Research Emissions of Levoglucosan, Methoxy Phenols, and Organic Acids from Prescribed Burns, Laboratory Combustion of Wildland Fuels, and Residential Wood Combustion LYNN R. MAZZOLENI,† BARBARA ZIELINSKA,* AND HANS MOOSMU ¨ LLER Desert Research Institute, Division of Atmospheric Science, 2215 Raggio Parkway, Reno, Nevada
Biomass combustion emissions make a significant contribution to the overall particulate pollution in the troposphere. Wildland or prescribed burns and residential wood combustion emissions can vary due to differences in fuel, season, time of day, and the nature of the combustion. Inadequate understanding of the relevance of these biomass combustion emissions is due to the lack of characterization of open combustion emissions and the limited understanding of the differences between these and residential wood combustion. To provide new insight to biomass combustion emissions, sampling was conducted in several types of conditions. Semi-volatile organic compounds (SVOC) were collected during four separate prescribed burns in three different ecosystems, Mariposa Sequoia Grove within Yosemite National Park, CA, desert brushes of central rural Nevada, and Toiyabye National Forest near Lake Tahoe, NV. SVOC samples were also collected under controlled conditions for several wildland fuels, including conifer needles, wildland grasses, and sagebrush. Fireplace emissions from simulated residential wood combustion were also collected and are included here for comparison. A high degree of variability was found in the emissions of organic carbon, elemental carbon, levoglucosan, methoxy phenols, and organic acids. The variability in the emissions of levoglucosan does not correlate with the PM2.5 gravimetric mass and thus may affect source apportionment estimates.
Introduction Biomass combustion emissions arising from natural fires, prescribed burns, and residential wood combustion contribute significant amounts of particulate matter (PM) to the troposphere (1). Wood smoke emissions from residential fireplaces have been shown to be a major source of PM in a number of communities in the United States (2-4). In addition, wood smoke emissions from wildfires and prescribed burns are responsible for occasional severe episodes of air pollution (5, 6). Wildland managers use prescribed * Corresponding author phone: (775) 674-7066; e-mail:
[email protected]. † Present address: Colorado State University, Atmospheric Science Department, Fort Collins, CO 80523. 10.1021/es061702c CCC: $37.00 Published on Web 02/24/2007
2007 American Chemical Society
fires and controlled natural ignition fires under the “let burn” policy to protect manmade structures, residential communities, and natural resources while trying to limit their impact on air quality and atmospheric visibility. This is accomplished by reducing accumulated dead biomass fuels (such as duff and logs on the forest floor) and thinning the under-story foliage density when burning conditions are not severe. Biomass combustion smoke contains a mixture of gaseous and particulate carbonaceous compounds with a complex chemical composition (7-10). Variability in combustion emissions arises from differences in fuel composition and structure, the fire intensity, aeration, ambient