions. However,since the model includes only -80% of the source emissions, the predictions are judged to be slightly high. In Figure 3g-i, the results for the acidic fraction alone (computed as the difference between the neutral and acid+neutral fractions) are shown. For this fraction, the agreement between the modeled and measured values is poor. The model, based solely on transport of primary emissions, consistently and substantially underpredicts the concentrations of aerosol organic acids. This indicates that the acidic organics emitted as primary material from the measured sources are not nearly sufficient to account for the concentrations of acidic aerosol organics measured at these ambient sites. Even if all the organic material emitted by the unmeasured sources listed in Table I were in the form of elutable acidic organics, there would not be sufficient primary emissions to account for the large differences between the modeled and measured concentrations of ambient aerosol
I i
-
JAN FEB MAR APR MAY JUN JUL. AUG SEP OCT NOV DEC 1982
Flgure 2. Modeled vs measured monthly concentrations of total fine organic carbon at Los Angeles Basin ambient sites: (a) Pasadena, (b) downtown Los Angeles, and (c) West Los Angeles.
error (4) but from day to day variability in ambient concentrations combined with the fact that 5-6 days were sampled out of each month. If f 2 a is visualized as an approximationto a 95% confidenceinterval on the ambient data, then the transport model results can be viewed as being statistically insignificantly different from the ambient data at most times. The seasonal variations measured, with peak levels in winter months, are reproduced fairly well by the model, as shown in Figure 2. The modeled concentrations agree well with measurements made at Pasadena but tend to be higher than observations made at downtown Los Angeles and West Los Angeles. The absence of any substantial deficit in the primary organic aerosol carbon concentrations computed for these three sites by the transport model when compared to actual ambient organic carbon concentrations suggests that most of the total aerosol organic carbon concentrations observed over long averaging times during 1982 were due to primary emissions of organic carbon particles rather than to secondary organic aerosol formation.
Environ. Sci. Technoi., Voi. 27, No. 10, I993
2049
m VI
00
'
45
2
.
RESULT - ( $ L UMODEL RED
'
PASADENA ACIDIC
'
. (h) M o k R;SULT
2
'
MEASURED VALUE e
d o w h r d w ~L A' ACIDIC
40-
.
,77--T
(1)
' MODEL RESULT -
,
T
_
_
WE,& L A
'
ACIDIC
MEASURED VALUE
35-
-8 E
2
302520-
. '
15-
10-
'
,
___I
I
,
I
,
.
e 8
e 0
05 L
00
'
,
I
.
I
r--y
:
'
,
,
I
1982
7 L___-/r
LI---,-_ I
1982
-1
I
,
I
I
,
,
,
-_r
,
-
1982
Flgure 3. Modeled vs measured monthly concentrationsof fine organics that elute between the CI2and Cae+alkanes at Los Angeles Basin ambient sites: (a) Pasadena, neutral fraction; (b) downtown Los Angeles, neutral fraction; (c) West Los Angeles, neutral fraction; (d) Pasadena, acidi-neutral fractlon; (e) downtown Los Angeles, acid+-neutral fraction; (f) West Los Angeles, acid+neutral fraction; (9) Pasadena, acidic fraction; (h) downtown Los Angeles, acldlc fraction; and (I)West Los Angeles, acidic fraction.
