Secondary particle formation and evidence of heterogeneous chemistry during a wood smoke episode in Texas

Authors


Abstract

[1] To evaluate the impact of regional wildfires in east Texas on fine particulate matter concentration and composition, source apportionment calculations were performed on a set of samples collected at three sites in Houston during a wood smoke episode. Separately, samples collected at the same sites on days not influenced by wood smoke, were analyzed for comparison. The analysis of the data collected on non–wood smoke episode days indicated that the major contributors to PM2.5 mass were secondary sulfate, diesel and gasoline powered vehicles, other organics (difference between the measured organic mass by carbon (OMC) and the sum of the primary OC source contributions), wood combustion and meat cooking. Secondary sulfate (not apportioned to a primary source) represented almost 100% of the sulfate measured whereas primary sources were found to account for the majority of the measured organic carbon at three sites (64–69%). On the wood smoke episode days, major sources of PM2.5 were found to be the same as on the days without wood smoke, except for the contribution of the meat cooking source, which became insignificant on the wood smoke episode days. The contribution of the wood combustion source increased by an average of 200% at all sites on wood smoke episode days, whereas the contributions of other primary sources did not increase significantly during the wood smoke episode. PM2.5 mass almost doubled during the wood smoke episode, largely because of the contributions of the secondary sources. The mass concentrations of secondary sulfate and organics not attributed to primary sources increased 68% and 228% at each site, respectively, during the wood smoke episode. The increase in the contributions of secondary sulfate aerosols during the wood smoke episode was examined using a 3-D photochemical grid model. The simulations, together with ambient data, indicated that the increases in sulfate concentrations observed during this wood smoke episode were consistent with heterogeneous/surface reactions on wood smoke particles.

1. Introduction

[2] Biomass burning, including prescribed burning and forest wildfires, leads to substantial atmospheric emissions of particulate matter and other pollutants. These emissions may significantly impact air quality over length scales of hundreds to thousands of kilometers during periods of intense fire activity [Crutzen and Andreae, 1990]. A variety of methods have been used to quantify the contributions of biomass burning emissions to ambient air quality, including measuring the chemical composition of atmospheric particles, using back trajectory analysis [Andreae, 1983], and determining the composition of trace gases [Blake et al., 1997].

[3] In this work, the air quality impacts of fires in Texas will be examined. Dennis et al. [2002] and Junquera et al. [2005] have examined the overall magnitude of both prescribed burns and wildfires in Texas and have demonstrated that emissions from fires regularly have a significant impact on regional air quality. Fraser and Lakshmanan [2002] quantified the molecular marker in particle phase, levoglucosan, of biomass combustion during a regional haze episode in Texas to track the transport of biomass combustion aerosols, and demonstrated that air quality impacts can be substantial during extreme fire events.

[4] To further examine the role of fires on air quality in Texas, the impact of wildfires on fine particulate matter concentration and composition during an August-September 2000 episode will be evaluated. Yue and Fraser [2004b] have examined sources of organic particulate matter during this fire episode by quantifying polar organic compounds in samples collected at three sites in Houston. Significantly higher concentrations of levoglucosan and other high molecular weight acids were observed on days when fires occurred, indicating that wood combustion was a major source of particulate matter generally, and polar compounds specifically, in this episode. To study the significance of fires as a source of particulate matter during this period, this paper will report source apportionment calculations for three sites in Houston. Source contributions will be reported for days when particulate matter concentrations were greatly influenced by forest fires, and separately on other days not influenced by wood smoke [Junquera et al., 2005]. The Chemical mass balance (CMB) approach will be used to estimate the contributions of primary and secondary sources to PM2.5 levels. These results will indicate that, in addition to being a significant source of carbonaceous aerosol, fires can lead to significant enhancements of particulate sulfate concentrations, and that this excess sulfate can be accounted for if heterogeneous sulfate formation reactions are occurring on the wood smoke surfaces. Regional photochemical modeling of days with significant concentrations of wood smoke particles will be performed to assess the impact of heterogeneous reactions on sulfate concentrations.

2. Sampling and Analysis

2.1. Wood Smoke Episode

[5] In late August and early September 2000, Texas experienced hot, dry weather for several days creating conditions conducive to wildfires across the state. Multiple wildfires occurred and smoke from the wildfires reduced visibility and impaired air quality in the Houston area. This period will be referred to as the wood smoke episode. This episode coincided with a major air quality field study, the Texas Air Quality Study (TexAQS 2000). Consequently, for this period, extensive modeling data sets and observational data are available.

[6] An inventory of the emissions associated with wildfires during the wood smoke episode have been assembled by Junquera et al. [2005] using procedures first described by Dennis et al. [2002]. The period of most intense fire activity was from 30 August to 8 September, with the highest fire emissions occurring on 4 and 6 September, as shown in Figure 1. In contrast, relatively little fire activity occurred from the beginning of August to 29 August in the Houston area.

Figure 1.

(a) Acres burned. (b) Estimates of emissions of PM2.5 from fires in southeast Texas during August and September 2000 [Junquera et al., 2005].

[7] These wildfires led to significant increases in ambient fine particulate matter concentrations. 24-hour averaged PM2.5 mass concentrations for the day with peak fire activity (6 September 2000) averaged 39 μg/m3 at multiple sampling sites throughout Houston, as compared to an average of 10 μg/m3 on the days (25–28 August 2000) immediately preceding the fires [Texas Commission on Environmental Quality (TCEQ), 2005; Junquera et al., 2005]. For sulfate, the 24-hour averaged mass concentrations were 13.5 and 3.8 μg/m3, on 6 September and the days preceding the fire events, respectively.

