Particle Size Distributions of Organic Aerosol Constituents during the

Jun 29, 2006 - Sampling sites were located on Turtleback Dome (37.7125° N; 119.7042° W; 1615 m above sea level), located above the entrance to Yosem...
0 downloads 0 Views 288KB Size
Environ. Sci. Technol. 2006, 40, 4554-4562

Particle Size Distributions of Organic Aerosol Constituents during the 2002 Yosemite Aerosol Characterization Study PIERRE HERCKES,† GUENTER ENGLING, SONIA M. KREIDENWEIS, AND JEFFREY L. COLLETT, JR.* Colorado State University, Department of Atmospheric Science, Fort Collins, Colorado 80523-1371

The Yosemite Aerosol Characterization Study (YACS) was conducted in the summer of 2002 to investigate sources of regional haze in Yosemite National Park. Organic carbon and molecular source marker species size distributions were investigated during hazy and clear periods. More than 75% of the organic carbon mass was associated with submicron aerosol particles. Most molecular marker species for wood smoke, an important source of particulate matter during the study, were contained in submicron particles, although on some fire influenced days, levoglucosan shifted toward larger sizes. Various wood smoke marker species exhibited slightly different size distributions in the samples, suggesting different, size dependent emission or atmospheric processing rates of these species. Secondary biogenic compounds including pinic and pinonic acids were associated with smaller particles. Pinonaldehyde, however, exhibited a broader distribution, likely due to its higher volatility. Dicarboxylic acids were associated mainly with submicron particles. Hopanes, molecular markers for vehicle emissions, were mostly contained in smaller particles but exhibited some tailing into larger size classes.

Introduction Variations in aerosol particle composition with particle size are well-recognized for many species. Differences in composition reflect variations in particle generation mechanisms, from the mechanical generation of coarse (generally supermicron) particles to nucleation processes that produce much smaller particles (1). Additional particle mass can also be formed through vapor condensation onto existing particles and cloud processing. While numerous studies have addressed particle size distributions of inorganic species, data are relatively scarce for organic carbon and, especially, for individual organic compounds. Some investigators have reported particle size distributions of organic carbon (2-8), and particle size distributions of polycyclic aromatic hydrocarbons (PAH) have been studied rather extensively (e.g., refs 9-12). Finally, a few studies have reported particle size distributions of other organic species, often including secondary species of biogenic origin (2, 13-17) or dicarboxylic acids (e.g., refs 12 and 18-20). * Corresponding author phone: (970) 491-8697; fax: (970) 4918449; e-mail: [email protected]. † Present address: Arizona State University, Department of Chemistry and Biochemistry, Tempe, AZ 85287-1604. 4554

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 15, 2006

In recent years, a number of air pollution studies used particular organic species, commonly referred to as molecular markers, to apportion sources of atmospheric particulate matter (e.g., refs 21-23). This is accomplished by linking ambient concentrations of ambient species through chemical mass balance (CMB) models to known emission profiles. Implicit in the CMB approach is an assumption that the marker species are conserved during transport. One element of this assumption is that the size distributions of carbonaceous particles produced by individual source types are reflected in the size distributions of the source marker species. Levoglucosan, for example, is assumed to have the same particle size distribution as wood smoke particles in general. If this were not true, one might anticipate that the ratio of levoglucosan to OC in the source plume would not be conserved. For example, a marker preferentially associated with larger particle sizes, relative to the general source particle size distribution, might undergo more extensive cloud processing and experience different rates of dry and wet deposition. Unfortunately, little is known about particle size distributions of key molecular marker species. This represents a serious shortcoming in the use of these compounds for source apportionment purposes. Determining molecular marker particle size distributions is especially important in cases involving long-range transport, when opportunities for cloud processing and both wet and dry deposition processes are increased. The present study was designed to increase our knowledge of atmospheric molecular markers by investigating their size distributions in ambient aerosol samples collected during an intensive field study in Yosemite National Park, CA. Samples were collected in both hazy and clean periods. A special focus was on biogenic marker compounds, mainly species derived from biomass burning and secondary organic aerosol formation, due to their important contribution to local summertime haze (24).

