Envlron. Scl. Technol. 1991, 25, 1311-1325
article has been funded in part by the U.S. Environmental Protection Agency under Assistance Agreement R-815709 to Montana State University through the Hazardous Substance Research Center for US.E P A regions 7 and 8 headquartered at Kansas State University, it has not been subjected to the
Agency’s peer and administrative review and, therefore, may not necessarily reflect the views of the Agency, and no official endorsement should be inferred. This research was also supported by the US.Geological Survey (Project 14-08-0001G1284).
Quantitative Characterization of Urban Sources of Organic Aerosol by High-Resolution Gas Chromatography Lynn M. Hlldemann,?Monica A. Mazurek,s and Glen R. Caw”
Environmental Engineerlng Science Department and Environmental Quality Laboratory, California Institute of Technology, Pasadena, California 91 125 Bernd R. T. Slmonelt
College of Oceanography, Oregon State University, Corvallis, Oregon 97331 ~
~~~~
~~
~
~
~
Fine aerosol emissions have been collected from a variety of urban combustion sources, including an industrial boiler, a fireplace, automobiles, diesel trucks, gas-fired home appliances, and meat cooking operations, by use of a dilution sampling system. Other sampling techniques have been utilized to collect fine aerosol samples of paved road dust, brake wear, tire wear, cigarette smoke, tar pot emissions, and vegetative detritus. The organic matter contained in each of these samples has been analyzed via high-resolution gas chromatography. By use of a simple computational approach, a quantitative, 50-parameter characterization of the elutable fine organic aerosol emitted from each source type has been determined. The organic mass distribution fingerprints obtained by this approach are shown to differ significantly from each other for most of the source types tested, using hierarchical cluster analysis.
Introduction Most of the existing data on organic aerosol emissions from sources are derived from bulk chemical analyses that quantify the total amount of organic carbon present. Total aerosol emissions of organic carbon have been measured by combustion techniques for a large variety of sources (e.g., refs 1and 2). Gravimetric procedures have been used to quantify the amount of extractable organic material emitted from sources such as boilers (e.g., refs 3-5), motor vehicles (e.g., refs 6-9), and fireplaces (e.g., refs 10 and 11). Some separation of the organic species based on vaporization temperature also has been achieved by using thermal evolution analysis to produce “thermograms” (12-17). These approaches provide little or no information about the underlying molecular structure of the organic aerosol compounds emitted. Other researchers have used costly, time-consuming procedures to identify specific organic compounds (e.g., polycyclic aromatic hydrocarbons) through approaches such as gas chromatography coupled with mass spectrometry (GC/MS) and high-pressure liquid chromatography with fluorescence detection. Daisey and co-workers (18) have published a thorough review of the literature between 1972 and 1986 that identifies specific organic compounds in various source emissions. The individual compounds identified typically account for only a few percent of the ‘Present address: Department of Civil Engineering, Stanford University, Stanford, CA 94305-4020. t Present address: Environmental Chemistry Division, Bldg. 426, Brookhaven National Laboratory, Upton, NY 11973. 0013-936X/91/0925-1311$02.50/0
total organic mass emitted. Hence these data are of limited use in designing pollution abatement programs that must achieve changes in the total amount of aerosol present. Aerosol carbon emissions arise from a large number of different source types. Recent studies of Los Angeles (19, 20) have shown that at least a dozen different source types must be considered in order to account for close to 80% of the primary organic aerosol emissions in that airshed. To determine the effect of these emission sources on ambient air quality via receptor modeling techniques, at the very least there must be more distinctive attributes of the organic aerosol than there are different source types. The predictions of transport-oriented models in relating source contributions to the ambient organic aerosol likewise require detailed characterization of the emission sources for model verification. In both air quality modeling approaches, a procedure is needed for characterizing the organic aerosol emission sources that is more economical than GC/MS analysis, but provides a fuller description of the differences between the source types than is possible from bulk carbon analysis. In the present study, gas chromatography (GC) analyses alone are used to characterize the organic aerosol emissions from a variety of sources in a way that reveals the distinctive features of the different source types using a quantitative, multiparameter description. This approach shows promise as a tool for characterizing organic emissions in a way that can be used to better support air quality modeling studies. Experimental Section Sample Collection. Fine organic aerosol (particle diameter d, C 2 ,urn) was collected