organic acids seen at downtown Los Angeles and West Los Angeles. Hence, it appears that secondary processes in the atmosphere and/or primary sources of organic acids not included in the current inventory (e.g., airborne vegetative detritus) are contributing substantially to the acidic fraction of the organic aerosol seen at the ambient sites. Primary acidic organicaerosol concentrations computed from the air-quality model can be subtracted from the measured ambient fine aerosol organic acids concentrations, as shown in Table 11. From the data obtained, it is possible to estimate the long-term average organic acids concentration that could be due to formation of fine secondary organic aerosol. Provided that the model does not greatly overestimatethe primary source contributions to ambient acidic organic levels, this concentration estimate will represent an upper bound for elutable secondary acidic organic aerosol, because it assumes that the untested sources do not emit any acidic organic aerosols. Given that most of the predicted ambient acidic organic aerosol originates from the San Nicolas Island background airquality measurements supplied to the model rather than from local sources, it is unlikely that the acidic organic concentrations have been greatly overestimated by the model. On an annual basis, the difference between modeled primary acids and total measured acids in the fine aerosol averages 0.6, 1.6, and 1.5 pg m-3 at Pasadena, downtown Los Angeles, and West Los Angeles, respectively. This can be compared to an average concentration of total fine 2050
Envlron. Scl. Technol., Vol. 27, No. 10, 1993
organics (1.2 X OC) of 8.0,8.6, and 6.9p.g m-3,respectively, showing that elutable secondary organic acids constitute no more than 9 % , 1 976, and 26 % of the total fine organic aerosol present on an annual basis at these three sites. It should be noted, however, that some of the noneluting organic material may also be secondary in origin. From Table 11, it can be determined that the fraction of total fine ambient organic aerosol which eluted from our GC column was smaller for Pasadena (62% , using the annual average quantities) than for downtown Los Angeles and West Los Angeles (83% and 77 7% ,respectively). If some of the nonelutable organics were also secondary, the total secondary organic acids present could have been higher than the percentages given above, especially at the Pasadena site. This upper range of 9-26% is consistent with the inference of Gray (15),based on total carbon to elemental carbon ratios, that secondary organic aerosol constituted on average no more than 27-38 7% of the total fine organics present at Los Angeles area monitoring sites during 1982. It is also similar to the estimate of Pandis et al. (12)) based on photochemical trajectory modeling, that secondary organics constituted 5-22 5% of the organic aerosol measured in the Los Angeles area during portions of the 1987 Southern California Air Quality Study. Alternatively, this deficit in modeled concentrations of acidic organics can be compared to HRGC elutable fine organics measured at these three sites, which averaged 4.9,7.2, and 5.4 pg m-3. On this basis, secondary organic acids constitute no more than 13% ,22 % ,and 29 % of the
Table 11. Extent of Ambient Elutable Acidic Organic Aerosol in Excess of That Due to Primary Emission Sources excess of elutable organic acids' (wg/m3)
total fine ambient organic aerosolb (wglm3)
Jan Feb Mar APT May Jun Jul Aug Sep Oct Nov Dec av
0.67 0.46 1.46 0.88 0.71 0.55 0.34 0.39 0.20 0.37 0.99 0.64
9.94 7.85 7.73 6.07 5.78 5.06 7.66 7.37 8.82 7.78 8.93 12.42 7.95
Jan Feb Mar APr May Jun Jul Aug SeP Oct Nov Dec av
2.37 0.96 0.66 1.16 0.97 1.32 2.06 1.05 1.07 1.98 2.42 3.53 1.63
11.88 8.28 8.26 6.72 5.69 5.16 7.03 6.46 7.15 10.39 10.70 15.86 8.63
Jan Feb Mar APr May Jun Jul Aug Sep Oct Nov Dec av
0.83 1.88 1.38 1.98 1.87 1.70 1.46 1.52 1.66 0.66 1.22 1.95 1.51
7.56 7.44 6.43 5.64 4.15 3.53 4.80 4.39 6.48 9.10 10.51 13.14 6.94
month
fine elutable (acid+neutral) ambient organic aerosol (wg/ms) Pasadena 7.07 5.76 4.24 4.98 4.74 4.88 4.95 3.97 3.51 3.97 4.53 6.21 4.90 Downtown Los Angeles 9.81 6.70 6.41 5.67 5.32 5.57 6.78 5.25 6.02 7.68 8.86 11.83 7.16 West Los Angeles 6.40 6.02 5.13 5.30 4.93 4.15 4.53 3.86 4.77 5.30 6.48 7.52 5.37
excess/total fine ambient organic aerosol ( % )
excess/fine elutable ambient organic aerosol ( 7% )
8.5 6.0 24.0 15.2 14.0 7.2 4.6 4.4 2.6 4.1 8.0 9.0
11.6 10.8 29.3 18.6 14.5 11.1 8.6 11.1 5.0 8.2 15.9 13.2
19.9 11.6 8.0 17.3 17.1 25.6 29.3 16.3 15.0 19.1 22.6 22.3 18.7
24.2 14.3 10.3 20.5 18.2 23.7 30.4 20.0 17.8 25.8 27.3 29.8 21.9
11.0 25.3 21.5 30.8 45.0 48.2 30.4 34.6 25.6 7.3 11.6 14.8 25.5
13.0 31.2 26.9 37.4 37.9 41.0 32.2 39.4 34.8 12.5 18.8 25.9 29.3
a Calculated as the difference between the measured ambient concentration and the model prediction. Calculated as 1.2 X organic carbon concentration as measured by thermal evolution/combustion analysis (22, 23).