and vascular moisture, duration of smoldering and flaming conditions, and the physical surroundings such as the ground slope. Several studies have been conducted to characterize the organic composition of the PM emissions from residential wood combustion (8, 11-14); however, only few studies have been performed on the combustion of wildland fuels (15) and prescribed burns (16). Major pyrolysis products arising from cellulose, lignin, and resins have been proposed as tracers for biomass combustion emissions (5, 17-19). The anhydrosugar levoglucosan, a pyrolysis product of cellulose and hemi-cellulose, has been proposed as a molecular marker for the long-range transport of biomass combustion aerosols (5, 20). In this paper, we present PM2.5 associated semi-volatile organic compound (SVOC) profiles for a variety of biomass burning fuels and scenarios including residential wood combustion and in-field prescribed burning. Information regarding the emission rates for the source tests of wildland fuels can be found in Chen et al. (21)
Experimental Procedures Fuels and Fuel Combustion. Prescribed Burns. Samples of emissions from prescribed burning were collected in the Nevada Great Basin, in the Yosemite Mariposa Sequoia Grove, CA, and in the Toiyabe National Forest, north of Lake Tahoe near Incline Village, NV (Table 1). Samplers were set up within 1 mile downwind of ambient prescribed fires conducted in (i) Yosemite National Park’s Mariposa Sequoia Grove, where a plot of 50 acres of land was burned with low intensity for fuel reduction, (ii) an Ecological Study Plot managed by the University of Nevada, Reno near Austin, NV, where about 5 acres were burned with heavy drip-torch application to ignite desert brushes for restoration of the native vegetation, and (iii) the Toiyabe National Forest, north of Lake Tahoe near Incline Village, NV, where two plots of 10 acres (mixed conifer forest with chaparral vegetation including White fir, Jeffery pine, and Ponderosa pine) were burned for fuel reduction and wildfire prevention. Sampling was conducted for approximately 30 min near each of the four plots during the summer months. Wildland Fuels. A variety of wildland fuels from the U.S. and Africa were selected to investigate physical and chemical differences in their combustion emissions. Biomass samples included White pine needles (WPN), Ponderosa pine needles (PPN), desert sagebrush (SAG), Zambia savanna grass (ZG), Ponderosa pine sticks (PPS), excelsior poplar insulation (EXC), an Alaskan feather moss tundra core, fresh wet Montana grass, and a collection of mixed wood branches. These wildland fuels were burned in the combustion laboratory of the United States Department of Agriculture, Forest Service Fires Sciences Laboratory (FSL) located in Missoula, MT. The fuel was burned in a large room under controlled VOL. 41, NO. 7, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2115
TABLE 1. Overview of Biomass Sample Collection and Selected Measurements (µg/m3) sample ID
biomass sample
no. of firesa PM2.5 massb organic carbonb elemental carbonb levoglucosanb
combustion type
Wildland Fuels FSL Source Samples PPN B PPN F1 PPN F2 PPN H WPN B WPN F1 WPN F2 WPN H PPS F1 PPS F2 EXC F1 EXC F2 SAG B SAG F1 SAG F2 SAG H ZG B ZG F1 ZG F2 ZG H MT grass tundra core mixed woods