2.2. Ambient Sampling

[8] In order to resolve the sources of fine particulate matter on days with low and intense fire activity, samples of atmospheric fine particles were collected at three sites in Houston as described in the work of Yue and Fraser [2004a, 2004b]. The three sampling locations are shown in Figure 2 and included one near the coast (La Porte), one in an industrial location (HRM-3) and one in a suburban neighborhood (Aldine). In La Porte, the sampling site was located on the grounds of a municipal airport located at the edge of the small residential community of La Porte. Industrial sources bordered this residential community on the north and east. The sampling site HRM-3 was located directly adjacent to the highly industrialized Houston Ship Channel. Finally, the sampling site at the suburban Aldine site was located on the grounds of an elementary school, approximately 20 kilometers from the industrial source region.

Figure 2.

Location of the three sites in Houston.

[9] Samples were collected every other day between 15 August and 30 September 2000, and the samples collected on 15, 21, 27 August and 2, 6, 8, 14, 20, and 30 September were analyzed for organic speciation.

[10] The organic speciation of fine PM is described in detail in the work of Yue and Fraser [2004a, 2004b], and will only be summarized here. Samples of fine particles were collected by a high-volume air sampler (TSP Volume Controlled High-Volume Air Sampler; Thermo Andersen, Smyrna, Georgia) attached to a fine particles inlet (High-Volume Virtual Impactor; MSP, Minneapolis, Minnesota) at a flow rate of 1.1 m3 min–1 for 24 hours. The filters were then stored in prebaked glass jars and frozen until analysis. Organic particulate matter was extracted with a suite of solvents under ultrasonic agitation and the extract was reduced in volume with rotary evaporation and nitrogen blowdown. After the extraction, the samples were analyzed using a Hewlett-Packard 6890 GC and a HP 5973 MS detector using authentic standards to calibrate instrument response [Yue and Fraser, 2004a, 2004b].

3. Methods

3.1. Receptor Modeling

[11] The impact of forest fires on fine particulate matter concentrations and composition was examined using the Chemical Mass Balance model (CMB) [Watson et al., 1998] to calculate the contributions of the important sources to total PM2.5 mass at the receptors.

[12] The CMB model estimates source contributions using the chemical composition of primary emissions and the concentrations of these same constituents in the ambient atmosphere. Thus CMB requires both accurate speciation of pollutants in ambient air and representative source profiles.

[13] In the CMB model, the concentration (cik) of chemical species i at receptor k, is given by

display math

where Cik is the concentration of species i at the receptor k, aij is the concentration of species i in the emissions from source j, sjk is the mass contribution of source j to receptor k, and eik is the difference between the measured and calculated species concentration [Watson et al., 1990]. The CMB model estimates the contributions from chemically distinct source types by using an effective variance weighted least squares solution to equation (1) to solve for the unknown source contributions, (sjk). In the present work, equation (1) was solved using the chemical mass balance model CMB 8.2, which is described in detail by Watson et al. [1998].

[14] Several considerations must be taken into account in constructing source profiles for use in CMB modeling. First, the analytical procedure used in the analysis of ambient samples should be equivalent to that used in the analysis of source samples to prevent procedural biases in the analysis. Also, the chemical composition of source profiles used in CMB model must be significantly different from each other so that they do not lead to collinearity between sources. Profiles of the chemical and molecular composition of emissions from fine particle sources used in this work were taken from the results of different studies, all of which used analytical procedures equivalent to those used in the analysis of the ambient samples [Yue and Fraser, 2004a, 2004b].

[15] The composition of fine particulate matter emitted from catalyst-equipped gasoline-powered vehicles obtained from Schauer et al. [2002] was used as the representative source profile for gasoline vehicles. The chemical and molecular composition of emissions from two heavy-duty tractor trailers operated on the Urban Dynomometer Driving Schedule for Heavy-Duty Vehicle cycle was used as the diesel-powered vehicle source profile [Fraser et al., 2002].

[16] Dennis et al. [2002] found that wildfires represent a significant fraction of the emission inventory of southeast Texas. Wiedinmyer et al. [2001] determined that pine and hardwood oaks cover a majority of land use areas in Texas. Hence pine and oak wood smoke profiles were adopted from the work of Schauer et al. [2001] as representatives of wood combustion source category. Since pine and oak wood smoke sources are very similar to each other, using two profiles at the same time caused collinearity problems. Therefore model performance was tested using oak and pine profiles separately. Since the use of the pine smoke profile improved the model performance, and since the fires that occurred during the episode under study were primarily in pine forests [Junquera et al., 2005], the pine smoke profile was used in this work.

[17] Meat cooking emission profiles were taken from the results of the source tests reported by Schauer et al. [1999]. In that work, significant quantities of n-alkanoic and n-alkenoic acids were measured in particles generated in meat cooking operations. Therefore these fatty acids were used in this study as markers of meat cooking emissions.