Experimental Procedures Aerosol samples were collected in Yosemite National Park, CA, in summer 2002 in the framework of the Yosemite Aerosol Characterization Study (YACS) (25-31). Sampling sites were located on Turtleback Dome (37.7125° N; 119.7042° W; 1615 m above sea level), located above the entrance to Yosemite Valley, and in Yosemite Valley in a clearing next to the Merced River (37.7431° N; 119.5842° W; 1219 m above sea level). Aerosol samples were collected using ThermoAndersen Hi-Volume samplers, equipped with five-stage impactor inlets (TA 235, Thermo Andersen, Smyrna, GA). These samplers allow for the collection of ambient particles in six size classes by five impaction stages (cutoff aerodynamic diameters of 7.2, 3, 1.5, 0.95, and 0.49 µm) (32) and one back-up filter, collecting particles smaller than 0.49 µm. Samples were collected on quartz fiber filters prebaked at 600 °C for at least 8 h. Sampling times varied from 24 h to several days depending on atmospheric conditions (hazy days vs clear days). Blank samples were taken by mounting blank filters into the collector and operating the collector for 2 min. None of the individual species detailed in this paper has been detected in the blank samples. Total carbon (TC) as well as organic (OC) and elemental carbon (EC) concentrations were determined by a Thermal Optical Transmission method (33) using a Sunset Laboratories semicontinuous instrument in offline mode. A significant positive artifact in particulate organic matter determinations might result from adsorption of volatile gaseous organic compounds on the pre-fired quartz fiber filters. To 10.1021/es0515396 CCC: $33.50

 2006 American Chemical Society Published on Web 06/29/2006

TABLE 1. Overview of Analyzed Samples location

start and end times/dates

YS071401

sample

clean

condition

Turtleback Dome

YS072501

mostly clean 1

Turtleback Dome

YS072901

mostly clean 2

Turtleback Dome

YS081501

smoke 1

Turtleback Dome

YS081601

smoke 2

Turtleback Dome

YS082801

local fire

Turtleback Dome

YS0816BLK YV090301

blank valley weekend

Turtleback Dome Merced River

YV090501

valley weekday

Merced River

07/13/02 1055 07/14/02 0705 07/22/02 0830 07/25/02 0743 07/28/02 0811 07/29/02 0732 08/14/02 0822 08/15/02 0707 08/15/02 0740 08/16/02 0759 08/25/02 0845 08/28/02 0733 08/16/02 0830 08/30/02 0831 09/03/02 0818 09/03/02 0849 09/05/02 1020

correct for this artifact, the filters were sliced into front and back halves (cf. ref 34). It was assumed that all particulate material is captured on the front half of the filter and that the front and back halves of the filter collect equal amounts of volatile organic carbon by adsorption. Organic carbon was measured on the front and back halves, and the OC measured on the back half was subtracted from the front half value to correct for this positive adsorption artifact. The adsorption artifact proved to be especially important for the larger size fractions with their lower OC concentrations. In some cases, the adsorbed gas fraction was as important as the particulate fraction resulting in corrections of 50% or more. No corrections were made for possible negative sampling artifacts resulting from losses of collected, semivolatile material. Organic molecular markers were determined by gas chromatography coupled to mass spectrometry (GC/MS) following extraction with methylene chloride (Fisher, Optima grade). The procedure is described in detail elsewhere (35). In brief: filters were spiked with a series of deuterated internal standards and then extracted 3 times with methylene chloride under sonication. The extracts were combined and reduced to 250 µL before being split into three aliquots. The first aliquot was injected underivatized into the gas chromatograph (GC) and analyzed for lower polarity compounds including polycyclic aromatic hydrocarbons (PAH), methoxyphenols, and n-alkanes. A second aliquot was derivatized by diazomethane for the determination of carboxylic acids as their methylesters. The final fraction was derivatized with bis(trimethylsilyl)trifluoroacetamide (BSTFA) and chlorotrimethylsilane (TMCS), added as a catalyst, to transform anhydrosugars and sterols into their trimethylsilyl esters prior to GC/MS quantification. Samples were chosen to represent different locations and conditions encountered during the study, based on backtrajectories as well as co-located measurements and observed variability in source contributions (29, 31). Table 1 gives an overview of sampling conditions for selected periods. A special focus of the YACS study in general and of the present work in particular is the measurement of molecular markers for wood smoke. A general overview of organic molecular markers in fine particulate matter and their variation during the study is given elsewhere (31).

