Secondary Organic Aerosols from Aromatic Hydrocarbons and their Contribution to Fine Particulate Matter in Atlanta, Georgia

Abstract

Tracers of secondary organic aerosols (SOA) from thirteen aromatic hydrocarbons were quantified in laboratory smog chamber experiments. Class-specific SOA tracers emerged, including 2,3-dihydroxy-4-oxo-pentatonic acid (DHOPA) from monoaromatic volatile organic compounds (VOCs), phthalic acid from naphthalene and 1-methylnaphthalene, and methyl-nitrocatechol isomers from o,m,p-cresol oxidation. Organic carbon mass fractions (f_SOC) for these and other tracers were determined and extend the SOA tracer method widely used to apportion biogenic SOC. The extended SOA tracer model was applied to evaluate the sources of SOC in Atlanta, GA during summer 2015 and winter 2016 after modifying the chamber-derived f_SOC values to reflect SOA yields and local VOC levels (f_SOC’). Monoaromatic, diaromatic, and cresol SOC contributed an average of 24%, 8%, and 0.12% of organic carbon (OC) mass during summer and 17%, 5%, and 0.27% during winter, respectively. Cresol SOC peaked during winter and was highly correlated with levoglucosan (r=0.83, p<0.001), consistent with it originating from biomass burning. Together, aromatic, biogenic, and biomass burning derived SOC accounted for an average of 77% and 28% of OC in summer and winter, respectively. The new understanding of SOA composition from aromatic VOCs advances the tracer-based method by including important precursors of SOC and enables a better understanding of the sources of atmospheric aerosol.

Graphical Abstract

1. Introduction

Secondary organic aerosol (SOA) forms in the atmosphere through the photooxidation of biogenic and anthropogenic volatile organic compounds (VOCs), and comprises a significant fraction of organic aerosol (Goldstein and Galbally, 2007; Hallquist et al., 2009; Kroll and Seinfeld, 2008; Zhang et al., 2007). SOA from biogenic precursors (BSOA) dominates the global SOA flux (Hallquist et al., 2009), with isoprene and monoterpene contributing 55% of the annual emission of all VOCs (Guenther, 1999). Aromatic hydrocarbons represent 17% of the global inventory of anthropogenic VOCs (Piccot et al., 1992), and contribute up to 30% of VOCs in urban atmospheres (Calvert et al., 2002). Nonetheless, laboratory and ambient studies indicate that estimations of anthropogenic SOA (ASOA) is largely underestimated (Henze et al., 2008; Volkamer et al., 2006), especially in urban and industrial environments (Ding et al., 2012; Stone et al., 2010). A tracer-based approach to apportioning SOA to its sources enables resolution of the relative contributions of specific anthropogenic and biogenic precursors (Kleindienst et al., 2007).

SOA tracers are oxidation products that are unique to specific VOC precursors. Tracer-based SOA estimation (a.k.a. the “SOA-tracer method”) is the quotient of ambient measurements of SOA tracers and laboratory-generated mass fractions of these tracers in SOA (Kleindienst et al., 2007). Because the mass fractions are determined on a compound-by-compound basis, this approach provides distinction of the VOC precursors to SOA. Additionally, the measurement of SOA tracers is compatible with measuring molecular markers of primary aerosol sources and can be combined to estimate primary and secondary sources of organic aerosol (Schauer et al., 1996). Intensive laboratory studies have focused on identifying oxidation products that can be used to trace SOA to their precursors of origin, particularly BSOA from isoprene, monoterpenes, and sesquiterpenes (Christoffersen et al., 1998; Claeys et al., 2004; Claeys et al., 2007; Edney et al., 2005; Glasius et al., 2000; Jaoui et al., 2007; Surratt et al., 2007; Szmigielski et al., 2007; Wang et al., 2005). In contrast, only a few tracers have been developed for quantitative estimations of ASOA (Kleindienst et al., 2012b; Kleindienst et al., 2007). Proposed ASOA tracers include 2,3-dihydroxy-4-oxopentanoic acid, phthalic acid, 4-methylphthalic acid, and some nitromonoaromatics because of their specificity, predominant partitioning to the particle phase, and detectability in fine particulate matter (PM_2.5) (Al-Naiema and Stone, 2017). Additionally, 3,5-dinitrosalicylic acid and 4-nitrophthalic acid are proposed ASOA tracers based on their correlation with other ASOA tracers, low vapor pressure, and seasonal and diurnal trends that are consistent with secondary origins (Ikemori et al., 2019).

To further develop these compounds as ASOA tracers, we determine their mass fractions in laboratory-generated SOA produced from the photooxidation of 13 aromatic hydrocarbons, including benzene and its alkyl derivatives, naphthalene and 1-methylnaphthalene, and three cresol isomers. With these data, we expand the SOA-tracer method in terms of the number of aromatic SOA tracers and the VOC precursors that are represented. We present new strategy to relate laboratory chamber experiments to ambient atmospheres by weighing SOA profiles by their SOA yields and ambient VOC concentrations. This approach is increases the applicability of SOA measurements generated in chambers under specific experimental conditions to broader atmospheres. To demonstrate the utility of this expanded SOA-tracer method, we estimate the SOA from aromatic hydrocarbons in PM_2.5 samples collected from Atlanta, Georgia. This urban city is characterized by high emissions of biogenic VOCs in summer and anthropogenic emissions that enhance BSOA formation (Xu et al., 2015a). Prior studies have established the importance of isoprene SOA during summer and biomass burning during winter (Budisulistiorini et al., 2016a; Kleindienst et al., 2007; Rattanavaraha et al., 2017). In contrast, little information is available about the abundance and variability of SOA from aromatic hydrocarbons. In this study, we estimate the relative importance of BSOA and ASOA in summer and winter seasons.