from combustion sources by a dilution stack sampling system, and other sources were sampled by grab sampling techniques. The source types tested and the number of samples analyzed are itemized in Table I. Details of the sampling procedures used for each of the sources have been described elsewhere (191, and the design of the stack sampler has been documented previously (21). Briefly, hot stack gases were diluted many-fold with purified dilution air that had passed through an activated carbon bed and an absolute filter. During the dilution process, the source effluent was cooled to ambient temperature and kept at near-ambient pressure, causing the aerosol-forming organic vapors present to condense onto preexisting aerosol as would normally occur in the plume downwind of the stack. The diluted emissions passed through a 2-pm size-cut cyclone, and the fine particulate
@ 1991 American Chemical Soclety
Environ. Sci. Technol., Vol. 25, No. 7, 1991
1311
Table I. Characteristics of Elutable Organics from Urban Sources of Organic Aerosol source (no. of samples analyzed)
eluted on GC," %
boiler, no. 2 fuel oil (3) 57-67 motor vehicles catalyst-equipped 133 f 17 automobiles (1) noncatalyst automobiles (1) 142 f 8 92 f 3 diesel trucks (1) roofing tar pot (2) 154-171 tire dust (2) 44-49' fireplace natural wood (2) 45-51 synthetic log (1) 77 f 4 c cigarettes (2) 81-87 brake dust (1) >2.7d hamburger cooking charbroiling (extraiean) (1) 43 f 7 charboiling (regular) (1) 27 k 2 60 f 19 frying (1) home appliances, natural gas (1) 94 f 26 vegetative detritus dead leaves (2) 19-25 green leaves (2) 27-33 paved road dust (2) 21-30
U:Rb ratio acid + neutral neutral 3.8-4.2
3.1-3.8
16.5
14.5
9.0 8.3 7.7 6.8
7.9 8.2 7.6 10.1
3.0-3.4 3.7 2.3 4.6
3.2-3.9 3.4 2.9 7.0
4.1 3.9 5.5
5.2 7.4 4.7 1.0
1.3
0.8 1.1
5.3
0.9 1.5 12.7
Error bounds given for sources where only one sample was analyzed solely represent the uncertainty in the measurement of the mass of organic carbon present. *U:R, unresolved to resolved ratio. 'GC programming extended t o a run time of 120 min for this sample. Delineation between elemental and organic carbon uncertain due to interference with laser transmittance. Lower bound value shown assumes that all the carbon was present as organic matter. a
matter then was collected on prebaked quartz fiber filters (Pallflex 47 mm 2500 QAO). An additional quartz fiber filter was located downstream of one Teflon filter used for aerosol collection in order to monitor the potential for organic aerosol sampling artifacts (19). The sampled fdters were stored in annealed glass jars with Teflon-lined lids at -25 "C until extraction. Extraction. The isolation and quantification procedures used in the extraction of the organic source samples have been described elsewhere in detail (22,23). Each set of filters was spiked with a known aliquot of perdeuterated standard(s) (n-C24D50for source samples, a suite of six compounds for the blanks) in order to quantify the extraction efficiency and recovery of organic compounds from the filters. The source samples varied greatly in composition, consisting of from 1 to 45 filters, and containing from 300 pg to 20 mg of organic carbon. To ensure that an appropriate amount of standard(s) was added to each sample, a trial organic filter sample was collected concurrently with the actual sample to be analyzed and was spiked, extracted, and analyzed before beginning extraction of the main source sample. In this manner, blind analysis of the critical source samples was eliminated. The preliminary analyses allowed determination of the appropriate mass of standard to be added to each source sample, permitting accurate quantitation of the elutable organic material for each source type. The typical sample, which consisted of 15 filters and contained -1 mg of organic carbon, was spiked with 50 pL of n-C,,D50 at a concentration of 154 ng/pL. Sequential additions of hexane (2 X 60 mL additions) and benzene/2-propanol(2:1) (3 X 60 mL additions) were used to extract the organics from the filters under ultrasonic agitation. Following filtration and combination of the extracts, each sample was reduced to a volume of 1312
Environ. Scl. Technoi., Vol. 25, No. 7, 1991
200-500 pL for gas chromatographic analysis. Gas Chromatographic Analysis. The concentrated organic extracts were analyzed with a Varian 4600 highresolution gas chromatograph equipped with a conventional Grob splitless injector, a 30-m fused-silica OV-1701 column (bonded 86% dimethyl-14% (cyanopropyl) phenylpolysiloxane, 25-pm film thickness, 0.32 mm i.d., J&W Scientific, Rancho Cordova, CA), and a flame ionization detector (FID). Peak integration was performed electronically by using a Varian Vista 402 computerized data system. The extracts were introduced onto the capillary column by a flash vaporization technique and with an injector temperature of 300 "C (22, 23). Temperature programming consisted of the following steps: (1) isothermal hold a t 65 OC for 10 min; (2) temperature ramp of 10 "C/min to 275 "C; and (3) isothermal hold at 275 "C for 49 min. The GC trace obtained when the organic extract was injected without derivatization was considered to represent the nonpolar or neutral