Ponderosa pine needles Ponderosa pine needles Ponderosa pine needles Ponderosa pine needles White pine needles White pine needles White pine needles White pine needles Ponderosa pine sticks Ponderosa pine sticks excelsior excelsior sagebrush sagebrush sagebrush sagebrush Zambia grass Zambia grass Zambia grass Zambia grass Montana grass (wet) Alaskan tundra core mixed woods
backing flat 1 flat 2 heading backing flat 1 flat 2 heading flat 1 flat 2 flat 1 flat 2 backing flat 1 flat 2 heading backing flat 1 flat 2 heading heading and backing flat flat
oak almond Tamarak pine eucalyptus cedar 1 cedar 2 wheat straw rice straw
oak wood almond wood Tamarak pine wood eucalyptus wood cedar wood cedar wood wheat straw rice straw
fireplace grate fireplace grate fireplace grate fireplace grate fireplace grate fireplace grate fireplace grate fireplace grate
NLT 1 AM NLT 2 AM NLT 3 PM NLT 4 PM UNRR 1 UNRR 2 YOSE 1 YOSE 2
Toiyabye Forest 1 Toiyabye Forest 1 Toiyabye Forest 2 Toiyabye Forest 2 Nevada Basin Nevada Basin Yosemite Mariposa Grove Yosemite Mariposa Grove
10 acre plot 1 AM 10 acre plot 2 AM 10 acre plot 1 PM 10 acre plot 2 PM 10 acre plot 1 10 acre plot 2 50 acre plot 1 50 acre plot 2
3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 2 1 1
209 ( 4.5 1930 ( 4.5 327 ( 4.5 364 ( 4.5 402 ( 4.5 758 ( 4.5 640 ( 4.5 1300 ( 4.5 361 ( 4.5 337 ( 4.5 652 ( 4.5 457 ( 4.5 583 ( 4.5 517 ( 4.5 575 ( 4.5 506 ( 4.5 88.6 ( 4.5 316 ( 4.5 148 ( 4.5 65.6 ( 4.5 346 ( 4.5 201 ( 4.5 520 ( 4.5
97.5 ( 7.3 731 ( 76.5 149 ( 9.7 257 ( 14.4 287 ( 16.1 471 ( 25.4 463 ( 25.1 934 ( 49.3 85.9 ( 8.0 76.2 ( 7.2 253 ( 15.4 225 ( 14.6 192 ( 11.2 234 ( 13.6 196 ( 11.9 206 ( 12.1 96.2 ( 7.2 198 ( 11.9 105 ( 8.1 64.0 ( 6.3 300 ( 17.0 183 ( 9.9 361 ( 19.1
146 ( 23.1 462 ( 75.0 196 ( 30.9 103 ( 16.3 95.0 ( 15.1 133 ( 21.0 112 ( 17.7 72.7 ( 11.5 316 ( 50.0 253 ( 39.8 275 ( 43.4 243 ( 38.5 198 ( 31.2 123 ( 19.5 134 ( 21.2 94.4 ( 15.0 47.1 ( 7.5 72.1 ( 11.5 58.2 ( 9.4 31.3 ( 5.0 41.0 ( 6.6 11.0 ( 1.8 104 ( 16.4
5.60 ( 0.7 13.2 ( 1.7 9.70 ( 1.3 25.2 ( 3.3 43.2 ( 5.6 100 ( 13.1 70.0 ( 9.1 130 ( 16.9 3.35 ( 0.45 3.80 ( 0.51 41.3 ( 5.4 38.3 ( 5.0 4.74 ( 0.6 12.4 ( 1.6 3.79 ( 0.51 11.1 ( 1.5 8.43 ( 1.1 66.0 ( 8.6 16.1 ( 2.1 5.68 ( 0.75 126 ( 16.5 26.1 ( 3.4 53.4 ( 7.0
139 ( 8.6 121 ( 7.5 287 ( 17.8 711 ( 44.0 1360 ( 84.3 1230 ( 75.8 253 ( 15.7 25.1 ( 2.3
9.83 ( 0.7 23.4 ( 1.5 94.7 ( 6.2 30.2 ( 2.0 359 ( 23.3 111 ( 7.2 129 ( 8.4 4.42 ( 0.5
5.11 ( 0.26 7.60 ( 0.38 2.09 ( 0.10 14.4 ( 0.72 14.4 ( 0.72 12.7 ( 0.63 28.6 ( 1.4 0.84 ( 0.04
288 ( 15.3 26.6 ( 1.8 495 ( 26.1 96.7 ( 5.3 134 ( 9.1 89.6 ( 6.3 121.1 ( 8.3 38.9 ( 3.4
41.4 ( 3.0 3.05 ( 0.35 25.3 ( 1.8 11.4 ( 0.89 17.7 ( 4.5 8.27 ( 2.2 17.0 ( 4.3 3.84 ( 1.1
12.3 ( 1.5 0.49 ( 0.06 44.4 ( 5.3 6.05 ( 0.73 3.85 ( 0.51 4.56 ( 0.61 5.31 ( 0.71 4.28 ( 0.57
Residential and Agricultural Fireplace Source Samples 1 1 1 1 1 1 1 1
223 ( 4.5 453 ( 4.5 348 ( 4.5 504 ( 4.5 2720 ( 4.5 883 ( 4.5 1250 ( 4.5 15.6 ( 4.5
In-Field Prescribed Burn Samples
a
Individual fires were collected onto one sampling media set.
equation: analyte uncertainty )
b
1 1 1 1 1 1 1 1
Measurements include the analytical uncertainties calculated by the following
x(precision × concentration)2 + (MDL)2.
