Distinguishing Primary and Secondary Tracers to Quantify Naphthalene and Methylnaphthalene Contributions to Secondary Organic Aerosol in the Pearl River Delta

Abstract

Naphthalene and methylnaphthalene (Nap and MN) are the most abundant polycyclic aromatic hydrocarbons (PAHs) and are important precursors of secondary organic aerosol (SOA) in the atmosphere. 1.2-Phthalic acid (1,2-PhA) and 4-methylphthalic acid (4-MPhA) are usually treated as tracers of SOA from Nap and MN. However, the two tracers also have primary sources, and directly using the tracers to estimate SOA would lead to an overestimation. In this study, we conducted a one-year synchronous observation of the two-ring PAH SOA (SOA_2-rings) tracers at nine sites in the Pearl River Delta (PRD) region. We measured and filtered the suitable emission characteristics of SOA_2-rings tracers for biomass burning, coal combustion, industrial processes and vehicle exhaust sources. Then, we developed a method to distinguish 1,2-PhA and 4-MPhA from primary emissions and secondary formation. The average proportions of 1,2-PhA_pri and 4-MPhA_pri in 1,2-PhA and 4-MPhA were 26.7% and 29.2%, respectively. The direct application of measured 1,2-PhA for estimating SOA_2-rings would lead to an overestimation exceeding 30% in the PRD. Furthermore, we estimated SOA_2-rings using the separated 1,2-PhA_sec and 4-MPhA_sec by the tracer-based method. The average contribution of MN to SOA was around three times that of Nap. In addition, when combined with monocyclic aromatic SOA (SOA_1-ring) and biogenic SOA, the contributions of SOA_1-ring (21%) and SOA_2-rings (25%) to total SOA were comparable. SOA_2-rings was even the largest contributor to total SOA (~44%) in winter. This study revealed that whether to separate the SOA_2-rings tracers for primary emissions and secondary formation is essential in SOA estimation and highlighted that two-ring PAHs make a significant contribution to SOA in the PRD.

1. Introduction

Secondary organic aerosol (SOA) is formed through homogeneous or heterogeneous reactions of volatile organic compounds (VOCs) in the atmosphere [1,2], and it is the most abundant component in organic aerosol (OA) [3,4,5]. Human activities emit quantities of anthropogenic pollutants into the atmosphere, leading to the formation of substantial anthropogenic SOA. Field observations have further confirmed that anthropogenic SOA could even dominate over biogenic SOA (BSOA) at urban sites [6]. Aromatic hydrocarbons, mainly emitted from coal combustion, transportation, solvent use and biomass burning, have been regarded as the major class of anthropogenic SOA precursors for a long time [6].

Recent studies demonstrated that polycyclic aromatic hydrocarbons (PAHs), especially two-ring PAHs with relatively higher volatility, also contribute substantially to anthropogenic SOA formation and should be taken into account [7,8]. Fang et al. reported that SOA derived from two-ring PAHs could contribute 12.4% of total SOA in the Pearl River Delta (PRD) region [8]. Two-ring PAHs are mainly originated from incomplete combustion of various fuels and solvent use [9,10,11]. Nap and MN are the most abundant two-ring PAHs and serve as key precursors of anthropogenic SOA in the atmosphere [7]. A smog chamber experiment simulated the photochemical oxidation process of diesel vehicle exhaust [12]. They found that Nap and MN are the important components of exhaust, and their oxidation products could contribute up to 54% of SOA [12]. This indicated that two-ring PAHs, especially Nap and MN, could also be important anthropogenic SOA precursors.