[18] There is strong evidence that plant leaf abrasion contributes an unspecified amount of fine particulate leaf wax particles to the atmosphere [Mazurek et al., 1991]. Leaves act as a sink for anthropogenic and natural airborne material, which under suitable conditions are resuspended into the atmosphere. Odd carbon number n-alkanes as a group ranging from C27 to C33 can serve as a marker that could be used to trace fine particulate vegetative detritus released to the urban atmosphere [Rogge et al., 1993a]. Organic compound speciation of PM2.5 analyzed by Rogge et al. [1993a], and the chemical speciation from Hildeman et al. [1991] were combined as the source profile for vegetative wax particles.

[19] The composition of samples of paved road dust collected and analyzed by Hildeman et al. [1991] and the molecular composition of the same samples reported by Rogge et al. [1993b] were used as the source profile of road dusts.

3.2. Photochemical Modeling

[20] As described in section 4, source resolution analyses of fine particulate matter during the wood smoke episode led to the conclusion that the formation of secondary sulfate was significantly enhanced by the presence of the wood smoke. Regional photochemical models were used to investigate processes that could lead to this enhanced secondary sulfate. Regional photochemical models simulate emission, chemical transformation, horizontal advection and diffusion, vertical transport and diffusion, dry deposition, and wet deposition of species in the atmosphere. Although any comparable photochemical grid model could be used, the Comprehensive Air Quality Model with extensions (CAMx) [ENVIRON International, 2005] was selected for this study because it is currently being used by the State of Texas for attainment demonstrations in areas that have violated the National Ambient Air Quality Standards for ozone. Several model performance evaluations have been performed on the photochemical model formulations used in this work [ENVIRON International, 2003; Zhang et al., 2003].

[21] The modeling domain was a nested regional/urban scale 36-km/12-km/4-km grid, shown in Figure 3. The episode period was 22 August to 6 September 2000. Meteorological inputs required by the model were based on results from the Mesoscale Meteorological Model, version 5, MM5 [TCEQ, 2004c]. The volatile organic compound (VOC) and NOx emission inventories used as input for the modeling episode were prepared by the Texas Commission on Environmental Quality (TCEQ) in accordance with U.S. EPA guidance. A MOBILE6-based inventory was developed for on-road mobile source emissions; emissions for nonroad mobile and area sources were developed using emission factors and the U.S. EPA's NONROAD model, using local activity data when available. Biogenic emission inventories were estimated using the GLOBEIS emission model with locally developed landcover data [TCEQ, 2004a]. Point source emissions were developed through a special inventory survey and were also estimated on the basis of ambient data collected in the source region. Details of the VOC and NOx emission inventory development are available at TCEQ [2004b].

Figure 3.

Modeling domain used in the study: The Regional, East Texas and Houston-Galveston-Beaumont-Port Arthur nested domains had 36, 12 and 4 km horizontal resolution, respectively.

[22] For the work described in this paper, emissions of SO2 from all sources and particulate matter emissions from fires were added to the model inputs. The major sources of SO2 were point sources. Data for point source SO2 emissions in Texas for the year 2000 were obtained from the TCEQ point source database [TCEQ, 2004a]. For regions outside of Texas, SO2 emissions were obtained from the US EPA 1999 National Emission Inventory (NEI99) (available at http://www.epa.gov/ttn/chief/net/1999inventory.html#final3crit, accessed on 15 March 2005). Annual emissions of some facilities in the NEI99 database were updated with Louisiana Department of Environmental Quality's emission inventory data set for the year 2000 (available at http://www.deq.louisiana.gov/portal/Portals/0/evaluation/eis/2000.xls, accessed on 15 March 2005). The spatial distribution of these SO2 point emissions in Texas and Louisiana is shown in Figure 4a. SO2 emissions from on-road and nonroad mobile sources were also retrieved from the NEI99 database, and were adjusted to year 2000 values using growth factors.

Figure 4.

(a) SO2 point source emissions (tons/yr) in Texas and Louisiana. Fire locations on (b) 6 September and (c) 8 September, based on the emission inventory of Junquera et al. [2005].

[23] Emissions from fires were based on the work of Junquera et al. [2005] and the locations of the fires on 6 September 2000 are shown in Figure 4b. The original work of Junquera et al. [2005] did not include SO2 emissions from fires. These emissions were added, on the basis of an emission factor reported by Reddy and Venkataraman [2002]. The extent of SO2 emissions from fires (Figure 4b) was approximately 1% of the point source SO2 emissions on days with high fire activity.

4. Results and Discussion

4.1. Measurements of PM2.5 and Source Attribution Using Organic Molecular Markers

[24] The filter data were separated in two sets for analysis. In the first case, the concentrations of organic compounds in PM2.5 samples collected on 15, 21, and 27 August and 2, 14, 20, and 30 September were averaged together and used in CMB calculations [Yue and Fraser, 2004a, 2004b] as days not affected by the wood smoke period. These data will be labeled non–wood smoke period. As a second case, in order to investigate the influence of the regional haze episode attributed to forest fires in eastern Texas on particulate matter concentrations in Houston [Junquera et al., 2005], the contribution of the same sources was calculated for a set of the samples that were collected on two days (6 and 8 September) when the fire activity was most intense. Note that 2 September was included in the averaging for the days with low contributions for fires because the ambient sites used in this work were not impacted by fires on this date. The fires on 2 September were east of the sites and winds were out of the west, advecting the fire plumes away from the sampling sites.

[25] The PM2.5 mass concentrations and the concentrations of major chemical components of PM2.5 are given in Table 1 for both the non-wood smoke period and the wood smoke episode.