2. Materials and Methods

2.1. Smog Chamber Photooxidation

Steady-state SOA was generated through a set of controlled photochemical reactions in a 14.5 m³ solid walled irradiation chamber. The PTFE Teflon coated reaction chamber was operated as a continuous stirred tank reactor, producing a steady-state aerosol distribution that was sampled. The residence time of gases in the chamber was typically 4 to 6 h. Conditions of the experiments are published elsewhere (Xie et al., 2017). Briefly, 13 hydrocarbons (Table 1) were individually introduced in to the chamber and photooxidized in the presence of NOx. In addition, six of these VOCs were also oxidized by OH in the absence of NO_x via the photolysis of H₂O₂. In all, 19 experiments are presented here. Individual hydrocarbons, as well as nitric oxide (NO), were injected through mass flow controllers from high pressure cylinders containing neat compound in air, by passing air through an impinger containing the neat liquid at a set temperature, or from a syringe pump containing the neat liquid.

Table 1:

Aromatic VOC photooxidation conditions and particle data for NO_x and H₂O₂ experiments, in which [H₂O₂]_o and [NOx]_o represent the initial hydrogen peroxide or oxides of nitrogen concentration introduced to the chamber, respectively, ΔHC is the change in hydrocarbon concentration upon oxidation, and yields are corrected for the wall losses.

Hydrocarbon	[H2O2]o (ppm)	[NOx]o (ppm)	ΔHC (ppmC)	SOA (μg m−3)	SOC (μg m−3)	SOA Yield (%)	SOC Yield (%)
benzene	-	0.036	2.82	530.4	226.1	35.5%	16.4%
toluene	-	0.115	2.53	162.4	73.6	12.0%	6.0%
ethylbenzene	-	0.130	1.99	135.0	64.6	12.6%	6.7%
o-xylene	-	0.216	3.22	222.7	115.6	12.8%	7.4%
m-xylene	-	0.174	3.23	103.2	57.3	5.9%	3.6%
p-xylene	-	0.175	3.37	70.5	38.3	3.9%	2.3%
1,3,5-trimethylbenzene	-	0.276	2.80	111.6	74.6	7.4%	5.5%
1,2,4-trimethylbenzene	-	0.252	3.20	80.8	49.3	4.7%	3.2%
naphthalene	-	0.165	0.65	90.3	47.8	26.6%	15.0%
1-methylnapthalene	-	0.178	0.28	69.2	38.8	46.7%	28.2%
o-cresol	-	0.029	1.12	226.8	118.1	32.3%	21.6%
p-cresol	-	0.038	1.12	251.4	129.8	35.6%	23.7%
m-cresol	-	0.035	0.66	119.3	56.3	28.8%	17.5%
benzene	6.982	-	1.61	53.8	16.8	6.3%	2.1%
toluene	7.922	-	1.65	143.2	64.3	16.3%	8.0%
ethylbenzene	7.834	-	1.72	126.4	60.3	13.6%	7.2%
naphthalene	8.055	-	1.16	238.2	121.1	39.3%	21.3%
1-methylnapthalene	7.814	-	0.67	197.9	95.7	56.3%	29.3%
m-cresol	10.093	-	0.19	78.5	37.1	65.0%	39.5%

Inlet and chamber concentrations of reactant hydrocarbons were measured by gas chromatography with flame ionization detection. Neutral ammonium sulfate seed aerosol at approximately 1 μg m⁻³ was generated by nebulizing a 10 mg L⁻¹ aqueous solution (model 9032; TSI, Inc., Shoreville, MN). The seed aerosol stream was then equilibrated to the dynamically controlled relative humidity (RH) in the chamber. In experiments conducted with NO, RH was typically 30%. In experiments utilizing H₂O₂, the chamber was operated without the addition of water vapor (RH < 2%). All experiments were conducted in the presence of UV light. SOA was collected on Teflon-impregnated glass fiber filters (Pallflex Fiberfilm T60A20). Detailed descriptions about the chamber operation and mass measurements are explained elsewhere (Edney et al., 2005; Kleindienst et al., 2007).

2.2. Site Description and PM_2.5 Sampling

Ambient PM_2.5 samples were collected during summer of 2015 (July 29 to August 27), and winter of 2016 (January 12 to February 18) in Atlanta, GA. A medium volume sampler (3000B, URG Corp.), operating at 90 liters per minute, was positioned on top of the School of Earth and Atmospheric Sciences (~30-40m above ground level), Georgia Institute of Technology (33°46’44.2” N, 84°23’46.2” W), an urban location 2 km north of downtown Atlanta, GA. Samples were collected on 90 mm quartz fiber filters (Pallflex® Tissuquartz™, Pall Life Sciences) from 13:30 to 12:30 the following day (local time). Air flow was monitored before and after each sample was collected using a rotameter (Gilmont Inst.). One field blank was collected for every five samples. Prior to sampling, all filters were pre-cleaned by baking for 18 h at 550 °C. After sample collection, filters were enveloped with prebaked aluminum foil, placed inside petri dishes, sealed with Teflon tape, and stored frozen at −20 °C until analysis.