fraction of the organic compounds present in the extract. A separate aliquot of the extract was derivatized through the addition of diazomethane to convert labile organic acids to their methyl ester analogues and acidic hydroxy compounds to their methyl ester analogues. This derivatized aliquot also was injected, giving a GC trace representing the acid plus neutral ("acid + neutral") fraction of the organic extract. The acidic fraction, which was determined as the mass difference between the results of the neutral and the acid + neutral fractions, contained those oxidized polar compounds that were soluble in the extraction solvents used. Resolved component mass contributions were determined by recalculation of the raw data at elevated signal-to-noise (SN) ratios to exclude the portion of the GC trace that contained no resolved peaks. The appropriate elevated SN ratio was chosen as the lowest signal-to-noise ratio that produced a base line that excluded all of the unresolved area. The unresolved mass component was calculated by the difference in area counts between the results a t SN = 1 and a t the elevated SN ratio. Regular injections of a 17-component n-alkane standard suite (n-Clo-n-C36homologues) throughout the course of the source analysis program were used to assist in subdividing the GC traces into elution zones based on the retention times of the n-alkane homologues. For some source extracts, homologous series with retention times corresponding to the n-alkanes were clearly visible in the GC trace of the source sample and could be used directly. For other source extracts, the retention times obtained from an n-alkane standard injection were used to determine the elution zones. Area counts for compounds eluting between each pair of adjacent n-alkanes were summed in order to define the relative mass distribution of compounds in each sample. A similar approach has been used previously (7,24)to characterize the organics in diesel exhaust. Mass Quantification. Conversion of area counts to organic mass must account for the FID response to the particular injection, as well as any losses that might have occurred during sample extraction and concentration. It has been observed from standard injections of homologous series in both the present and previous studies (e.g., ref 25) that instrument response for a given mass of a component in a series varies with retention time, becoming less sensitive for the higher molecular weight components in the series. Hence, it was desired to account for this variable sensitivity. Two standard suites of homologous compounds, the n-alkanes and the normal fatty acid methyl esters (n-
Ce-n-Cm alkanoic acid homologues) were injected regularly during the source analysis program. As will be shown later in this paper, the majority of area counts were contributed by neutral compounds for most of the source extracts. Accordingly, it was decided that the instrument response to the n-alkane series (which consists of neutral compounds) would be used to quantify the aerosol organic mass in each chromatographic subregion eluting between two consecutive members of the n-alkane series. The quantification approach utilized the relationship between n-alkane area counts and n-alkane mass to calculate the mass of organics eluting between adjacent alkanes, as follows: mass between C, and Cn+l = mass n-C24D, spiked onto sample X counts n-Cz4D, counts between C, and Cn+l X RRFc, (1) RRF,.,, is the relative response factor for nwhere RRF,.c CuDWand R R e i s the relative response factor for alkane
c,.
Multilevel calibration procedures were used to determine the RRF values. The RRF for n-Cz4D, was calculated from regular injections of the n-CZ4DW standard with 1phenyldodecane at two standard concentrations, while the RRFs for the n-alkanes were determined from repeated injections of the standard n-alkane suite with l-phenyldodecane, also at two standard concentrations. The RRF values for a compound A were calculated as counts l-phenyldodecane mass A RRFA X mass l-phenyldodecane counts A (2)
Analytical Blanks. Rigorous quality control is a vital part of trace organic analytical procedures. Two types of blanks were analyzed in conjunction with this project. With each group of 10-12 source samples, a sample containing blank, baked quartz fiber filters was also spiked, extracted, and analyzed to monitor the extraction procedures and detect any contaminants present. The few significant contaminants seen were identified via GC/MS as low molecular weight solvent impurities and degradation products. These solvent byproducts also appeared in the source sample GC traces, and their resolved area counts were excluded from the data set, along with the resolved area counts of the perdeuterated standards. Samples obtained by filtration of the precleaned dilution air used in the dilution stack sampler were collected before each source type was sampled. These dynamic system blanks, which were intended to identify any contaminants in the dilution air and any residual contamination remaining after the sampler's components were cleaned, were extracted and analyzed in the same way as the source samples, and contaminant levels were seen to be very low.