conditions on a continuously weighed fuel bed. The resulting smoke was captured by a 3.6 m diameter inverted funnel into a 1.6 m diameter stack (Figure S1). The room was pressurized with outside air that was conditioned to standard temperature and relative humidity (25 ( 1 °C and 25 ( 5%) and was then vented through the stack at a flow rate of 2.5 m/s, completely entraining the emissions from fires burning beneath the funnel. A sampling platform surrounded the stack 18 m above the fuel bed. Samples were drawn from the stack into a small residence chamber with ports for PM2.5 sample collection. The total residence time for gas-to-particle phase partitioning was approximately 20 s. Additional ports of the residence chamber were utilized for low flow optical and particle morphological measurements, which are discussed elsewhere (21-23). Wildland fuels were burnt in a series of three small fires with an identical setup on a flat fire bed (except for combustion of Alaskan feather moss tundra core, fresh wet Montana grass, and collected mixed wood branches, where fires were not replicated). Four of the fuels were also burned on an angled fire bed placed at a 45° angle to simulate heading and backing fires (Table 1). Small amounts of biomass, nominally 250 g (1250 g for tundra cores), were combusted in each fire with a sampling period of approximately 5 min. To collect enough PM mass for suitable organic speciation, a series of three fires was sampled as a composite onto a set of sampling media. 2116
9
776 ( 4.5 58.0 ( 4.5 1560 ( 4.5 344 ( 4.5 592 ( 4.5 376 ( 4.5 354 ( 4.5 106 ( 4.5
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 7, 2007
Residential Wood and Agricultural Combustion. Residential wood and agricultural combustion fuels were collected during the California Regional Particulate Air Quality Study to investigate biomass combustion emissions significant in the Central Valley of California. These fuels include oak, pine, eucalyptus, cedar, and almond woods, rice straw, and wheat straw. Experiments were conducted at Desert Research Institute (DRI) in a concrete block shed, 11 ft wide × 9 ft deep × 8 ft tall. Residential wood smoke was simulated using a commercial wood burning fireplace. An 18 ft × 8 in. i.d. stainless steel exhaust flue extended through the center of the roof from the fireplace. Samples were collected from the flue using the DRI dilution sampling system (8). The dilution sampler was a modified version of the model built and described by Hildeman et al. (24). Emissions from the exhaust duct were diluted with HEPA and charcoal filtered ambient air (∼0-15 °C) in the dilution tunnel by a factor of approximately 20 with 80 s of residence time for gas-toparticle phase partitioning. Samples were collected over a 4 h period, except for agricultural straws where samples were collected over a 2 h period. Residential wood and agricultural straws were purchased in the Fresno, CA area. Replicate fireplace emission samples from oak and cedar wood samples were collected. The fires were started with 0.5 kg of kindling underneath 5 kg of wood and ignited with a propane torch. After the kindling was
completely burned, the sampling was started. The fire was then recharged with 5 kg of wood after approximately 30% of the original load burned to maintain a vigorous flame (but not so vigorous as to extend into the flue) throughout the sampling period. Agricultural burning emissions of rice and wheat straw were generated within the commercial wood burning fireplace. Straw was difficult to burn with this setup. The rice straw clumped tightly together, and combustion tended toward smoldering; therefore, the clumps were periodically mixed to prevent the fire from fading and extinguishing itself. Note that the fireplace combustion of rice and wheat straws may not yield emissions directly relevant to the combustion methods common in agricultural practice. Sample Collection Methods. PM2.5 samples were drawn through cyclone separators with a cutoff diameter of 2.5 µm and then collected onto Teflon-impregnated glass-fiber (TIGF) filters followed by PUF/XAD/PUF “sandwich” cartridges using a DRI fine particle semi-volatile organic sampler. Simultaneous collection of samples for PM2.5 gravimetric mass on Teflon filters (Pall-Gelman, 47 mm) and for thermal optical carbon analysis on quartz fiber filters (Pall-Gelman, 47 mm, heat treated) was conducted as described by Chow (25). After sampling, filters were immediately refrigerated and kept cold during overnight shipping. Samples were stored in a freezer until analysis. Analytical Methodology. Thermal carbon analysis was conducted following the IMPROVE protocol (26). Several isotopically labeled internal standards were