Table 1. PM2.5 Mass and Chemical Composition During the Observation Perioda
ComponentAldineHRM-3La Porte
Non–Wood Smoke PeriodWood Smoke EpisodeNon–Wood Smoke PeriodWood Smoke EpisodeNon–Wood Smoke PeriodWood Smoke Episode
  • a

    Data are divided into wood smoke period (filters collected on 6 and 8 September) and non–wood smoke period (all other filters analyzed). All concentrations are in micrograms per cubic meter.

  • b

    Unscaled value of organic carbon mass.

  • c

    The mass of secondary sulfate is scaled to represent ammonium sulfate.

PM2.5 mass15.028.517.025.311.026.8
Organic carbonb4.710.14.28.02.78.1
Elemental carbon0.50.60.60.60.30.5
Nitrate0.30.30.40.30.20.2
Sulfatec5.38.85.49.55.38.7
Ammonium1.22.91.23.11.22.9
Crustal material0.50.31.50.30.30.3

4.1.1. Source Apportionment for the Days Not Influenced by Wood Smoke Episode

[26] A total of 23 organic compounds plus 3 inorganic constituents, as shown in Figure 5, were selected to be used in chemical mass balance calculations. An important consideration in the selection of marker compounds is that the species be chemically stable (species that cannot be significantly depleted nor formed by chemical reactions during the transport from sources to receptors), and that the species adequately span all source categories to be isolated [Schauer et al., 1996]. Even though oleic acid is an olefinic compound susceptible to ozone attack, this compound is a key tracer for aerosols from meat cooking and thus should be included in the analysis [Schauer et al., 1996].

Figure 5.

Comparison of source allocation model calculations of molecular marker species concentrations to measured ambient concentrations for the days not affected by wood smoke.

[27] The precision of mass concentration of each species should also be included in the model calculations. For the individual organic compounds analyzed using GC-MS, 20% of the concentrations of the compounds were assigned as the precision values based on the precision of the measurements determined experimentally [Yue and Fraser, 2004a, 2004b]. Bulk chemical composition of particulate matter was determined using X-ray fluorescence (XRF) and thermal-optical transmission (TOT) by Research Triangle Institute (RTI) in compliance with established EPA protocols [U.S. Environmental Protection Agency, 1999]. The method detection limits of the three inorganic constituents involved in the chemical mass balance calculations were used for their precision values.

[28] A comparison of the calculated and measured ambient concentrations of each marker species at each site is given in Figure 5. In general, reconstructed molecular marker concentrations agreed with measured concentrations over a span of 3 orders of magnitude. Since the source contribution estimates calculated by CMB are based on least squares linear regression, they are not unique and therefore several performance measures and statistics such as t-statistics, R2 and chi-square values, are needed to help evaluate the accuracy of CMB source contribution estimates. All t-statistics values were greater than 2, which is a valid range as stated by Watson et al. [1990]. R2 values were in the range of 0.67–0.73 and the corresponding chi-square values were less than 6.5, all of which indicate a good fit between reconstructed compound concentrations and measured values.

[29] Fine particulate matter mass contributions on non-wood smoke days from six primary sources including gasoline-powered vehicle exhaust, diesel-powered vehicle exhaust, meat cooking, vegetative detritus, wood combustion and road dust, were calculated as well as other organics and secondary sulfate concentrations (Table 2). Also, OC attributed to each individual primary source is listed separately in Table 2. The road dust contribution was not statistically different than zero at Aldine and La Porte; therefore this source was excluded from the model. Although road dust had a small contribution to fine PM mass at HRM-3, it was statistically different from zero, and thus was included in the model calculation for this site. The presence of the road dust source in the calculations of source contribution estimates at HRM-3 may be a result of unpaved road surfaces in industrial facilities located near the Houston Ship Channel.

Table 2. Source Contributions at Three Sites to Ambient PM2.5 Mass and Organic Carbon (OC) Concentrations for the Days Not Affected by Wood Smokea
Source CategoriesAldineHRM-3La Porte
PM2.5 MassPM2.5 OCPM2.5 MassPM2.5 OCPM2.5 MassPM2.5 OC
  • a

    Organic carbon masses have not been scaled to represent mass of organic compounds. Unit is micrograms per cubic meter.

Gasoline vehicles2.51 ± 0.941.10 ± 0.412.02 ± 0.670.89 ± 0.291.00 ± 0.400.44 ± 0.18
Diesel vehicles3.16 ± 0.441.01 ± 0.143.77 ± 0.451.21 ± 0.142.19 ± 0.280.70 ± 0.09
Vegetative detritus0.59 ± 0.090.19 ± 0.030.32 ± 0.050.10 ± 0.020.31 ± 0.040.10 ± 0.01
Meat cooking1.14 ± 0.260.39 ± 0.090.52 ± 0.140.18 ± 0.050.71 ± 0.160.24 ± 0.05
Wood combustion0.74 ± 0.180.41 ± 0.100.84 ± 0.150.47 ± 0.080.44 ± 0.100.25 ± 0.06
Road dustn.i.n.i.0.17 ± 0.040.05 ± 0.01n.i.n.i.
Sum of apportioned PM2.5 mass8.14 ± 1.09 7.64 ± 0.83 4.65 ± 0.52 
Sum of apportioned OC 3.11 ± 0.46 2.90 ± 0.34 1.73 ± 0.21
Measured OC 6.54 5.88 3.77
Other organics 3.43 2.98 2.04
Secondary sulfate5.28 5.40 5.30 
Sum of PM2.5 mass apportioned to primary sources, other organics and secondary sulfate16.85 16.02 11.99 
Measured PM2.5 mass15.00 16.99 11.00 