2.3. Chemical Analysis

2.3.1. Extraction and Analysis of SOA Tracers

SOA chamber samples, ambient PM_2.5, and field blank filters were extracted and analyzed following a previously reported method (Al-Naiema and Stone, 2017). Briefly, a sub-filter fraction was punched, placed in a silanized glass jar, and spiked with isotopically-labelled internal standards. Filters were sonicated in three 10 mL portions of acetonitrile for 10 min each. Extracts were then combined, rotary evaporated to 2 mL, filtered with a 0.25 μm PTFE syringe filter, further evaporated to 100 μL under a gentle stream of high-purity nitrogen (N₂), and stored frozen until analysis. A 10 μl aliquot of each extract was blown to dryness with N₂ and immediately reconstituted with 5 μL of N,O-Bis(trimethylsilyl)trifluoroacetamide with trimethylchlorosilane (BSTFA + TMCS, 99:1, Fluka Analytical 99%), capped and heated for 90 min at 100 °C. Derivatized samples were analyzed using gas chromatography coupled with mass spectrometry (GC-MS; Agilent 7890A GC, coupled with 5975C MS) equipped with a DB-5 column, electron ionization (EI) source (70 eV), and a GC inlet temperature of 300 ◦C. Details about method performance metrics and quantification approaches are found elsewhere (Al-Naiema and Stone, 2017).

2.3.2. Inorganic Ion Speciation

The concentrations of inorganic ions were determined in aqueous filter extracts by ion chromatography with suppressed conductivity detection on a Dionex ICS-5000. Method performance and instrumentation details are published elsewhere (Jayarathne et al., 2016).

2.4. Co-located Measurements

PM_2.5 mass, trace gases (O₃, CO, SO₂, NH₃ and NO_x), and meteorology were obtained from SEARCH monitoring data at Jefferson street (JST) located 2 km west of the site of PM_2.5 collection.

2.5. Statistical Analysis

Spearman’s correlation analysis was used to examine the strengths of monotonic relationships between ambient concentrations organic species and source contributions using MiniTab 19 software. Correlation coefficients (r) were interpreted as follows: very high (0.9-1.0), high (0.7-0.9), moderate (0.5-0.7), low (0.3-0.5), and negligible (0.0-0.3) (Mukaka, 2012). The statistical significance of correlations was evaluated at the 95% confidence interval (p < 0.05).

3. Results and Discussion

3.1. Tracer Mass Fractions from Aromatic Hydrocarbons

Thirteen aromatic hydrocarbons were oxidized with OH in the presence of NO_x, and six of them were oxidized with OH in the absence of NO_x (Table 1). The mass fractions of organic species were determined in the resultant SOA (f_SOA) and SOC (f_SOC) (Table S1) following equations 1 and 2, respectively.

$$f_{\text{SOA}}=[\text{tracer}]∕[\text{SOA}]$$ [Eq.1]

$$f_{\text{SOC}}=[\text{tracer}]∕[\text{SOC}]$$ [Eq.2]

These experiments cover a range of important aromatic VOC precursors and two oxidation conditions. Analytical uncertainties in f are propagated from analytical uncertainties in the measurements of SOA tracers and either SOC or SOA; these values are generally on the order of 10-30% (Table S2), while uncertainties in secondary source contributions using f are in the range of 35%, but can be larger. Because only one chamber experiment was performed for each VOC precursor and oxidant combination, changes to f_SOA and f_SOC in response to hydrocarbon concentration, oxidant concentration, hydrocarbon-to-oxidant ratios, seed-particle chemistry, extent of reaction, relative humidity (RH), etc. could not be characterized. To better understand the uncertainties in f values and to confirm that the same tracers are formed under differing conditions (e.g. VOC versus oxidant limited, or differing RH) an impractical number of chamber experiments would be required. For example, the rate and composition of SOA formation is complex and is not expected to transition from low/no NO_x to high-NO_x conditions linearly or as a step function. To assess the influence of the NO_x concentration on f_SOA and f_SOC, for example, roughly ten more chamber experiments would need to be performed for each precursor / oxidant combination. In total, more than 100 experiments would be needed, which is time and cost-prohibitive. The results described herein advance the precursor and pathway-specific handling of aromatic SOA tracers, but changes to f_SOA, and f_SOC in response to key experimental parameters, such as NO_x concentrations, should be the focus of further study.

Our discussion focuses on tracers that are consistently formed from a class of aromatic VOCs (monoaromatic, diaromatic, or cresols). These SOA tracers include 2,3-dihydroxy-4-oxopentanoic acid (DHOPA), phthalic acid, and methyl-nitrocatechols (MNCs) that were previously recommended as tracers of aromatic SOA based on their detectability, gas-particle partitioning, and source specificity (Al-Naiema and Stone, 2017; Iinuma et al., 2010; Kleindienst et al., 2012b; Kleindienst et al., 2007). Additional organic species measured in the SOA generated from the studied aromatic hydrocarbons are reported in Table S1.

DHOPA, an aromatic ring-opening SOA product, has previously been reported as a tracer for toluene-derived SOA (Kleindienst et al., 2007). It was detected in SOA generated from all of the monoaromatic VOC precursors studied for oxidation in the presence and absence of NO_x (Table 2), indicating that it is a common SOA product for monoaromatic VOCs. The observed f_SOC of DHOPA from toluene in the presence of NO_x (0.0070±0.0008) was similar to the average f_SOC reported for nine toluene photooxidation experiments under NO_x oxidation conditions reported by Kleindienst et al. (0.0079±0.0026) (Kleindienst et al., 2007). For benzene and toluene, smaller f_SOC were observed in the presence of NO_x compared to the absence of NO_x (Table 2), consistent with previous smog chamber observations (Henze et al., 2008; Ng et al., 2007a).