elutable fine organic aerosol present on a yearly average at the three sites. The results also can be averaged on a seasonal basis by combining the three sites. The ratio of excess elutable acids to total elutable organics in the Los Angeles Basin averages 18%for the winter months, 27 % for the spring, 23% for the summer, and 19% for the fall. Hence, over the course of a year, this basin-wide average, representing an upper limit on the fraction of elutable organic aerosol present in Los Angeles air that may be secondary in origin, varies between 18and 27 5%. Whether this percentage range represents an upper or lower bound on total average secondary ambient organics depends on the nature of the nonelutable material (as well as the accuracy of the model predictions). If the nonelutable organics consist mainly of primary emissions, the percentages should represent an upper-bound estimate; if it includes substantial secondary material, they may be somewhat low. It should be noted that for short-term measurements during air pollution episodes, a much higher fraction of the total organic aerosol in the Los Angeles area has been
attributed to secondary formation (24). Nonetheless, the calculations in this paper suggest that the bulk of the fine organic aerosol in the Los Angeles Basin over long averaging times during 1982 was primary in origin. This explains why the monthly model predictions based solely on primary organic aerosol emissions are in reasonable agreement with ambient aerosol measurements, both for the case of total fine organic carbon (Figure 2) and fine elutable organics (Figure 3). (iii) Mass Distribution of Elutable Organics. Monthly composites of ambient organic aerosol samples from each of the three monitoring sites studied were analyzed by HRGC. The mass concentration of organic species that elute between each of the normal alkanes in the range C12-C36 was determined for both the neutral and the acid+neutral fractions as previously described for the source samples; the seasonal characteristicsof these ambient measurements are discussedelsewhere (20). Since most of the source samples were taken in or near Pasadena, the Pasadena air-monitoring site was selected to be examined in detail. Environ. Sci. Technol., Vol. 27, No. 10, 1993 2051
O"
-
0.5
::+--y
0.1
0.0
I
0.4 -- FEBRUARY
A C D I NEUTRAL FRACTION ~ y NEUTRALFRACTION
0.4
u
0.1
. . . . . . . .
Acm FRACTION
,.:-
..,--
..,
0.0
,
7
0.3
0.3 0.2
0.1
0.1
0.0
0.0
0.4
u'
0.0
5
0.4 -FEBRUARY
u
0.3
0.2
;',
, , , ,
,
-1 NOVEMBER
0 0.2
AUGUST
-:.
, ,-,-3 ..,............ -.' ..
1 ' 1
b .
I q
.
0.4
4JUNE
I
0.0
k] /q
! ::: 2 4 5 -
3 00 . 1
......... ____,
...,..
r
2
0.2
01
. . . . . . . . .
8 "E 2
..