[30] The apportioned mass due to the primary sources of PM2.5 given in Table 2 (non–wood smoke days) represents 54% of measured PM2.5 mass at Aldine, 45% of PM2.5 at HRM-3, and 42% of PM2.5 at La Porte. The primary sources contributing to PM2.5 mass at each site include diesel-powered vehicles with 21%, 22%, and 20%, gasoline vehicles with 17%, 12%, and 9%, wood combustion with 5%, 5% and 4%, and meat cooking with 8%, 3%, and 6% of apportioned PM2.5 mass at Aldine, HRM-3 and La Porte, respectively. The contribution of diesel-powered vehicles was found to be considerably higher than gasoline vehicles at each site. The higher contribution of the meat cooking source at Aldine is reasonable when considering that this site is located near a residential area where meat cooking operations are expected to be more frequent than they are at the other nonresidential sites. The measurements in the work of Yue and Fraser [2004b] also showed that the levels of octadecenoic acid (C18:1), which was found in the emissions from meat cooking processes, were the highest at Aldine. Vegetative detritus was a minor contributor to PM2.5 mass at all locations.

[31] The concentrations of organic carbon and sulfate attributable to primary sources were calculated by multiplying the primary source contribution estimates by the coefficients determined from the chemical composition profile of each source. The OC contribution from individual sources is listed separately in Table 2. At all three sites, less than 1% of sulfate was attributed to the primary sources, so the sulfate contribution is not listed separately. The OC and sulfate measured in the PM2.5 samples in excess of the primary contributions was attributed to secondary sources. Since secondary ammonium sulfate aerosol is a major source of fine PM in the southeast Texas region, the excess sulfate is labeled as “secondary sulfate.” With bulk compositional analysis showing sufficient PM2.5 ammonium to neutralize sulfate, with a molar ammonium to sulfate ratio averaging 1.7 ± 0.34 [Russell et al., 2004], the secondary formation of sulfate will likely be in the form of ammonium sulfate, and the mass of secondary sulfate was scaled to represent ammonium sulfate.

[32] For organic carbon during the non–wood smoke period, primary sources have been found to account for the majority of the measured fine particle OC concentrations at each site. Secondary organic aerosol (SOA) is not included in the OC apportionment calculations. The sum of OC from the source contributions determined by CMB represents 66%, 69% and 64% of the measured OC concentration in Aldine, HRM-3 and La Porte, respectively. Five sources have been identified that contribute to the organic carbon in PM2.5 at Aldine and La Porte and six sources at HRM-3 (Table 2). According to the source apportionment model results, the major sources contributing to the organic carbon in PM2.5 are gasoline vehicles (16–23%), diesel vehicles (21–29%), wood combustion (9–11%), and meat cooking (4–9%). Contributions of gasoline and diesel vehicles to organic carbon mass in Table 2 are roughly equivalent at Aldine, while the contribution of diesel-powered vehicles is considerably higher than gasoline vehicles at HRM-3 and La Porte.

[33] Excess organic carbon mass is the difference between the measured organic mass by carbon (OMC) and the sum of the contributions of the primary OC sources identified by the mass balance. OMC is usually calculated as 1.4 times the OC mass [Watson et al., 1988]. This excess OC mass may originate either from secondary organic aerosol formation or from sources not included in the model calculations, therefore this excess mass will be called “other organics” in the rest of this work.

[34] The sum of other organics and secondary ammonium sulfate concentrations plus the apportioned mass concentrations represented 94% of PM2.5 mass at HRM-3 and was overestimated by 8% at La Porte and 11% at Aldine. The lower percentage of the PM2.5 mass at HRM-3 that could be explained by the mass balance is probably because HRM-3 is located in a highly industrialized area, and there may be other sources such as refineries and petrochemical plants contributing to fine particle levels that were not included in the mass balance model. In addition to the important primary sources of PM2.5, secondary sources were important contributors to fine particle mass with secondary sulfate contributing 31%, 31%, and 49%, and other organics contributing 20%, 18%, and 19% of the fine particle mass at Aldine, HRM-3 and La Porte, respectively.

4.1.2. Source Apportionment on Wood Smoke Episode Days

[35] The contribution of the same sources was calculated separately for the wood smoke episode when regional haze episode attributed to forest fires in eastern Texas influenced particulate matter concentrations in Houston [Junquera et al., 2005]. Table 3 shows the source contribution estimates at three sites on these two days and the contribution of each source to organic carbon levels. Calculations of secondary sulfate and other organics were done as described above. PM2.5 mass was dominated by the same sources as in the first case, including secondary sulfate, other organics, diesel vehicles, gasoline vehicles, wood combustion and vegetative detritus. The contribution of the road dust and meat cooking sources (as evidenced by the considerable decrease in octadecenoic acid levels during the episode) were not statistically different than zero at the three sites, and therefore these sources were not included in the model. The contribution of the wood combustion source to the PM2.5 mass was 200% (on average) higher at each site than the contributions in the previous case, as expected. This result was consistent with, and primarily driven by, the concentration measurements of levoglucosan, which showed an increase of 173% during the wood smoke episode at all three sites [Yue and Fraser, 2004a, 2004b]. The results for the two cases (non–wood smoke and wood smoke) were compared graphically in Figures 6 and 7with Figure 6 showing the contribution of primary sources, secondary sulfate and other organics to PM2.5 mass and Figure 7 showing the contribution of primary sources and other organics to PM2.5 OC mass.