added to each TIGF filter and PUF/XAD/PUF cartridge sample prior to extraction. The isotopically labeled standards used included hexanoic-d11 acid, benzoic-d5 acid, succinic-d4 acid, decanoicd19 acid, adipic-d10 acid, suberic-d12 acid, levoglucosan-13C6, homovanillic-2,2-d2 acid, myristic-d27 acid, heptadecanoicd33 acid, oleic-9,10-d2 acid, tetradecanedioic-d24 acid, and cholesterol-2,2,3,4,4,6-d6. The TIGF filter was extracted separately from the PUF/XAD/PUF cartridge for separate analyses of wildland fuel source samples and prescribed burn samples to evaluate the semi-volatile nature of these polar compounds. For organic compound speciation, all of the samples were extracted by accelerated solvent extraction (Dionex ASE 300 Accelerated Solvent Extrator) with dichloromethane followed by acetone. These extracts were then combined and concentrated by rotary evaporation followed by a moisture filtered (ChromPack gas purifier) ultrahigh purity nitrogen blow down and then split into two fractions. A fraction was further evaporated to 100 µL under moisture filtered (ChromPack gas purifier) nitrogen and transferred to 300 µL of silanized glass inserts (National Scientific Company, Inc.). Samples were further evaporated to 50 µL, and then 25 µL of pyridine, 25 µL of acetonitrile, and 150 µL of BSTFA with 1% TMCS [N,O-bis-(trimethylsilyl) trifluoroacetamide with 1% trimethylchlorosilane (Pierce)] were added. These derivatizing reagents convert the polar compounds into their trimethylsilyl derivatives for analysis of organic acids, methoxyphenols, cholesterol, sitosterol, and levoglucosan. The deactivated glass insert containing the sample was put into a 2 mL vial and sealed. The sample was then placed into a thermal plate (custom-made) containing individual vial wells at 70 °C for 2 h. The calibration solutions were freshly prepared and derivatized just prior to the analysis of each sample set. All samples were analyzed by gas chromatography interfaced with mass spectrometry (GC/ MS) within 18 h to avoid degradation. The samples were analyzed by an electron impact ionization GC/MS technique using a Varian CP-3400 gas chromatograph with a model CP-8400 auto-sampler and interfaced to a Saturn 2000 ion trap mass spectrometer.
Results and Discussion Overview. Variations in combustion conditions and biomass fuel types are expected to alter the fractions of emission products; however, there has been a lack of replicated biomass source testing especially for wildland fuels to understand these relationships. In this study, sample collections of several fires were replicated, and additionally, four wildland fuels were selected for flat, heading uphill, and backing downhill combustion tests. Measurements of PM2.5 gravimetric mass and carbon fractionation for organic (OC) and elemental carbon (EC) by thermal optical reflectance are summarized in Table 1 and Figure 1. Variations in mass and carbon fractionation were observed among fuel types and within fuel types. Carbonaceous matter dominates PM2.5 mass and in some cases exceeds gravimetric mass measurements due to the positive and negative sampling artifacts. Sample collection for these two analyses was conducted using different filter types; quartz fiber filters and Teflon membrane filters were used for carbon and PM2.5 mass analyses, respectively. Quartz fiber filters are known to collect a portion of the semi-volatile compounds present in the gas phase, while particles on Teflon membrane filters are known to lose a portion of the semi-volatile compounds present in the particle phase (27, 28). Note that the dominant artifact was not quantitated for these samples. Overall, a large degree of variation was observed in the mass fractions of OC and EC, between fuels and within fuel types (Figure 1). The largest OC/EC ratio was observed in emissions from burning Californian eucalyptus wood and Alaskan tundra core biomass. Unfortunately, there was only one sample for each of these fuel types. Differences in PM2.5 mass and OC/EC ratio were found among flat, heading, and backing fires within the fuel types (Figure 1). The greatest variability within sets of four samples for the wildland fuels was found for Ponderosa pine needles (PPN) and Zambia grass (ZG). In all sets of four samples, we observed that the backing (i.e., downhill) fires burned with a slower spread and a narrower flame zone (2-5 cm), while the heading (i.e., uphill) fires spread faster to cover the entire fuel bed with a somewhat higher flaming intensity. The narrower flame zone may explain the higher EC mass fraction emitted in backing fires. Higher mass fractions of OC in heading fires were observed for only PPN and sagebrush (SAG). The heading fire of White pine