Results and Discussion Total Carbon. Total (TC) carbon concentrations for several representative days are given in Figure 1. On most days, the majority of TC is found in the smallest size class (daero < 0.49 µm). In all samples, more than 75% of the TC is present in

characteristics clear day mostly clear day with night smoke clear with limited smoke influence regional smoke Oregon fires regional smoke Oregon fires local Fire blank Labor day weekend Friday through Tuesday Tuesday through Thursday after Labor day

submicron aerosol particles. These observations are consistent with other measurements made during YACS. Malm and co-workers (26) found that the composition of fine particles during YACS was dominated by organic matter. McMeeking and co-workers (27), using a combined differential mobility analyzer-optical particle counter system, observed that most particles were smaller than one micrometer. On very hazy days, we observed that the carbon mass distribution shifted toward larger particle sizes but still resided mostly in the submicron range. McMeeking and co-workers (28) observed a similar shift in the total particle size distribution between clear and smoky periods. Other data (29), including multiwavelength aethalometer measurements of aerosol black carbon (BC) and fine particle K+ concentrations, indicate that there was a strong smoke contribution to ambient particle concentrations on these days. In sample smoke 1, a second supermicron mode was observed, but its origin remains unknown. For the two valley samples, we observed little difference between the Labor Day weekend sample (increased visitor traffic) and the sample collected after that weekend. We obtained size resolved OC (Figure 2) and EC data as well. These data are somewhat less reliable than TC measurements due to the need to properly account for optical transmission changes (to appropriately determine the OC/ EC split) through nonuniform filter loadings resulting from the impaction jets. The very low concentrations of EC observed in all YACS samples also add uncertainty to EC size distribution measurements. OC size distribution data track closely with the TC distributions. The EC data show an apparent second size mode on clear days, although this might be a measurement artifact associated with the very low EC concentrations. The observations from Yosemite are consistent with previous observations that organic matter is preferentially associated with small (submicron) particles. Temesi and coworkers (6) found that total carbon in rural Hungary showed a distribution maximum around 0.53 µm, and very little carbon mass was associated with the supermicrometer aerosol. Maenhaut and co-workers (3) found that most OC was in the submicrometer size range, as did Plewka and coworkers (5) and Jaffrezo and co-workers (8). Viidanoja and co-workers (7) compared OC and BC in PM2.5 and PM10 samples and found that 90% of BC and 73% of OC are associated with PM2.5, although on a few events, coarse OC could account for up to 67% of the total OC. Finally, Miguel and colleagues studied seasonal variability in OC, EC, and organic trace species size distributions in Claremont, CA (4). VOL. 40, NO. 15, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

4555

FIGURE 1. Total carbon size distributions for different periods during the Yosemite Aerosol Characterization Study (YACS).

FIGURE 2. OC size distributions for different sampling conditions during YACS. During the summer months (June and July), they observed that the largest amount of EC was in particle sizes below 0.1 µm, with small amounts of EC in larger particle sizes (>1 µm). Some studies report bimodal size distributions of organic carbon, but comparisons are difficult as different studies use different size cuts and size resolutions (2, 3). Wood Smoke Marker Size Distributions. Levoglucosan. A widely used marker for biomass burning is the anhydrosugar levoglucosan (36). It is a major species emitted by combustion of cellulose (36-39). Despite its wide use in source apportionment studies (40, 41), only two studies discuss particle size distributions of levoglucosan (42, 43). Levoglucosan size distributions observed during YACS are presented in Figure 3. Levoglucosan is observed on all days, clear and hazy, illustrating the ubiquity of biomass burning contributions to sum4556

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 15, 2006

mertime aerosol in Yosemite National Park, at least during summer 2002. Concentrations are nevertheless lowest during the clear periods. We observe higher concentrations of levoglucosan during the fire-influenced periods represented by samples smoke 1 and smoke 2. Most of the levoglucosan is observed in submicron particles for all sample periods, consistent with findings from other recent studies (42, 43). Interestingly, for sample smoke 1, the distribution shifted toward larger particles, while on the next day (smoke 2), also a regional haze day with strong smoke influence, the distribution maximum was flatter, shifting back toward the smallest size range (the height of the histogram for the smallest particle size range is determined in part by the assumed lower cutoff of this size bin, chosen to be consistent with typical size distributions measured by the differential mobility analyzer). The shift to smaller particle sizes for

FIGURE 3. Particle size distributions of levoglucosan.