Table 2:

Mass fractions of class-specific SOA tracers generated in laboratory chamber oxidation experiments as a function of SOA and SOC with propagated analytical uncertainties. DHOPA is a tracer of SOA from monoaromatic VOCs, phthalic acid from diaromatic VOCs, and methyl-nitrocatechols from cresols. Measurement uncertainties for individual species are given in Table S1.

SOA Tracer	Precursor	Oxidant	fSOA (uncertainty)	fSOC (uncertainty)
DHOPA
	Benzene	NOx	0.00013 (0.00002)	0.00029 (0.00004)
	Toluene	NOx	0.0032 (0.0004)	0.0070 (0.0008)
	o-Xylene	NOx	0.0024 (0.0003)	0.0047 (0.0006)
	m-Xylene	NOx	0.0036 (0.0004)	0.0064 (0.0008)
	p-Xylene	NOx	0.0039 (0.0005)	0.0072 (0.0009)
	Ethylbenzene	NOx	0.00015 (0.00002)	0.00031 (0.00004)
	1,2,4-Trimethylbenzene	NOx	0.0022 (0.0003)	0.0035 (0.0004)
	1,3,5-Trimethylbenzene	NOx	0.00035 (0.00004)	0.00052 (0.00006)
	Benzene	H2O2	0.00013 (0.00002)	0.00042 (0.00005)
	Toluene	H2O2	0.0068 (0.0008)	0.015 (0.002)
Phthalic acid
	Naphthalene	NOx	0.024 (0.002)	0.046 (0.005)
	1-Methylnaphthalene	NOx	0.0075 (0.0008)	0.013 (0.001)
	Naphthalene	H2O2	0.055 (0.005)	0.108 (0.011)
	1-Methylnaphthalene	H2O2	0.0068 (0.0007)	0.0141 (0.0014)
Methyl-nitrocatechols1
	o-Cresol	NOx	0.041 (0.010)	0.073 (0.072)
	m-Cresol	NOx	0.279 (0.069)	0.494 (0.122)
	p-Cresol	NOx	0.057 (0.010)	0.111 (0.020)

¹⁾The sum of 4-methyl-5-nitrocatechol and 3-methyl-6-nitrocatechol

Phthalic acid is produced from naphthalene and 1-methylnaphthalene in the presence and absence of NO_x. Photooxidation of these diaromatic polycyclic aromatic hydrocarbons (PAHs) generates phthalic anhydride that hydrolyzes to form phthalic acid (Chan et al., 2009; Kautzman et al., 2010). In naphthalene SOA produced in the presence of NO_x, the f_SOA (0.024±0.002) is within the range of results from five naphthalene photooxidation experiments in the presence of NO_x reported by Kleindienst et al. (0.0199±0.0084) (2012b). In the absence of NO_x, f_SOA of phthalic acid from naphthalene doubled (0.055±0.005), indicating a greater formation of phthalic acid relative to the SOA produced in the absence of NO_x, consistent with Kleindienst et al. (2012b). Phthalic acid also formed in SOA from xylenes (Table S1), especially o-xylene that had an f_SOC of 0.045±0.004. The contribution of xylenes to estimates of diaromatic SOA should be considered on a site-by-site basis and is discussed in section 3.3.1 in the context of Atlanta. The quantitative yields of phthalic acid from other alkyl derivatives of naphthalene (i.e., 2-methylnaphthalene and 1,2-dimethylnaphthalene) should be determined to further improve the estimate of SOA from diaromatic species.

Methyl-nitrocatechols (MNC) are produced from the photooxidation of cresols emitted by wood combustion (Iinuma et al., 2010). 4-Methyl-5-nitrocatechol and 3-methyl-6-nitrocatechol were consistently detected in SOA from o,m,p-cresols in the presence of NO_x (Table S1). 4-Methyl-5-nitrocatechol accounted for nearly half of the SOC mass generated from m-cresol (f_SOC = 0.49±0.12), making it the major product of m-cresol photooxidation. 3-Methyl-5-nitrocatechol was not detected in this study, but was previously observed to be only a minor product of m-cresol oxidation (Iinuma et al., 2010). The sum of MNCs mass fractions from m-cresol detected in this study (Table 2) were 2.8 times higher than that reported by Iinuma et al. (2010), likely due to differences in chamber conditions including the initial concentrations of m-cresol, NO_x, and seed particles, as well as the resulting SOA yields. The sum of 4-methyl-5-nitrocatechol and 3-methyl-6-nitrocatechol are recommended for use as the SOA tracer concentration in the SOA tracer method.

In using the reported f_SOA or f_SOC (Table 2) to apportion SOA following the SOA tracer method, the following guidance is provided. First, knowledge of NO_x levels in the ambient environment should be used to select f values from the predominant SOA formation pathway (i.e. absence or presence of NO_x). Second, when measurements of aromatic VOCs are well-characterized for an ambient study site, weighted-average f_SOA’ and f_SOC’ values (as described in section 3.3.1) should be used to account for the relative concentrations of aromatic VOCs in ambient air and their SOA-forming potential. Third, these compounds should only be considered aromatic SOA tracers when they are not expected to have appreciable primary sources, such as an industrial facility. An example of the application of the newly developed f_SOC’ ratios for SOA apportionment is provided in section 3.3.1.