I
I
0.2 0.1
0.0
00
12 1416182022242628 3 0 3 2 3 4 3 6 n-ALKANE CARBON NUMBER
0.4 0.3
Flgure 4. Measured monthly elutable fine organic mass distributions for Pasadena in 1982. 0.0
Figure 4 presents the mass distribution results for the Pasadena site for each month of the year 1982. It is observed that both the neutral and the acid+neutral fractions display a bimodal distribution, with the peak of the first mode occurring between the elution points of the c16 and c23 n-alkanes, and the second mode peaking between the and c 3 2 n-alkanes. The previously observed wintertime peak in elutable organic aerosol concentrations such as that seen in Figure 3f is found to be due mainly to a substantial wintertime increase in the intensity of the second mode. A substantial quantity of acidic organics is also observed, with most of the acidic organics mass concentration occurringin the range between the elution points of the c16 and C23 n-alkanes. (See ref 20 for a more detailed discussion of the ambient results.) The mass distribution of elutable organics due to primary aerosol sourceswas computed from the air-quality model for the Pasadena site for each month of the year 1982. Predictions for the neutral fraction for each month are compared with the ambient measurements in Figure 5. The degree of agreement between model results and ambient measurements of the neutral fraction mass distributions is encouraging, especially if one mentally adjusts the modeled distributions vertically by the amount necessaryto account for the differences between measured and modeled total elutable organics that are shown in Figure 3. (Recall that the modeled concentrations are basedon transport calculations for each hour of the month while the measured concentrations represent a composite of only six days of the month; hence perfect agreement between modeled and measured total mass concentrations is not expected.) Thus, the model predictions for September are higher than observed, while the May, June, and December model results are somewhat lower than the 2052
Environ. Sci. Technol., Vol. 27, No. 10, 1993
I
I
I
,
,
I
,
......
,
I
I
,____
I
I
I
I
I
I
I
I
I
I
I
I
I
!
. . . . . . . . . ..... .....:-..;.......5 ...... ..,... j
__,-.
. _ . ,.. ___#..
I
I
,
'
I
I
,
I
,
I
I
I
I
I
I
"E
I
I
\-y
1
I.--;-
I
I
I" ," ' ....................... ,____ ....................... ..............
' - - - I
I
NOVEMBER
I
VL'y 0.3
.....
_,_._.
'I
.......
MAY
0.4
W
0.3
12 14 16 18 20 22 24 26 28 30 32 34 36 n-ALKANECARBON NUMBER
I
I OCTOBER
iAPRIL
:iu 0.4
___,
, , , , , ,__
I
-1 OCTOBER
0.0
0.1
~
SEPTEMBER
-
0.4
.-;.'
0.3
,A
00
m
.......
I
I
I
I
1
___,-
I
I
I
I
1
, , I
I
I
I
I
, 1 I
12 14 16 18 20 22 24 26 28 30 32 34 12 14 16 18 20 22 24 26 28 30 32 34 36
n-ALKANECARBON NUMBER n-ALKANECARBON NUMBER Flgure 5. Comparison between modeled primary and measured neutral organic mass distributions for Pasadena in 1982.
observations, both as expected from Figure 3. The general shape of the modeled neutral fraction mass distributions shown in Figure 5 is quite close to observations during March, July, August, and October when both the measured and modeled total elutable organics concentrations (shown in Figure 3) are in close agreement. Some characteristic differences between the model predictions and the ambient measurements for the Pasadena site can be discerned. The modeled mass distributions for the neutral organic compounds tend to be lower than observations in the C12-Cls n-alkane range. As a result, the model does not show the bimodal distribution apparent in the ambient samples. Also for the neutral fraction, the model tends to overpredict the amount of organics mass between the CZOand C25 n-alkanes. Because of the generally good agreement between measured and modeled results for the neutral organic mass distributions, the influence of the different emission sources on the shape of the distribution curve and its seasonal variation can be examined. In Figure 6, the sources contributing to the organic mass distribution predicted for Pasadena are shown for 4 months spanning the year 1982. Surprisingly, the neutral organics found n-alkanes are mainly the result between the C15 and of background organics already present in the air mass as it passes over San Nicolas Island. As the model underpredicts aerosol concentrations in the elution range of the C15-Cl7 n-alkanes, it is possible that additional organics like those found in San Nicolas Island background air