FIGURE 4. Particle size distributions of retene. levoglucosan is somewhat surprising considering the shift toward larger sizes observed for organic carbon and mass on these days, which we attributed to wood smoke based on other supporting observations (K+, aethalometer measurements, and back-trajectories arriving from known fire source regions). Differences in the size distributions for OC and levoglucosan in this wood smoke influenced period could result from important contributions by other organic compounds. Engling et al. (31) suggest that biogenic SOA compounds, for example, remain quite important even during periods of strong smoke influence and that production of SOA in the smoke plume itself may account for much of the observed organic carbon during this period. The differing size distributions could reflect the degradation of levoglucosan during multi-day transport of the smoke plume, which is believed to have come from fires burning in southern Oregon. Acid hydrolysis of levoglucosan (35, 37) in solution, for example, could lead to its degradation during cloud processing, although cloud pH values along the northern California and Oregon coasts, in California’s Central Valley,

and in Yosemite are typically fairly high (44-46). Laboratory studies also reveal relatively rapid loss of levoglucosan in authentic cloud samples after collection, perhaps due to degradation through biological pathways. Finally, ongoing work suggests that levoglucosan emissions differ widely with burning regime (e.g., flaming vs smoldering) (47), and nothing is known of the impact on size resolved emissions. Methoxyphenols. Methoxyphenols, derived from the thermal degradation of lignin, are also common constituents of smoke from biomass combustion (36, 38, 39). Methoxyphenol emission rates are strongly dependent on fuel type (38, 4853). Methoxyphenols are emitted mainly by softwood combustion and are present only in small amounts in smoke from other types of biomass. While these compounds have been determined in source samples (39, 48-53) and are frequently included in aerosol characterization studies (23, 35), no previous information has been published concerning their particle size distributions in the ambient atmosphere. Figure A (Supporting Information) gives an example of acetovanillone (4-acetyl-2-methoxyphenol) size distributions VOL. 40, NO. 15, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

4557

FIGURE 5. Particle size distributions of pimaric acid observed during YACS. observed during YACS. On the regional haze/smoke days (smoke 1 and smoke 2), we observe the highest overall concentrations. While concentrations on clean days are low, acetovanillone is nevertheless ubiquitous during the YACS study. Acetovanillone is mainly present in smaller particle sizes, but small amounts also appear to be associated with larger, supermicron particles. For the smoke influenced days (smoke 1 and smoke 2), we do not see a shift toward larger particles, somewhat in contrast to the observations for total particle mass and OC. Interestingly, on Labor Day weekend, the acetovanillone size distributions in the valley do shift toward larger particles. Many methoxyphenols may be photochemically degraded in the atmosphere (54). The semivolatile nature of many methoxyphenols can also affect their real and apparent size distributions. Artifact formation in sampling, due to losses to the gas phase and recondensation on lower stages, can shift apparent size distributions to smaller particles. Condensation of atmospheric gaseous methoxyphenols onto quartz impaction substrates can result in an apparent distribution broadening, especially toward larger particle sizes collected on initial impaction stages. Methoxyphenols might also redistribute in the atmosphere across the particle size distribution due to volatilization and recondensation processes. These phenomena could have contributed to broadening of the measured methoxyphenol size distributions as compared to levoglucosan. Retene. Retene has been used for many years as a molecular source marker for wood smoke in atmospheric aerosols (55). It is emitted by combustion of pine wood (56) and has been identified in numerous aerosol studies (e.g., ref 35). Despite its popularity as a wood smoke marker, the size distributions observed during YACS (Figure 4) are the first known measurements in ambient aerosol. We observe that in general, the size distributions are rather broad with distribution maxima around 1 µm. In all cases, there is a significant fraction of retene in supermicron particles; however, as in the case of methoxyphenols, broadening of the distribution might be an artifact due to the semivolatile nature of the compound. The largest concentration appears to be in the sample YS082801, which was influenced by a local fire. During regional smoke events, the distributions are rather broad and very different from the previously described marker size distributions. 4558