3.2. PM_2.5 Concentrations and Composition in Atlanta during Summer and Winter

Daily PM_2.5 concentrations in Atlanta, Georgia ranged 6.8-21.8 μg m⁻³ (averaging 12.0±3.4 μg m⁻³) from July 29-August 27, 2015 (summer) and 5.0-18.2 μg m⁻³ (averaging 9.0±3.2 μg m⁻³) from January 12-February 18, 2016 (winter). The chemical analysis of PM_2.5 (Figure 1, Table S2) indicated organic carbon (OC) was the most abundant measured PM component, contributing an average of 35±16% of PM_2.5 mass during summer and 34±11% during winter. Meanwhile, elemental carbon (EC) contributed 2.4±0.9% in summer and 4.0±2.0% in winter. OC:EC ratios averaged 15 in summer and 8 in winter, indicative of OC contributions from SOA, particularly in the summertime. The secondary inorganic ions, sulfate (SO₄^2-), nitrate (NO₃^-), and ammonium (NH₄⁺) contributed 15±6%, 0.7±1.0%, and 5.5±2.3% of PM_2.5 mass in summer and 15±6%, 18±12%, and 10±4% during winter, respectively. Higher NO₃^- and NH₄⁺ levels in winter are attributed to the formation of ammonium nitrate in cooler ambient temperatures (Ford and Heald, 2013). Together, OC, EC, secondary inorganic ions, and other minor ions (Table S2) accounted for 60% and 83% of summer and winter PM_2.5, respectively. The remaining PM_2.5 is expected to result from elements associated with OC (or organic matter, estimated to be 2.17 times OC in prior work) and road dust previously estimated to contribute 1-3% of PM_2.5 mass at JST in Atlanta, GA (El-Zanan et al., 2009; Zheng et al., 2002).

Figure 1:

Time series of PM_2.5 mass (black dots) and major components (μg m⁻³) measured at an urban site in Atlanta, GA, A) daily in summer 2015, B) on average in summer, C) daily in winter 2016, and D) on average in winter. Other PM_2.5 was calculated as the difference between the PM_2.5 mass and the measured chemical species (left axis) and is expected to mainly represent other organic elements and road dust. One filter was used to collect PM_2.5 over two consecutive days: Jan. 23-24 and Feb. 11-12.

3.3. Determination of OC Sources

3.3.1. Estimations of SOC from Monoaromatic and Diaromatic Precursors

SOC derived from three groups of aromatic hydrocarbons (monoaromatic, diaromatics, and cresols) was estimated following the SOA-tracer method. This calculation utilized weighted-average f_SOC values (f_SOC’) that were calculated by equation 3:

$$f_{SO{C}^{{}^{\prime }}}=\frac{{\sum }_{i=1}^n{f}_{SOC,i}{Y}_{SOC,i}[C_i]}{{\sum }_{i=1}^n{Y}_{SOC,i}[C_i]}$$ [Eq.3]

where i represents all the VOCs within the compound class (e.g., the eight monoaromatic VOCs listed in Table 2), Y_SOC is the SOC yield from each VOC precursor (Table 1), and [C_i] is the ambient VOC concentration (μg m^-3). Rather than a straight average of the values in Table 2, f_SOC’ accounts for variability of SOC yields across the precursors and the abundance of each precursor in the ambient atmosphere. The SOA tracer model assumes that the laboratory-determined f_SOC and yields are representative of the Atlanta atmosphere and that the ambient VOC concentrations are proportional to the SOA formed from each precursor. The use of readily measured ambient VOC concentrations, rather than the unknown amount of reacted VOC neglects differences in reaction rates among VOC within a compound class. For instance, rates of monoaromatic VOC reacting with hydroxyl radical vary by about a factor of 10 and increase with alkyl substitution (Ng et al., 2007b). Consequently, this approach underestimates the relative contributions of aromatic VOC with greater alkyl substitution to SOA. The application of this assumption is reasonable for moderately reactive aromatic VOC, but would be problematic for highly reactive VOC (e.g. b-caryophyllene with ozone) or very slowly reacting VOCs (e.g. alkanes). While these assumptions remain untested, this approach represents our best estimate of aromatic VOC contributions to SOA with the available data.

For SOC apportionment in Atlanta during summer 2015 and winter 2016, f_SOC’ values were calculated using f_SOC from NO_x oxidation experiments [Eq.3] and are reported in Table S4. The 24-h average monoaromatic VOC concentrations measured every 6^th day were drawn from the EPA AQS Datamart and were averaged over summer and winter periods (Table S5) (EPA, 2018). Measurements of naphthalene and methyl-naphthalene were not available for Atlanta, so f_SOC’ was calculated using naphthalene-to-methyl-naphthalene ratios of 6.0 during summer and 4.6 during winter based on the only available urban concentration data, which corresponded to the L.A. Basin (Reisen and Arey, 2005). Although phthalic acid is emitted from primary combustion sources, like motor vehicles (Kawamura and Kaplan, 1987), it was primarily attributed to SOA during summertime in the L.A. Basin, with approximately 0.1 ng m⁻³ of phthalic acid expected to derive from primary sources (Kleindienst et al., 2012a). In Atlanta, the association of phthalic acid with SOA is supported by its strong and significant correlation with DHOPA in summer (r = 0.82, p < 0.001) and winter (r = 0.71, p < 0.001). The contribution of o-xylene to ambient concentrations of phthalic acid was estimated to be an average of 31% in summer and 44% in winter from the product of the monoaromatic SOC contribution (μg m^-3), relative contribution of o-xylene to the monomaromatic f_SOC’ (8.2% and 9.1%, respectively), and the o-xylene f_SOC for phthalic acid (0.045±0.004). Subsequently, daily phthalic acid concentrations were corrected for the monoaromatic o-xylene contribution prior to estimating diaromatic SOC. Ambient measurements of o,m,p-cresols were not available, so equivalent mass concentrations were assumed in calculating f_SOC’. The weighted-average f_SOC’ values were calculated from local measurements, when possible, and could be further improved by co-located measurements of monoaromatics, naphthalene, methyl-napthalene, and o,m,p-cresols.