2
0.18
2
0.15
+
2 &
" 0.15
0.12
&
0.09
Ew
0.06
,*
0.03
2-
0.00
E
I
12
, 14
, 16
, 18
, 20
,
22
, , . , , . 24 26 28 30 32 34 36 n-ALKANE CARBON NUMBER
uc
6
3
0.24
5
0.21
z_
0 U
(4
-
2
2
0
-
-
0.18 0.15
2
0.09
0.03 0.00
0.30
14
16
18
20
22
(4
0.27
24 26 28 30 32 34 36 n-ALKANE CARBON NUMBER
1982
0.24
0.21 0.18 0.15 0.12 0.09
Ew
0.06
PE-
0.06
U
u= 0.12
,*
0.09
BOILER - DIST OIL H.D.DIESELTRUCKS NONCATALYST CARS CATALYST CARS BACKGROUND (SNI)
0
U
z_
0.12
12
0.30 0.27
uc
SEITEMBER 1982
a4-
4
0.06
0.03
,*
0.03
0.00
2
0.00
E
12
14
16
18
20
22
24 26 28 30 32 34 36 n-ALKANE CARBON NUMBER
12
14
16
18
20
22
24 26 28 30 32 34 36 n-ALKANE CARBON NUMBER
Figure 6. Contributions of elutable acid-kneutrai organics from sources to modeled organic mass distributlons for Pasadena in 1982: (a) March, (b) June, (c) September, and (d) December.
(e.g., from grasses, which cover the island) may be added to air parcels as they move inland over the Los Angeles area. The appearance of a distinct mode in the modeled mass distribution with a peak around the elution point of the C27 n-alkane is seen to result mainly from automotive emissions, This mode becomes even more pronounced in the December plot: concentrations from automotive emissions are elevated due to wintertime meteorological conditions, and fireplace emissions become a significant additional part of the total elutable organics mass concentration. This evaluation of the major primary sources contributing to ambient organic aerosol concentrations could assist the design of pollution-control strategies for primary organic aerosol. The model results for the acidic fraction of the primary organic aerosol, shown in Figure 7, fall significantly below the amount of acidic organics present in the Pasadena atmosphere. In particular, the substantial concentrations of acidic organics that elute between the C16 and C23 n-alkanes in the ambient samples are not adequately represented by primary emission sources that are included in the model. Primary emissions from most of the sources tested contained little or no acidic organics; hence, most of the acidic organic material appearing in the modeled result is attributable to background aerosol as measured at San Nicolas Island (Figure 7). Three hypotheses can be offered to account for the difference between measured acidic organics and the modeled contributions due to anthropogenic primary emissionsplus background. First, the organic acids needed to complete a material balance on the acidic aerosol could result from reactions of gas-phase molecules to form
aerosol-phase products that elute in the acidic fractions. This would support the traditional view that secondary aerosol from gas-phase precursors is responsible for most of the organic acids (9-11). However, two other possibilities also should be mentioned. Since the tendancy of the model is to overpredict the neutral mass between C20 and C25, it is possible that the primary particulate emissions of neutral organics in this region are especially reactive and react in the atmosphere to form acidic aerosol products. Alternatively, acids contributed by the primary emissions of organic vegetative detritus due to leaf abrasion have been noted as a significant contributor to the primary aerosol (4,19),but those emissions are not included in the transport model because their emission rate is not known. Conclusions An atmospheric transport model was used to predict a set of characteristics for the fine primary organic particulate matter at ambient monitoring sites in the Los Angeles area, as would be expected if the primary organic aerosol emissions were transported without further chemical reaction. The source samples and ambient aerosol samples were collected and analyzed by comparable methods so that accurate comparisons could be drawn between predicted and observed organic aerosol properties. The neutral fraction of the organic aerosol can be attributed with reasonable accuracy to known primary emissionsources plus background air quality. The monthby-month model predictions for the year 1982 reflect the seasonal variations observed in the neutral organic fraction of the ambient aerosol, and the predicted concentrations Environ. Sci. Technol., Vol. 27, No. 10, 1993 2053
BACKGROUND FROM SAN NICOLAS ISLAND
z.