needles (WPN H) emitted 2-3 times more PM2.5 gravimetric mass with a considerably higher OC/EC ratio than either flat or backing downhill fires (Table 1). The combustion of Zambia grass yielded unusual results. As compared to the other Zambia grass burns, ZG F1 had very different characteristics including 2-3 times higher concentrations of PM2.5 mass and organic carbon and 6-10 times higher concentrations of levoglucosan (Table 1). Except for the ambient collection of in-field prescribed burning emission samples, carbonaceous compounds were found to constitute 30-190% of the PM2.5 gravimetric mass (Figure 1), where fractions larger than 100% are due to sampling artifacts noted previously. Combustion differences within and among source types may account for variable emissions of SVOC partially absorbed by quartz filters and desorbed from Teflon filters (27, 28), resulting in higher concentrations of total particulate carbon than PM2.5 gravimetric mass. Several samples had higher concentrations of TC than PM2.5 gravimetric mass, most notably: ZG H, ZG B, eucalyptus, cedar 2, and rice straw. Emission samples from prescribed burning and from almond and wheat straw combustion had much lower mass fractions of particulate carbon. These low mass fractions of TC indicate significant contributions from elements (Ca, K, and Cl) and inorganic ions, which are not presented here but can be found in Chow et al. (29) and Chen et al. (21). VOL. 41, NO. 7, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2117
FIGURE 1. Organic and elemental carbon by TOR analysis expressed as a percent weight fraction of PM2.5 gravimetric mass (a) and abundance of levoglucosan normalized by PM2.5 gravimetric mass (b).
FIGURE 2. Identified semi-volatile polar organic compounds normalized by PM2.5 gravimetric mass.
A comprehensive study of several individual polar organic compounds (POC) was conducted to examine compositional similarities and differences between biomass types and combustion conditions. Of these individual compounds, levoglucosan and methoxylated phenols were found to be the dominant species in most source samples. Other abundant compounds included organic acids such as benzoic 2118
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 7, 2007
acids, resin acids, and methoxy acids. The total POC identified for source types ranged from 5 to 80% of the PM2.5 gravimetric mass (Figure 2). Five to 20% of the PM2.5 gravimetric mass was identified for prescribed burns (NLT, YOSE, and UNRR). Twenty to 80% of the PM2.5 gravimetric mass was identified for the combustion of White pine needles, grasses (tundra and MT grass), straws (wheat and rice), and mixed woods.
FIGURE 3. PM2.5 gravimetric mass concentration vs levoglucosan concentration. Combustion of grasses, straws, and mixed woods tended more toward smoldering and emitted higher mass fractions of semi-volatile methoxy phenols. Levoglucosan. Extraction recoveries of levoglucosan-13C6 were very good on average for wildland fuels and prescribed burn samples. The median recovery from 44 samples (including 20 ambient samples) was 88%. Recovery estimates were calculated from a ratio of levoglucosan-13C6 peak areas of samples to standard calibration checks. Previously, 69% levoglucosan extraction recoveries were reported by Simpson et al. (30). An average recovery of 84% was calculated for the other internal standards. The variation of levoglucosan within fuel types and combustion conditions is shown in Figure 1. We observed substantial variation in levoglucosan emissions from the different biomass fuel types. In some cases, we also noted substantial variation in levoglucosan emissions from different burns with the same fuel type, based on a small number of replicate experiments. Variability was the most substantial for PPN, ZG, and SAG. Comparisons between the groups of flat burns (n ) 2) and all burns within a group (n ) 4) suggest that the simulated hill slope tests did not greatly increase variability of levoglucosan concentrations. However, it should be noted that the combustion conditions between individual tests were difficult to replicate as seen in particulate emissions from the flat burn test replicates of several fuels. It is interesting to note that the high variation in levoglucosan mass fractions does not appear to correspond to variations in the carbon mass fractions. In an attempt to understand the variability of levoglucosan concentrations, relationships between it and bulk measurements were evaluated. We examined PM2.5 gravimetric mass concentrations versus levoglucosan concentrations (Figure 3). The three types of biomass samples are noted differently to check if the combustion scenario is responsible for the levoglucosan