In Atlanta, SOC from monoaromatic, diaromatic, and cresol combined together accounted for an average (± standard error) of 1.2±0.2 μgC m⁻³ (ranging 0.2-4.1 μgC m⁻³) during summer and 0.59±0.07 μgC m^-3 (0.14-1.8 μgC m⁻³) during winter. For both seasons, the major aromatic SOC precursors were monoaromatics, followed by diaromatic, and then cresol (Figure 2). Monoaromatic SOC contributed an average of 24±4% of OC mass during summer (0.97±0.15 μgC m⁻³) and 17±2% during winter (0.50±0.06 μgC m⁻³, Figure 3). Higher monoaromatic SOC during summer can result from many factors including higher temperature that promotes evaporation and off-gassing of fuels, solvents, paints, etc.; actinic fluxes; and concentrations of C₇-C₉ aromatic VOCs due the seasonal changes in gasoline composition (Blanchard et al., 2010). Diaromatic SOC contributed an average of 5.5±0.9% of OC in summer (0.24±0.05 μgC m⁻³) and 3.1±0.6% in winter (0.08±0.01 μgC m⁻³, Figure 2). Naphthalene is emitted year-round and is the most abundant PAH emitted from coal-fired boilers and mobile sources (Al-Naiema et al., 2015; Marr et al., 1999). Although naphthalene can be emitted from biomass burning (Bruns et al., 2016), the seasonal trend in diaromatic SOA is opposite that of biomass burning, suggesting that biomass burning is not a major source of naphthalene in Atlanta. The significant and high correlation of diaromatic SOC with monoaromatic SOC in summer (r=0.70, p<0.001) suggest that they have similar anthropogenic origins in this region.

Figure 2:

Average source contributions to PM_2.5 OC (%) during summer winter samples investigated in this study. β-Caryophyllene accounted for 1% during summer and 1.1% during winter. Isoprene SOC represented 0.4% of OC during winter. Other OC was calculated from the differences between average OC mass concentrations and the sum of the all SOC sources and biomass burning, and mainly represents fossil-fuel-related primary sources.

Figure 3:

Contribution of SOC sources (μg m⁻³) in summer and winter PM_2.5 samples analyzed in an urban site of Atlanta, GA.

3.3.2. Biogenic SOC

The contributions from biogenic VOCs (isoprene, α-pinene and β-caryophyllene) to SOC were determined using previously developed SOC-tracer profiles (Kleindienst et al., 2007). Biogenic SOC at the urban sampling site in Atlanta accounted for 45±6% of OC during summer and 6.4±0.6% during winter (Figure 2). Isoprene SOC, estimated from ambient measurements of 2-methylglyceric acid and 2-methyltetrols, contributed 27±3% of OC during summer (1.2±0.2 μgC m⁻³) and 0.38±0.04% (0.01 μgC m⁻³) during winter. Monoterpene SOC, estimated from ambient measurements of 3-hydroxyglutaric acid, 2-hydroxy-4,4-dimethylglutaric acid, and cis-pinonic acid, accounted for 18 ±3% (0.79±0.15 μgC m⁻³) of OC during summer and 4.9±0.6% (0.14±0.02 μgC m⁻³) of OC during winter. β-Caryophyllene SOC accounted for 0.045±0.007 μgC m⁻³ during summer and 0.034±0.005 μg m⁻³ during winter, contributing approximately 1% of OC in both seasons (Figure 2). The low OC contributions from β-caryophyllene SOC (ranging from below the detection limit to 0.06 μgC m⁻³) from were also previously reported in JST between May and August 2005 (Kleindienst et al., 2010). The higher relative contribution of isoprene and monoterpene SOC in summer is consistent with prior studies in the Southeastern U.S. that document higher isoprene and monoterpene flux in summer (Geron et al., 1995; Guenther et al., 1995) as well as larger contributions from isoprene epoxide (IEPOX)-derived organic aerosol (Budisulistiorini et al., 2016b; Rattanavaraha et al., 2017; Xu et al., 2015b) and monoterpene SOC (Kleindienst et al., 2007). Further, our estimate of isoprene-derived SOA, averaging 27% of PM_2.5 OC during summer agrees closely with the 30% of non-refractory PM₁ organic aerosol attributed to isoprene SOA in August 2012 also at the Georgia Tech sampling site using positive matrix factorization of high-resolution aerosol mass spectrometry measurements (Xu et al., 2015b). Overall, biogenic SOC was substantial during summertime and with the largest contributions from isoprene followed by monoterpenes.