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 15, 2006

Resin Acids. Resin acids have also been used as molecular markers for wood smoke emissions (57-59). Measured size distributions for several resin acids are illustrated in Figure 5 and in the Supporting Information. Interestingly, we observe higher concentrations for resin acids in the Valley than at Turtleback Dome, in contrast to the other wood smoke markers. This difference might result from localized use of particular fuel types for campfires, which are concentrated in the Yosemite Valley, although information is not available to test this hypothesis. For dehydroabietic acid (see Figure B, Supporting Information) and 7-oxodehydroabietic acid (see Figure C, Supporting Information), most of the mass is in the submicron mode, and the smallest size fraction is usually the dominant one. Pimaric (Figure 6) and isopimaric (see Figure D, Supporting Information) acids, on the other hand, show much wider distributions that are shifted toward larger particle sizes, at times even appearing bimodal. Overall, most wood smoke markers, except retene and acetovanillone, are associated with submicron aerosol particles (>75% of the mass of the marker species is present in submicron particles). Wood smoke marker size distributions are generally similar to mass and OC size distributions, although they differ significantly on a few occasions. It is most intriguing that for specific samples, different molecular markers for wood smoke exhibit distinctly different size distributions (Figure 6 and Figure E in Supporting Information), even within a compound family such as resin acids (see Figure F, Supporting Information). One possible explanation is that the emission rates and resulting particle size distributions of wood smoke marker species depend on fuel type (39, 48, 50-53) and burn conditions (47, 60), something that needs to be further investigated. Overlapping contributions by multiple smoke sources, each with different fuel and burning conditions, might also contribute to divergent results. Different species emission and degradation rates could also contribute to observed differences. Longrange transport of smoke from Oregon wildfires was responsible for much of the smoke measured in Yosemite in early to mid-August 2002 (29), and some more labile species might have been degraded more rapidly during transport than others. This possibility stresses the need for a careful evaluation of molecular marker stability during long-range transport, including cloud processing.

FIGURE 6. Particle size distributions of selected wood smoke markers for sample smoke 2.

FIGURE 7. Particle size distributions of pinic acid. Secondary Biogenic Compounds. Terpene Oxidation Products. The chemical composition of fine particulate matter during YACS showed high concentrations of SOA compounds derived from biogenic precursors, including products resulting from monoterpene oxidation (31). Specifically, concentrations of pinic acid, cis-pinonic acid, pinonaldehyde, and nopinone have been determined. These compounds have been quantified in previous studies (61-64) in ambient particles, and a few studies even addressed size distributions of secondary biogenic compounds (13, 14). Pinic acid (Figure 7), cis-pinonic acid (see Figure G, Supporting Information), and nopinone (see Figure H, Supporting Information) were mostly found in the smallest particle size fraction. This is consistent with previous observations and with formation by nucleation in forested atmospheres (64, 65). We also observe, however, small amounts of these secondary biogenic compounds associated with larger particles, consistent with some observations (13) but

in contrast to others (14). The broadening of the size distribution might be the result of condensation of VOC reaction products onto larger existing aerosol particles or of particle coagulation, especially during long-range transport of the smoke plume, which certainly should contain a considerable biogenic VOC load in addition to smoke particles. Pinonaldehyde (Figure 8) exhibits a different pattern than other secondary biogenic species. Its size distribution is much wider, peaking around 1 µm in diameter, suggesting that condensation of pinonaldehyde onto existing particles is probably an important mechanism for introducing pinonaldehyde to atmospheric particles as suggested by Kavouras and Stephanou (66). However, while other authors see, even for pinonaldehyde, more than 90% in the submicron aerosol, we observed on average 30% of the pinonaldehyde associated to supermicron particles. An explanation might be that a more important supermicron particle population is present during YACS on which pinonaldehyde can condense. VOL. 40, NO. 15, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

4559

FIGURE 8. Particle size distribution of pinonaldehyde.

FIGURE 9. Particle size distributions of 17r 21β hopane. Dicarboxylic Acids (C4-C12). Another important class of organic species, frequently formed through atmospheric reactions, is dicarboxylic acids. Some authors have previously reported size distributions for dicarboxylic acids (e.g., ref 19); see Figure I in Supporting Information, which presents adipic acid as an example for size distributions of dicarboxylic acids observed during YACS. In nearly all cases, the dicarboxylic acids exhibit a monomodal distribution with a maximum below 0.49 µm. In a number of samples, no dicarboxylic acids were detected in particle sizes larger than 1 µm. In the clear day sample YS071401, for example, dicarboxylic acids were only detected in the smallest particle size fraction (