Figure 6.

Comparison of the source contributions to PM2.5 mass in two cases.

Figure 7.

Comparison of the source contributions to OC in PM2.5 in two cases.

Table 3. Source Contributions at Three Sites to Ambient PM2.5 Mass and Organic Carbon (OC) Concentrations for the Days During the Wood Smoke Episodea
Source CategoriesAldineHRM-3La Porte
PM2.5 MassPM2.5 OCPM2.5 MassPM2.5 OCPM2.5 MassPM2.5 OC
  • a

    Organic carbon masses have not been scaled to represent mass of organic compounds. Unit is micrograms per cubic meter.

Gasoline vehicles1.68 ± 0.590.74 ± 0.263.18 ± 0.991.40 ± 0.431.56 ± 0.470.69 ± 0.18
Diesel vehicles3.48 ± 0.431.11 ± 0.143.63 ± 0.481.16 ± 0.152.91 ± 0.350.93 ± 0.09
Vegetative detritus0.72 ± 0.090.23 ± 0.030.67 ± 0.090.21 ± 0.030.54 ± 0.070.17 ± 0.01
Meat cookingn.i.n.i.0.05 ± 0.010.02 ± 0.00n.i.n.i.
Wood combustion2.02 ± 0.271.13 ± 0.152.32 ± 0.371.30 ± 0.211.54 ± 0.210.86 ± 0.12
Sum of apportioned PM2.5 mass7.90 ± 0.78 9.85 ± 1.16 6.55 ± 0.63 
Sum of apportioned OC 3.21 ± 0.33 4.09 ± 0.51 2.65 ± 0.26
Measured OC 14.17 11.15 11.38
Other organics 10.95 7.06 8.73
Secondary sulfate8.80 9.45 8.68 
Sum of PM2.5 mass apportioned to primary sources, other organics and secondary sulfate27.65 26.36 23.96 
Measured PM2.5 mass28.45 25.30 26.85 

[36] The apportionment of organic carbon in PM2.5 for the wood smoke case showed that the sum of the contributions of the primary sources to organic carbon concentrations is only slightly higher during the wood smoke episode (3–35% increase), therefore most of the increase in measured OC mass is attributable to the other organics (137–328% increase) for the wood smoke episode (Figure 6). From Figure 7, one can observe slight changes in the contributions of gasoline vehicles, diesel vehicles, vegetative detritus, whereas a significant increase can be observed in wood combustion source contribution to organic carbon (173%, 176% and 250% at Aldine, HRM-3, and La Porte, respectively) which lead to an increase in the total mass attributed to primary sources (Tables 2 and 3).

[37] Although the sum of the mass attributed to all primary sources during the wood smoke episode showed some increase due to the increase in the contribution of wood combustion, the dramatic increase in total PM2.5 mass during the wood smoke episode can be attributed to other organics (including secondary formation) and secondary ammonium sulfate (Figure 6). The contribution of other organics to total PM2.5 mass showed an increase of 137–328%, secondary sulfate increased 64–75% at each site in the wood smoke episode case. Possible reasons for the increase in the contribution of secondary sulfate and other organics are presented in section 4.2.

4.2. Secondary Aerosol Formation

[38] The source apportionment analysis performed on wood smoke episode days and non–wood smoke episode days shows that secondary sulfate and other organics increased significantly during the wood smoke episode. Sulfate contributions averaged 5.3 and 9.0 μg/m3 and contributions of other organics averaged 2.8 and 8.9 μg/m3 on non–wood smoke episode days and during wood smoke episode, respectively.

[39] The increased mass of other organics during the wood smoke episode could be due to inaccurate source profiles, increased partitioning of organic semivolatile species to the particle phase or heterogeneous reactions on the wood smoke particles producing additional condensed material. Inaccuracy in source profiles, specifically a low value of the OC/levoglucosan ratio used in the CMB calculations, may be a reason for the increase in other organics during the wood smoke episode. The OC/levoglucosan ratio used in this work was based on the data of Schauer et al. [2001]. Recent data have shown wide variability in this ratio depending on combustion conditions [Fine et al., 2002; Sanders et al., 2003; Shafizadeh et al., 2003] and the ratio may be influenced by atmospheric reactions of levoglucosan.

[40] A second potential cause of the increase in other organics during the wood smoke episode is increased partitioning of semivolatile species into the particle phase. The increased wood smoke concentration, of approximately 1.5 μg/m3, will cause some additional semivolatile species to partition into the particle phase. However, since primary OC concentrations in the wood smoke episode increase by a much smaller percentage than do the other organics concentrations, it seems unlikely that this phenomenon would explain all of the observed increase in other organics.

[41] Finally, heterogeneous reactions may explain the increase in other organics during the wood smoke episode. Jang et al. [2003] found that indigenous sulfuric acid produced from combustion of fossil fuels (e.g., wood smoke) could initiate the acid-catalyzed heterogeneous reactions on the particle phase. Jang et al. [2002] observed these heterogeneous reactions for a system involving ozone reaction with α-pinene, which Russell and Allen [2005] found was the dominant secondary organic aerosol precursor in southeast Texas. This pathway will be examined in more detail in future work. The remainder of this paper will examine pathways for the formation of additional sulfate during wood smoke episodes.