3.3.3. Biomass Burning influences in Primary and Secondary OC

3.3.3.1. Primary Biomass Burning

Biomass burning contributions to OC were estimated based on the ambient concentration of levoglucosan. During the summer, prescribed burning is the most prevalent type of biomass burning in southeastern U.S. (Christopher et al., 2009; Lee et al., 2005), which has a levoglucosan-to-OC mass fraction (± uncertainty) of 0.094±0.040 (Lee et al., 2005). During the winter, residential wood combustion is the major type of biomass burning in the Southeastern U.S. (Zhang et al., 2010a), which has an average levoglucosan-to-OC ratio of 0.15±0.03 for the five most prevalent wood types in EPA Region 5 burned in residential stoves (Sheesley et al., 2007). In this way, biomass burning was estimated to contribute an average of 0.54±0.09 μgC m⁻³ (13±2% of OC) during summer and 2.4±0.6 μgC m⁻³ (53±3% of OC) in winter. The wintertime levels are consistent with prior biomass burning estimates for the nearby JST site in January 2002 at 2.14 μgC m^-3 corresponding to 50% of OC (Zheng et al., 2007). The seasonal trend of low biomass burning contributions to OC in summer and relatively higher contributions in summer is consistent with non-refractory PM₁ source apportionment results at two sites in Atlanta in 2012-2013 (Budisulistiorini et al., 2016b; Xu et al., 2015b) and a broader study of biomass contributions to PM_2.5 in the southeastern US in 2007 (Zhang et al., 2010b).

3.3.3.2. Cresol SOC

The post-emission processing of VOCs emitted from biomass burning can form SOA, as has been previously demonstrated in the case of cresols (Bruns et al., 2016). In Atlanta, cresol SOC accounted for <1% of OC in both seasons, but was more than twice as large in winter compared to summer (Figure 2). Cresol SOC was strongly correlated with biomass burning tracer levoglucosan during winter (r=0.83, p<0.001), but not in summer (r= 0.48, p=0.007). Thus, the seasonal differences in cresol SOC are expected to be related to the seasonality of biomass burning, consistent with prior observations in Nagoya, Japan (Ikemori et al., 2019). Although cresols can also be formed from toluene photooxidation (Forstner et al., 1997), the high correlation with levoglucosan during winter supports their origin being biomass burning. While cresol SOC is a component of biomass burning derived SOC, it does not represent all biomass burning SOC. A more robust estimate of biomass burning derived SOC requires other methodological approaches and/or further expansion of the SOA-tracer method to include other SOA-forming gases emitted from biomass burning and their SOA tracers.

3.3.3.3. Other Potential Biomass Burning SOA tracers

SOA products observed in chamber experiments (Table S1) showed strong correlations with levoglucosan in wintertime: 4-nitrocatechol (r=0.75, p<0.001), MNCs (r=0.83, p<0.001), isophthalic acid (r=0.77, p<0.001), terephthalic acid (r=0.84, p<0.001), and 5-nitrosalicylic acid (0.84, p>0.001, Figure 4). The strong wintertime correlations suggest an association of these species with biomass burning, their production in chamber experiments indicates their secondary origin, and they have been associated with aged biomass burning (Al-Naiema and Stone, 2017). 4-Nitrocatechol and MNC have been detected in primary biomass burning emissions (Iinuma et al., 2007; Wang et al., 2017), although the extent to which they form from primary emissions versus secondary transformations has not been established. Meanwhile, 5-nitrosalicylic acid was not detected in primary emissions (Wang et al., 2017) and is expected to be formed from secondary processes.

Figure 4:

Time series of levoglucosan (right axis) and other PM_2.5 species that shows association with biomass burning in both summer and winter samples.

During summer, weaker correlations were observed between levoglucosan and 4-nitrocatechol (r=0.47, p=0.009), MNCs (r=0.48, p=0.007), isophthalic acid (r=0.72, p<0.001), and terephthalic acid (r=0.58, p=0.001). The weaker correlation of these species with levoglucosan during summer suggests that they may be derived from aromatic SOC in summer when biomass burning contributions to OC are smaller. Thus, these species are likely to have multiple secondary sources, such that multi-variant source apportionment techniques (i.e. positive matrix factorization) would be needed to distinguish their relative contributions from biomass burning and other sources.

3.3.4. Other sources of OC

OC not attributed to aromatic SOC, biogenic SOC, or biomass burning accounted for an average of 31±5% of OC during summer and 23±4% of OC during winter. Other OC is expected to include other primary sources of OC (in addition to biomass burning) previously observed in JST samples, including OC from gasoline exhaust, diesel exhaust, and meat cooking (Kleindienst et al., 2010). Due to the low data coverage of hopanes in the analyzed samples, primary fossil sources could not be apportioned, as hopanes are a key species needed to distinguish fossil fuel-related emissions (Schauer et al., 2002). In addition, some unapportioned OC is expected to come from biomass burning-derived SOC (other than that derived from cresols) and other SOC precursors not yet included in this SOA-tracer model.

3.4. Environmental Implications

Although aromatic hydrocarbons are major constituents of the VOCs in urban air (Calvert et al., 2002), the extent to which their SOA (or SOC) is contributing to aerosol mass is not well understood. Previously, the estimation of SOC from aromatic hydrocarbons was limited to a single tracer for toluene (DHOPA) and aromatic SOC was estimated using the DHOPA-to-SOC mass fraction averaged across toluene chamber experiments (Kleindienst et al., 2007). This study demonstrates that DHOPA forms from benzene; o,m,p-xylenes; ethylbenzene; 1,3,5-trimethylbenzene (TMB); and 1,2,4-TMB; in addition to toluene; indicating its suitability to broadly represent monoaromatic SOC. The use of the DHOPA f_SOC based on the average (± standard deviation) of nine chamber experiments of toluene (0.0079±0.0026) (Kleindienst et al., 2007) to estimate aromatic SOC is 2.1 times larger than the average DHOPA f_SOC averaged over eight aromatic hydrocarbons reacted in the presence of NO_x (0.0037±0.0030). Thus, prior estimates of toluene SOC could readily scaled by a factor of 2.1 (or preferably the appropriate factor based in local f_SOC’) to estimate SOC from aromatic more broadly. Additionally, f_SOC and f_SOA values were developed for phthalic acid and MNCs that are products from the photooxidation of diaromatics and cresol VOCs, respectively, which expands the SOA-tracer method to include other VOCs that are highly abundant and important precursors of SOC in ambient air (Iinuma et al., 2010; Kleindienst et al., 2012b; Shakya and Griffin, 2010).