Results Total vs Elutable Organics. Gas chromatography alone cannot measure the full spectrum of carbonaceous material present as collected aerosol; however, the measurable fraction of "elutable organics" that is detected by GC can be quantitatively linked to the total aerosol organics. For air quality modeling and emission control purposes, it is of interest to determine what fraction of the total organics in each source sample is elutable. The total amount of fine organic carbon aerosol collected from each source type was analyzed separately via a thermal evolution and combustion technique (26,27). The
emission rate of fine organic compounds measured from each source type by this technique is reported elsewhere (19). To determine the fraction of the total organic carbon present that is observed via GC analysis, the total concentration of organics obtained by integration under each GC trace was divided by the total concentration of organics as measured by the thermal evolution and combustion technique, using a factor of 1.2 to estimate total organics from total organic carbon (28). As seen in Table I, the fraction of the total organics emitted from each source (as measured by combustion analysis) that eluted on the GC during the 80-min elution cycle varied substantially between sources. For the tar pot, natural gas home appliances, and motor vehicles, a high fraction of the total fine organics collected were elutable. (Recoveries of greater than 100% most likely are due to an underestimate of the total organics based on the organic carbon measurement; a factor of 1.4 also has been used by some investigators to convert from organic carbon to total organics.) In contrast, less than one-third of the organics in paved road dust and vegetative detritus eluted on the GC. Organics that did not elute may have been insoluble in the solvents used, too acidic or basic in composition to migrate through the capillary column, thermally unstable, or of too high a molecular weight to elute during the 80-min GC cycle. Unreso1ved:ResolvedRatios. The ratio of unresolved components to resolved components (U:R) observed during GC analysis of atmospheric organic aerosol samples has been utilized in past studies as a qualitative indicator of the degree of anthropogenic contribution to the organic portion of airborne particulate matter (24-32). GC traces of petroleum and combustion products exhibit a large hump of unresolved compounds and hence have high U:R ratios. In previous analyses of urban aerosol samples (using a 30-m, 0.32 mm i.d. OV-1701 column) the neutral fraction showed U:R ratios of 2.6-25, while rural aerosol samples dominated by vegetation-derived organic aerosols characteristically showed ratios of less than 2.0 (33). From the results of the present study (using the same type of column), the U:R ratio characteristic of each major source type can be determined. While the precise U:R ratio for a GC trace will be influenced by the length and bore of the column, the similarity of the conditions of the present experiments to those employed in many of the previous studies permits the present data to be compared to the prior studies. Hence, the data set developed here could be used in later studies to explain qualitatively the differences in U:R ratios that have already been observed between air basins based on the types of emissions sources present. GC trace characteristics of the emissions from source types with both elevated and low U:R ratios are shown in Figure 1. Results for the catalyst-equipped automobiles (Figure l a ) show a large hump of unresolved compounds between CZ4and C,, with no predominant single peaks, yielding a high U:R ratio of 14.5. Fireplace combustion of pine wood provides a less extreme case (Figure lb), with a few important resolved peaks between C19and Czl and between CZ5and CZ6,but still exhibiting an intermediate U:R of 3.9. In contrast, both vegetative detritus (U:R = 1.4) and natural gas combustion emissions from home appliances (U:R = 1.0) (Figure lc,d) contain most of their organic mass in a few predominant resolved peaks, resulting in low U:R ratios. Table I lists the range of U:R ratios observed for each of the sources measured in this study. Petroleum products (tar pot, tire dust) had ratios of 7-10, while vehicular Environ. Scl. Technol., Vol. 25, No. 7. 1991
1313
CATALYST AUTOMOBILES
FIREPLACE, PINE WOOD
FINE AEROSOL EMISSIONS NEUTRAL FRACTION
FINE AEROSOL EMISSIONS NEUTRAL FRACTION
1
0
10
30
20
I2
14
16
50
40
30
2 0 2 2 2 4 2 6 20
I0
32
70
60
10
0
30
20 12
14
16
18
50
40
20 2 2 2 4 2 6 2 0
30
32
60
34
13
,
70~1ME
36
n ALKANE CARBON NUMBER
z 5
---
TOTAL NEUTRAL FRACTION RESOLVED FRACTION
1.6
00
TIME lm8ni
34 36 IALKANE CARBON NUMBER
360- U : R = 3 . 9
?