[42] Possible reasons for the high concentrations of sulfate on wood smoke episode days are direct emission of sulfate from the forest fires, displacement of chloride in the fire emissions by sulfate, and secondary aerosol formation through chemical reactions, possibly heterogeneous reactions. Source characterization for wood combustion have shown that sulfate is present in only trace levels from wood combustion [Hays et al., 2002; Schauer et al., 2001; Cachier et al., 1996; Robinson et al., 2004]. As a result, the source apportionment calculations show that direct primary emissions of sulfate from wood combustion did not play a role in creating high sulfate concentrations in ambient particulate matter. Particulate matter from fires is emitted at high temperatures, and at these temperatures, many inorganic elements such as KCl volatize and condense back onto particle surfaces. Chloride displacement occurs when SO2 and the atmospheric O2 react with KCl forming particulate phase K2SO4. Some chloride displacement may occur, but typical chloride concentrations in oak and pine wood smoke (up to a few percent [Hays et al., 2002; Schauer et al., 2001; Cachier et al., 1996; Robinson et al., 2004]) are not sufficient to account for the observed sulfate enhancements; so direct emissions from fires matter containing chloride, followed by chloride displacement, are also not the cause of the high sulfate concentrations observed during the fire events.

[43] If direct emissions of sulfate and chloride displacement do not explain the entire observed sulfate, then it is likely that chemical reactions transforming SO2 to sulfate are the dominant source. Chemical pathways for SO2 oxidation to sulfate include (1) the homogeneous oxidation of SO2 by OH• in the gas phase, (2) the condensed-phase reactions of SO2 with active oxidants such as peroxides and ozone, and (3) the heterogeneous reactions of SO2 on nonaqueous carbonaceous particles.

[44] 1. In the gas phase, SO2 reacts with OH radical to form H2SO4 and H2SO4 condenses on available particles. This homogeneous reaction can produce very high sulfate concentrations if the atmosphere has significant SO2 and OH• mixing ratios. At the La Porte site, 10-min samples shown in Figure 8 indicate elevated SO2 concentrations on days with and without intense fire activity. In addition, as shown in Figure 9, OH• measurements imply persistent diurnal distributions throughout the measurement period, with no evidence of elevated concentrations during the wood smoke episode. These data suggest that SO2 and OH• concentrations were not particularly elevated during days of intense fire activity, so observational data indicate that gas phase reactions do not explain the enhanced sulfate formation during the wood smoke episode; regional photochemical modeling calculations reported later in this section also confirm this.

Figure 8.

Ten-minute sulfur dioxide concentrations at La Porte site during the Texas Air Quality Study, in a unit of ppbv.

Figure 9.

Free radical concentrations measured at the La Porte site during the Texas Air Quality Study. A number of days exhibit high HO2 concentrations at night (adapted from Martinez et al. [2001]).

[45] 2. Hydration of wood smoke particles leads to increased aqueous phase volumes and aqueous phase reactions. Wood smoke particles are initially emitted as dry and relatively hydrophobic materials; these hydrophobic particles can then undergo atmospheric reactions, making them hydrophilic and able to take up water. The volume of water taken up by the particles will depend not only on the particle mass available, but also on the size distribution of the particles. An upper bound estimate of the volume of aqueous phase condensed onto wood smoke particles (roughly 50 μg/m3) was made by applying a growth factor of 1.4 [Gasparini et al., 2004] to particle size distributions measured on 6 September at the La Porte site. This volume of water together with upper bound estimates of the aqueous concentration of SO2 (based on the maximum observed gas phase SO2 concentration of 66 ppbv) and the aqueous oxidation rate (500% h−1 [Seinfeld and Pandis, 1998]) lead to a sulfate formation rate of only 2 × 10−13 g m−3h−1. This upper bound suggests that the increase in the volume of aqueous phase in the atmosphere due to wood smoke, with no new chemistry due to the wood smoke, is not likely to explain the high sulfate concentrations during the wood smoke episode.

[46] 3. Thus only gas-particle-phase surface reactions are left as a possible explanation for the sulfate formation observed in the wood smoke episode. Conversion of SO2 gas into sulfate aerosol has been observed on the surface of carbonaceous particles [Novakov et al., 1974; Brodzinsky et al., 1980; Mamane and Gottlieb, 1989]. In addition to directly forming sulfate, these reactions of SO2 with carbonaceous particles can convert hydrophobic particle surfaces into hydrophilic surfaces. A variety of mechanisms for SO2 oxidation reactions on carbonaceous surfaces have been proposed, involving a variety of oxidizing species [Novakov et al., 1974; Liberti et al., 1978; Tartarelli et al., 1978; Britton and Clarke, 1980; Baldwin, 1982]. In this work, our focus will not be on evaluating mechanisms, but rather demonstrating that rates of heterogeneous reactions can be sufficiently high to explain the observed sulfate enhancements during the wood smoke episode.