0.14
f
0.12
BACKGROUND FROM SAN NICOLAS ISLAND
i:
6
0.14
3 0.12 I
nn
us 0.08 z
z" 0.10
:
0.08
0.06
0.06
5
0.04
0.04
,"
0.02
Y
E
m
m
E
2 0.00
2 0.00 12
14
16
18
20
22
24
26
28
30
32
34
36
n
2
"' 0.20
0.18
-
0.16
-
U
- f 0.12 0.10 ue 0.08 &
---
14
16
18
20
22
24
26
28
30
32
34
36
n-ALKANE CARBON NUMBER
0.20 MEASUREDVALUE MODEL RESULT BACKGROUND FROM SAN NICOLAS ISLAND
$ 12
n-ALKANE CARBON NUMBER
(c)
n
0.02
SEPTEMBER
0.18
3
u
0.14
BACKGROUND FROM
0.16
DECEMBER
SAN NICOLAS ISLAND
0.14
24 0.10 -v2 0.12
4
I
r-
0.08
3
0.06 0.04
m
m
E
,mE 0.02 p 0.00
0.02
p 0.00 12
14
16
18
20
22
24
26
28
30
32
34
36
12
14
16
18
n-ALKANE CARBON NUMBER
20
22
24
26
28
30
32
34
36
n-ALKANE CARBON NUMBER
Flgure 7. Comparison between modeled primary and measured acldic organic mass distributions for Pasadena in 1982: (a) March, (b) June, (c) September, and (d) December.
are close to the observed values. However, the amount of neutral organics that elute in the range between the CZO and n-alkanes is overpredicted by the model. A distinct peak in the mass of organics that elute near the C Zn-alkane ~ is shown by the model to result mainly from motor vehicle exhaust emissions. This peak becomes more pronounced in the winter months, when emissions from fireplaces also contribute significantly. This indication that, on average, much of the ambient organic aerosol may originate directly from primary sources could simplify the implementation of effective control strategies. Although some organic acids are emitted from primary aerosol sources (4),these primary emissionsare insufficient to account for the ambient organic acid concentrations measured during this study. Primary acidic organicaerosol from the anthropogenic aerosol sources studied here is calculated by the model to contribute only 0.35,0.42, and 0.41 pg/m3 to the 0.97, 2.06, and 1.92 pg/m3 of acidic organics measured at Pasadena, downtown Los Angeles, and West Los Angeles, respectively, during the year 1982. Organic acids present in background air as measured at San Nicolas Island contributed 0.19 kg/m3 to the annual average elutable fine organic acids in the aerosol phase. Some of the apparent excess of ambient organic acid aerosols will be explained once the contribution of plant fragments to the aerosol has been determined based on matching the biomarkers present in vegetative source samples to their abundance in the ambient aerosol (4,19, 25). The remaining acidic organics needed to explain the ambient organic acid concentrations result from reactions 2054
Envlron. Scl. Technol., Vol. 27, No. 10, 1993
in the atmosphere, involvingeither gas-phase reactants or perhaps attack on primary neutral organic aerosol. Acknowledgments Portions of this paper were presented at the 1989AAAR Conference in Reno, NV, and a t the 1990 ASCE Specialty Conferenceon EnvironmentalEngineering in Washington, DC. This work was supported by EPA Grant R-81327701-0and by gifts to the Environmental Quality Laboratory. This paper has not been subject to the EPA's peer and policy review and, hence, does not necessarily reflect the views of the EPA. Mention of trade names or commercial products does not constitute EPA endorsement or recommendation for use. Literature Cited (1) Cass, G. R.; Boone, I?. M.; Macias, E. S. In Particulate Carbon: Atmospheric Life Cycle; Wolff, G. T., Klimisch, R. L., Eds.; Plenum Press: New York, 1982; pp 207-240.
(2) Daisey,J. M.;Cheney, J. L.;Lioy,P. J. J.AirPollut. Control ASSOC.1986, 36, 17-33.