6
TOTAL NEUTRAL FRACTION RESOLVED FRACTION
320-
---
7
2
:
6 m 0 C
15j g
2ws .-
2001
0.2
i-,--~
F.,--_.--..
.; 0 0 - . - E 12 14
%
I . .
IS
16
..
20
--___.
22
,----
120
..._ _._-
24 26 28 30 32 34 n-ALKANE CARBON NUMBER
36
12
E
14
16
18
22
20
24 26 28 30 32 34 n-ALKANE CARBON NUMBER
VEGETATIVE DETRITUS, GREEN LEAVES
NATURAL GAS H O M E APPLIANCES
FINE AEROSOL EMISSIONS NEUTRAL FRACTION
FINE AEROSOL EMISSIONS NEUTRAL FRACTION
T
i 0
IO
u" 2
2o
1
30
20 12
I4
1'6
18
50
40
2 0 2 2 2 4 2 6 26
30
32
60
70
00
10
0
TIME l m i n l
20 I2
34 36 IALKANE CARBON NUMBER
14
10
20 2 2 2 4 2 6 28 30
2
-
135
32
60
70
BO TIME l m m l
34 36 n ALKANE CAR8ON NUMBER
-
3
---
1-1 I
u
50
40
30 16
NATURAL GAS HOME APPLIANCES L.R=IO TOTAL NELTRAL FRACTION 0 RESOLVED FRACTION - - 120-
GREEN VEGETATIVE DETRITUS TOTAL NEUTRAL FRACTION RESOLVED FRACTION
36
5c:
r-
1 - -1 5rrl
E
E
Y
m
ti
e
2
1
I
4.5
3.0 15
F 0.0
12
14
16
18
20
22
24 26 28 30 32 34 n-ALKANE CARBON VUMBER
36
12
14
16
18
20
22
24 26 28 30 32 34 n-ALKANE CARBON NUMBER
36
Figure 1. Unresolved vs resolved organic mass distributions. X denotes a known solvent artifact; S, is the coinjection standard, l-phenyldodecane; and SR Is the recovery standard n-C2,D5,,. These artifacts and standards are not included in area count summations. (a) Catalyst automobiles; (b) fireplace burning pine wood; (c) vegetative detritus, green leaves; (d) natural gas home appliances. 1314
Environ. Sci. Technol., Vol. 25, No. 7, 1991
CIGARETTE SMOKE (SAMPLE I ) C I G A R E l T E SMOKE (SAMPLE 2)
---
TIRE DUST (SAMPLE l j TIRE DUST (SAMPLE 2)
---
16-
1614-
14-
12-
12-
10-
10-
8-
b 4-
2-
PAVED ROAD DUST (SAMPLE I ) PAVED ROAD DUST (SAMPLE 2)
,__---
E
---
DEAD VEGETATIVE DETRITUS (SAMPLE I ) DEAD VEGETATIVE DETRITUS (SAMPLE 2)
?
GREEN VEGETATIVE DETRITUS (SAMPLE I ) GREEN VEGETATIVE DETRITUS (SAMPLE 2)
---
._
1.-
Y
---
ROOFING TAR POT (SAMPLE I j ROOFlNG TAR POT (SAMPLE 2)
I - -
---
I2
12
14
16
18
20
22
24
26
28 30 32 34 n-ALKANE CARBON NUMBER
12
14
16
18
20
22
24
26
28 30 32 34 36 "-ALKANE CARBON NUMBER
Flgure 2. Organic mass distributions obtained from replicate sample analyses. (a) Cigarette smoke; (b) tire dust; (c) paved road dust; (d) green vegetative detritus; (e) dead vegetative detritus; (f) roofing tar pot.