The apportionment of SOC in Atlanta using weighted-average f_SOC’ values reveals an important contribution of aromatic VOCs to SOC and more generally a sizeable anthropogenic influence on SOA. One decade prior, contributions from toluene SOC at JST were estimated to be 0-0.26 μgC m⁻³ in May-August 2005 (Kleindienst et al., 2010). Meanwhile, radiocarbon (¹⁴C) measurements suggested that fossil SOC at JST from 2004-05 accounted for 1.60±0.67 μgC m⁻³ in summer and 2.35±1.32 μgC m⁻³ in winter (Ding et al., 2008), revealing a large gap in understanding of the precursors to anthropogenic SOC. With our extended SOA-tracer method, we estimate that monoaromatic and diaromatic SOC contributes an average (± standard error) of 1.2±0.2 μgC m⁻³ in summer and 0.58±0.07 μgC m⁻³ in winter. These new estimates of aromatic SOC improve mass closure on anthropogenic SOC, and suggest an important role for other anthropogenic SOC precursors, especially in wintertime, that are likely to include other diaromatic VOCs, alkanes, and alkenes to the fossil SOC. Because of the substantial contribution of monoaromatic and diaromatic SOC in wintertime, these SOA precursors are expected to also contribute even more SOC during spring and autumn when temperatures and actinic flux are higher.

The expanded estimate of aromatic SOC provides a more robust comparison of the relative contributions of anthropogenic and biogenic SOC (where anthropogenic SOC includes mono- and diaromatic SOC, but not cresol-derived SOC associated with biomass burning). In Atlanta during summertime, biogenic SOC accounted for 2.0±0.35 μgC m⁻³ with an average (± standard error) biogenic-to-anthropogenic SOC ratio of 1.8±0.2. In winter, biogenic SOC contributed 0.19±0.02 μgC m⁻³ and had a biogenic-to-anthropogenic SOC ratio of 0.35±0.17. The observed seasonal trends in the relative contributions of biogenic and anthropogenic SOC are consistent with ¹⁴C measurements at nearby JST in 2004-05 by which contemporary SOC was estimated to contribute 2.79±1.11 μgC m⁻³ in summer and 1.09±0.95 μgC m⁻³ in winter (Ding et al., 2008), which is equivalent to biogenic-to-anthropogenic SOC ratios of 1.7 in summer and a 0.45 in winter. Although the ¹⁴C measurements were made one decade prior to this study, they are considered to be the most relevant comparison for assessing the biogenic-to-anthropogenic SOC ratios because of their unique ability to clearly distinguish between modern and fossil carbon. Among the challenges in comparing PM_2.5 sources across 10 years, are inter-annual changes in PM_2.5 composition in the Southeastern U.S., particularly decreasing PM_2.5 sulfate concentrations (Attwood et al., 2014). Nonetheless, the aromatic VOC concentrations considered as precursors to SOA in this study are remarkably similar, with 2015 concentrations (Table S4) similar to the 2005 average concentrations for benzene (1.5 ppbC), toluene (4.2 ppbC), o-xylene (0.3 ppvC) and m/p-xylene (1.2 ppbC) (EPA, 2018). Improvements to the absolute estimates gained from the SOA-tracer method may be gained by further extending the SOA tracer method to include additional precursors and/or to more accurately represent atmospheric conditions (i.e. concentrations of precursors and oxidants) at the receptor site.

Biomass burning SOC remains difficult to apportion by the SOA-tracer method, due to the large number of SOC precursors emitted by this source. Cresol is one of many reactive gases that is emitted from biomass burning that forms SOA (Bruns et al., 2016). By our estimates, cresol SOC accounts for 0.12±0.02% of OC during summer and 0.27±0.04% during winter. This indicates the impact of biomass burning SOC in Atlanta, GA is 2.25 times larger in winter than summer. Yet, we believe that these estimates represent only a fraction of the total biomass burning SOC. Better constraints on biomass burning SOA requires more detailed understanding of its precursors, products, and yields.

The f_SOC and f_SOA data presented in this study can be used to extend the SOA-tracer method of estimating aromatic hydrocarbon contributions to SOC and/or SOA in different environments following the guidance provided in sections 3.1 and 3.3.1. The proposed f’ values provide a new approach to relate observations from chamber experiments to ambient atmospheres by weighing SOA profiles by their SOA yields and ambient VOC concentrations. This modification is expected to improve the representativeness of chamber data to ambient atmospheres, given that it is impractical to measure SOA and SOC yields for ambient conditions over a large number of geographic locations under varying seasonal, diurnal, and meteorological conditions. The advancement of the SOA tracer method to apportion ambient SOC is significant, in that it has been applied at many locations around the globe and can simultaneously estimate contributions from biogenic SOC precursors. By understanding the contribution from aromatic and biogenic SOC, the sources of ambient PM can be more accurately evaluated, which holds significance for strategies to maintain and/or improve air quality.