[47] In order to obtain a preliminary estimate of the sulfate enhancement due to heterogeneous reactions, some estimate of SO2 oxidation rate is necessary. One way to estimate an upper limit on the rate of heterogeneous sulfate formation is to calculate the rate of surface impingement of gas-phase SO2 on wood smoke particle surfaces, and then to estimate a fraction of the impinging molecules that adsorb and react. The fraction reacted is typically quantified in terms of a reactive uptake coefficient (γ) [Ullerstam et al., 2002, 2003]. The number of reactive collisions with the surface (the sulfate formation rate) is defined as the reactive uptake coefficient multiplied by the total number of surface collisions per unit time (Z).

display math
display math

where v is the mean molecular velocity of SO2, calculated as equation image and A is the effective particle surface.

[48] CAMx codes were modified to account for these heterogeneous reactions and to distinguish the sulfate mass produced from the heterogeneous pathway. The rate of impingement of SO2 onto fire particles was calculated for the modeling domain shown in Figure 3. The mass of fire particle emissions was taken from the emission inventory of Junquera et al. [2005] and the number of particles and available surface area was calculated by assuming a particle density of 1.5 g/cm3 and an average particle diameter of 0.25 μm. The average particle diameter was based on the work of Reid et al. [2004], who listed aerosol volume median diameters for fresh smoke in the 0.25–0.3 μm range. Note that the hydration state of smoke particles was not distinguished here. The regional photochemical model was modified so that the rate of impingement of SO2 onto fire particles was calculated for each 1-hour time step in each grid cell. In addition, assuming a value of γ of 10–2, the estimated particulate sulfate produced by heterogeneous reactions was calculated (wood smoke catalyzed sulfate). Results of the simulation for 6 September are shown in Figure 10. Figure 10 compares the concentrations of secondary sulfate formed through homogeneous/aqueous reactions and the concentrations of secondary sulfate formed via heterogeneous reactions on wood smoke particles.

Figure 10.

Ground level sulfate concentrations (μg/m3) on 6 September predicted by CAMx (left) via gas and aqueous-phase reaction pathways of SO2 from sources within the domain and (right) via impingement of SO2, from sources within the domain, onto wood smoke particles, with a reactive coefficient of 10–2 at (a) 0000 LT (b) 0600 LT (c) 1200 LT, and (d) 1800 LT. Observed concentrations (μg/m3) at the La Porte (LPT) monitoring site are shown (data are from M. R. Canagaratna, personal communication, 2005).

[49] The right hand side of Figure 10 shows the temporal evolution of a large plume of sulfate, produced by heterogeneous oxidation on wood smoke. The plume forms in Louisiana on the morning of 6 September and advects toward Houston. By the evening, the plume has passed the Houston measurement sites. This is consistent with the sulfate measurements made at La Porte (also shown in Figure 10). Although the location of the predicted plume is slightly south of the La Porte site, the general features are consistent with observations, particularly the time of onset of the plume and the time at which the plume dissipates. In contrast, the sulfate produced by conventional routes (through gas and aqueous phase chemistry), shown in the left hand side of Figure 10, does not replicate either the magnitude or the spatial and temporal distribution of sulfate concentrations on 6 September; that is, only small amount of sulfate was produced and sulfate does not accumulate in the Houston area.

[50] These analyses suggest that the heterogeneous oxidation of SO2 on wood smoke particles can explain the observed enhancement of sulfate concentrations during the wood smoke episode. If a reaction probability of 10–2 reactions per SO2 wood smoke particle collisions is assumed, the magnitude and spatial distribution of sulfate concentrations is consistent with observations.

5. Conclusion

[51] A chemical mass balance (CMB) model was used to estimate the contributions of primary and secondary sources of PM2.5 in Houston, Texas. The analysis consisted of running the model on data collected during a regional wood smoke episode and separately on data collected on other days not affected by the wood smoke episode. Analysis of the data collected on non–wood smoke episode days indicated that the major contributors to PM2.5 mass were secondary sulfate, diesel and gasoline powered vehicles, other organics, and wood combustion. Vegetative detritus, meat cooking and road dust sources were determined to be minor contributors. Secondary sulfate represented almost 100% of the sulfate measured whereas primary sources were found to account for the majority (64–69%) of the measured organic carbon at three sites. Including the secondary formation sulfate and other organic mass, the model accounted for 94% of PM2.5 mass at HRM-3 and was overestimated by 8–11% at Aldine and La Porte. On the wood smoke episode days, the same major sources were found to be contributing to PM2.5 levels at each site except that the contribution of meat cooking source was not significant on the wood smoke days, as evidenced by the considerable decrease in octadecenoic acid levels during the episode. The contribution of the wood combustion source showed an increase by an average of 200% at all sites whereas the contributions of other primary sources did not increase significantly during the wood smoke episode. This result is consistent with the increase in the concentrations of levoglucosan during the wood smoke episode at each site, while the concentrations of marker species for other sources were not dramatically higher on the wood smoke episode days. PM2.5 mass almost doubled on the wood smoke episode days, and the contributions of the secondary sources, namely secondary sulfate and other organics increased 68% and 228% at each site, respectively. The apportioned mass together with the secondary aerosol contributions accounted for 97% and 89% of PM2.5 mass at Aldine and La Porte, respectively and was overestimated by 4% at HRM-3. The increase in the contributions of secondary sulfate aerosols during the wood smoke episode was examined using a 3-D photochemical grid model. The simulations, together with ambient data, indicated that the increase in sulfate concentrations during this wood smoke episode could be explained by heterogeneous/surface reactions on wood smoke particles.

Ancillary