emissions showed ratios of 8-17. Somewhat lower values were observed for other types of petroleum combustion, with distillate oil-fired boiler emissions and synthetic log combustion (containing petroleum waxes) having ratios between 3.1 and 4.2. Combustion of plant material (fireplace combustion of natural wood, cigarettes) also resulted in emissions with elevated U:R ratios, between 2.3 and 3.9. The other anthropogenic sources tested likewise showed elevated U:R ratios, with the exception of natural gas combustion, which had U:R ratios in the range 1.0-1.3. The natural source tested, vegetative detritus, showed low U:R ratios as expected, ranging between 0.8 and 1.5. These measurements of the U:R ratio confirm that emissions from liquid petroleum products and their combustion, as well as combustion of some other types of carbonaceous materials, give U:R ratios greater than 2.0, while vegetative detritus has a U:R ratio near unity. However, the low ratio obtained for natural gas combustion emphasizes that the U:R ratio should be used only as a qualitative indicator of the anthropogenic content of an ambient aerosol sample, unless the contributions of those anthropogenic sources with a low U:R ratio can be determined independently. Organic Mass Distributions. GC analyses of source samples often yield traces that each contain hundreds of peaks corresponding to the different compounds that are present. While the identity of most of these compounds
cannot be determined by gas chromatography alone, it is still possible to more completely characterize the organics emitted from each source type by calculating the mass of organic compounds that elutes between each of a series of well-defined points in each GC trace. In the present work, source fingerprints characteristic of the organic compound distribution emitted by each source are assembled by summing the quantity of organic material that elutes between each of the normal alkanes in the range C12-C36under the experimentalconditions cited previously. (a) Replicate Analyses. The reproducibility with which the organic compound distribution from a single source can be characterized by the methods used in this study is demonstrated in Figure 2. Here, the derivatized mass distributions obtained from pairs of separately extracted and analyzed samples for six different source types are presented, using the percent of the total elutable organic mass that elutes between adjacent n-alkanes as the unit of measurement. The cigarette smoke data (Figure 2a) represent an exact analytical replicate, with each of the two samples containing the same number of simultaneously collected filters. The tire dust and paved road dust sample pairs (Figure 2b,c) represent separate resuspensions of the same grab sample, giving somewhat different mass loadings between the two samples. The pairs of green and dead vegetative detritus samples (Figure 2d,e) also were resuspended in separate experiments, using the same Environ. Sci. Technol., Vol. 25, No. 7, 1991
1315
s
3.5 but resulted in some type of visible injury or functional alteration from mists with pH I3.0. Few data exist for mists with pH values between 3.0 and 3.5. Ozone has been shown to injure crops (13)and to affect some tree species (14-1 7). Controlled exposures typically use square-wave patterns for ozone (no peaks); a few recent studies used more realistic exposure scenarios (18). Ozone and acidic cloudwater represent chemical stresses for high-elevation forests that are already subjected to natural competition, biotic stresses, and climate extremes ( 4 ) . Although a single co-occurrence event was observed in Maine (19),t,he dynamics of ozone and acidic cloudwater 'Oregon State University. *University of Washington.
exposures and their co-occurrence have not been identified previously. ~~~~~i~~ that an insultlrepair cycle is involved in pollutant injury to trees(2&22), the temporal characteristics of the ambient exposure regime are importantfor determining the degree of imposed stress. Despite a lack of clearcut laboratory evidence of ozone or acid injury to red spruce, visible decline symptoms are observed at high elevation in the Appalachian Mountains (3). The exclusion of ambient cloudwater and ozone from mature red spruce reduced these symptoms in one study (A. H. Johnson, personal communication). Therefore, we have asked how the ambient exposure of forests to ozone and acidity differs from controlled exposures. Experimental Section Ambient ozone and cloudwater were collected over the 1986-1988 growing seasons (April-October) at five sites in the Appalachian Mountains: Whiteface, NY (44'23' N, 73'5' W, 1483-m elevation), Moosilauke, NH (40'1' N, 71'50' W, 1000 m), Shenandoah, VA (37'72' N, 78'20' W, 1000 m), Whitetop, VA (36'39' N, 81'36' W, 1686 m), and Mitchell, NC (35'44' N, 82'17' W, 2006 m). The gases ozone [UV photometric technique (23)1, SO2 [UV fluorescence technique ( 2 4 ) ] ,and H202 [reaction with p-hydroxyphenylacetic acid (25)] were measured (Table I). Cloudwater was collected with a passive string collector (26) and analyzed for solution pH (electrometric technique), aqueous SO:-, and NO< (ion chromatograph). Following quality assurance procedures ensured a t least f15% accuracy and precision (27). Data Analysis Methodology Atmospheric pollution episodes were classified as one of five types (Figure 1)by comparing hourly ambient data to [ozone] I70 ppbv and pH I 3.2, cutoff levels that are stringent enough to be potentially damaging to trees, while accepting enough data for characterization of distributions. We assumed that the movement between the pollution episodes is stochastic so that, given the pollution history up to the present, the conditional probability of going to any other episode depends only on the current episode type (except respites between episodes, which depend on the current and previous episode type). Furthermore, we as-
0013-936X/9110925-1325$02.50/0 0 1991 American Chemical Soclety
Environ. Sci. Technol., Vol. 25, No. 7, 1991 1325