variation. In this plot, it is difficult to draw conclusions about combustion fuel types (e.g., wildland fuels, residential woods, and prescribed burns) or burning scenarios (e.g., open, fireplace, or laboratory controlled). By comparison of levoglucosan concentrations to OC concentrations, we have found two fairly distinct relationships (Figure S2). One of the relationships is for the wildland fire and in-field prescribed burning emissions, and the other one is for the residential wood combustion (RWC) (Figure S2). Again, it is difficult to draw conclusions in this case since the SVOC absorption and desorption artifacts were not quantitated and may differ significantly between winter residential wood combustion and summer prescribed burning. Overall, the plots are similar; however, the increased mass concentrations measured in the in-field prescribed burning samples and lower mass concentrations in the residential wood burning introduce more scatter to the overall relationship(s). Levoglucosan has been widely reported to be quite abundant as compared to other individual organic compounds in biomass smoke, but a comparison of the reported concentrations (13-15, 31-33) suggests that the emissions are highly variable. The data shown in Table S1 were adapted from various emission rates presented in individual papers and normalized by the TC emission rate. Since carbon fractionation protocols of different carbon methods split OC and EC differently, it is difficult to compare data across studies without normalization to TC. A few studies included in Table S1 reported levoglucosan emission rates without derivatization (13, 14, 31, 33). Trimethylsilylation has been reported to significantly improve detection and quantitation of levoglucosan (20, 30). The overall range of levoglucosan was from 3 to 49% TC. Since the analytical differences are likely small, the majority of this variability may be explained by differences of fuel types and combustion differences that are not well-understood. Three separate studies of fireplace VOL. 41, NO. 7, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2119
FIGURE 4. Guaiacols and syringols normalized by PM2.5 gravimetric mass. combustion of oak wood reported levoglucosan abundances of 4.8, 13, and 23% TC. In our study, levoglucosan from fireplace oak wood tests comprised 4% TC (or 2.3% PM2.5 mass), which is in good agreement with Hays et al. (15). The percent of levoglucosan from fireplace tests of eucalyptus wood, common in Southern California, was much lower in our study, 5% TC (or 2.9% PM2.5 mass) as compared to Schauer (14), 49%, and Nolte (32), 38%. Sheesley and colleagues (33) reported levoglucosan to represent 3% of the TC in rice straw combustion emissions, which is the same as reported in our study, but in another study (34), levoglucosan was reported as 12% TC. Variations of the levoglucosan mass fraction are potentially explained by the differences in combustion factors. Cellulose pyrolysis is a series of complex, parallel, and consecutive chemical reactions (35); thus, the yields of pyrolysis byproducts are subject to variability. In addition to temperature, another variable that has a strong influence on the yield of levoglucosan from cellulose is the presence of inorganic ions. Cationic effects were studied by Dobele and co-workers (36). They report a substantial increase in the yield of levoglucosan after removal of metal cations, such as potassium and calcium from wood (36). The effect of potassium on levoglucosan yields is particularly concerning in the context of source apportionment models since both components are considered to be markers of biomass smoke. Methoxylated Phenols. Guaiacols and syringols, arising from pyrolysis of wood lignin, are commonly found in biomass combustion emissions. In the lignin of hardwoods, structural units consist of both guaiacyl and syringyl types in similar proportions, but in softwoods, guaiacyl units are the predominant structural unit (8). Thus, during pyrolysis of coniferous lignin, syringols are generally not formed, but during pyrolysis of deciduous lignin, guaiacols and syringols are formed. Many researchers have included these compounds in source characterization analyses of residential woods (8, 13-15, 31). Results shown in Figure 4 indicate that in addition to hardwoods, sagebrush and grasses emit both guaiacols and syringols, but pine needles (especially White pine needles) have a high mass fraction of guaiacols (2-5% PM2.5 mass) with very little syringols, similar to softwoods. The prescribed burn samples collected in mixed coniferous forests (YOSE and NLT) also had high mass fractions of guaiacols (1-5% PM2.5 mass) and low mass fractions of syringols (