(3) Hildemann, L. M.; Cass, G . R.; Markowski, G. R. Aerosol Sei. Technol. 1989, 10, 193-204. (4) Hildemann, L. M.; Mazurek, M. A.; Cass, G . R.; Simoneit, B. R. T. Environ. Sei. Technol. 1991,25, 1311-1325. (5) Zelenski, S. G.; Hunt, G. T.; Pangaro, N. In Polynuclear Aromatic Hydrocarbons: Chemistryand Biological Effects; Bjorseth, A., Dennis, A. J., Eds.; Battelle Press: Columbus, OH, 1980; pp 589-597. (6) Friedlander, S. K. In Atmospheric Aerosol: SourcelAir Quality Relationships; Macias, E. S., Hopke, P. K., Eds.; American Chemical Society: Washington, DC, 1981; pp 1-19.
Laboratory, California Institute of Technology: Pasadena, CA, 1986. Markowski,G. R.; Cass, G. R. Enuiron. (16) Hildemann, L. M.;
(18)Mazurek, M. A,; Simoneit, B. R. T.; Cam, G. R.; Gray, H. A. Int. J . Anal. Chem. 1987,29,119-139. (19)Mazurek, M. A.; Caw, G. R.; Simoneit, B. R. T. Aerosol Sci. Technol. 1989,10,408-420. (20)Hildemann, L. M.; Mazurek, M. A,; Cass, G. R.; Simoneit, B. R. T. Aerosol Sci. Technol., submitted. (21)Larson, S. M.; Cass, G. R.; Gray, H. A. Aerosol Sci. Technol. 1989,10,118-130. (22) Huntzicker, J. J.; Johnson, R. L.; Shah, J. J.; Cary, R. A. In Particulate Carbon: Atmospheric Life Cycle;Wolff, G. T., Klimisch, R. L.; Eds.; Plenum Press: New York, 1982;pp 79-85. (23) Johnson, R. L.; Shah, J. J.; Cary, R. A.; Huntzicker, J. J. In Atmospheric Aerosols; SourcelAir Quality Relationships; Macias, E. S., Hopke, P. K., Eds.; American Chemical Society: Washington, DC, 1981;pp 223-233. (24) Turpin, B. J.;Huntzicker, J. J. Atmos. Enuiron. 1991,25A, 207-215. (25) Mazurek, M. A.; Cass, G. R.; Simoneit, B. R. T. Environ. Sci. Technol. 1991,25,684-694.
Sci. Technol. 1991,25,744-759. (17) Gray, H. A.; Cass, G. R.; Huntzicker, J. J.; Heyerdahl, E. K.; Rau, J. A. Environ. Sci. Technol. 1986,20,580-589.
Received for review October 23, 1992.Reuised manuscript received May 21, 1993.Accepted May 25, 1993.
(7) Behymer, T. A.; Hites, R. A. Environ. Sci. Technol. 1984, 18,203-206. (8)Rogge, W. F.; Hildemann, L. M.; Mazurek, M. A,; Cam, G. R.; Simoneit, B. R. T. Environ. Sci. Technol. 1993,27,18921904. (9)Grosjean,D.;Friedlander, S. K. J . AirPollut. Control Assoc. 1976,25,1038-1044. (10)Schuetzle, D.;Cronn, D.; Crittenden, A. L.; Charlson, R. J. Environ. Sci. Technol. 1975,9,838-845. (11) Cronn, D.R.; Charlson, R. J.; Knights, R. L.; Crittenden, A. L.; Appel, B. R. Atmos. Environ. 1977,11,929-937. (12) Pandis, S. N.;Harley, R. A.; Cass, G. R.; Seinfeld, J. H. Atmos. Environ. 1992,13,2269-2282. (13)Cass,G. R. Ph.D. Thesis, CaliforniaInstitute of Technology, Pasadena, CA, 1977. (14) Cass, G. R. Atmos. Enuiron. 1981,15,1227-1249. (15) Gray, H. A. EQL Report 23; Environmental Quality
Envlron. Scl. Technol., Vol. 27, No. 10, 1993 2056