However, it should be noted that the relative contributions of Nap and MN to SOA or secondary organic carbon (SOC) are still uncertain. During the field observation of a haze event in Beijing, Huang et al. estimated the SOA formed from the oxidation of Nap and MN. Using the precursor consumption-based method, Nap was calculated to contribute 7.6% of SOA, which was much higher than MN (2.6% of SOA). And the two-ring PAHs contributed 10.2% to the total SOA through the precursor consumption-based method, which was lower than the tracer-based method (14.9%) [7]. Based on the CMAQ model, Ye et al. also reported that Nap contributed 8.1% of aromatic-derived SOC, which was 3 times higher compared to MN (2.4%) in the Yangtze River Delta (YRD) region [13]. Yu et al. evaluated the contribution of VOCs to SOC in the YRD region based on the precursor consumption-based method [14]. After adding Nap and MN to the precursor list, the proportion of SOC explained by SOC_VOC increased from 25.3% to 39.5%, and MN replaced toluene as the most significant contributor. However, Nap did not contribute significantly to SOC in the study. Fang et al. estimated SOA from two-ring PAHs in Guangzhou using the precursor consumption-based method, and they found that contributions of Nap and MN to SOA were approximately equivalent [8]. The discrepancies in contributions of Nap and MN to SOA or SOC still exist because different methods and parameters are used in various studies.

The tracer-based method was commonly used to estimate atmospheric SOA in field observations. 1.2-PhA is generated from the hydrolysis of phthalic anhydride [12], which has been confirmed to generate from Nap at high yields [15]; thus, 1,2-PhA could be used as a tracer of SOA from the photooxidation of Nap [16]. 4-MPhA is a product of MN photooxidation and is highly correlated with other anthropogenic SOA tracers like 2,3-dihydroxy-4-oxopentanoic acid (DHOPA) and 1,2-PhA in ambient aerosols [17,18,19]; therefore, it could be used as a tracer of SOA from MN [16,17].

However, some studies found that 1,2-PhA and 4-MPhA also have primary emission sources (e.g., biomass burning, industrial processes, coal combustion, and vehicle exhaust) [11,20,21,22], and their contributions cannot be neglected. Using 1,2-PhA and 4-MPhA as tracers without any modification would lead to a gross overestimation for contributions of two-ring PAHs to SOA or SOC [23,24,25]. For example, during a flight observation over China, 1,2-PhA was used to estimate the two-ring PAH SOC; the calculated two-ring PAH SOC even exceeded the measured OC by a factor of 6 [26], highlighting the need to distinguish the sources (primary and secondary) of 1,2-PhA and 4-MPhA in the air.

In order to clarify the contributions of Nap and MN to SOA or SOC more accurately, we carried out a large-scale and long-term observation activity in the PRD region. The PRD region, in the southern coastal part of China, is one of the most urbanized and industrialized areas of China; anthropogenic emissions in this region are substantial [27,28]. And the concentrations of Nap and MN in the PRD showed an apparent seasonal variation, with the lowest in summer and the highest in winter [29]. The increase in two-ring PAHs during winter would have a high potential contribution to the regional SOA. In this work, we measured the primary emission characteristics from the main primary sources for 1,2-PhA and 4-MPhA. Furthermore, we distinguished the primary emission and secondary formation parts of 1,2-PhA and 4-MPhA based on field observations at nine sites in the PRD region. Finally, using the tracer-based method, we used the revised 1,2-PhA and 4-MPhA data to estimate SOA_2-rings and evaluate the contribution of SOA_2-rings to total SOA in the PRD region.

2. Methodology

2.1. Field Sampling

In order to get a more comprehensive understanding of the pollutant levels in the PRD region of China, a concurrent sampling campaign was conducted at 9 sites during 2015 (Figure 1). These sites included 3 urban sites in Zhaoqing (ZQ), Guangzhou (GZ), and Dongguan (DG); 2 suburban sites in Zhuhai (ZH) and Nansha (NS); and 4 rural sites in Boluo (BL), Taishan (TS), Heshan (HS), and Tianhu (TH). Detailed information about these sites can be found in our previous studies [27,30]. At each site, 24 h samples were collected every 6 days from January to December in 2015 on pre-baked quartz filters (8 × 10 inches) using a high-volume PM_2.5 sampler at a flow rate of approximately 1.1 m³ min⁻¹. Field blanks were collected at each site every month. Samples collected in January, April, July, and October were selected to represent PM_2.5 properties during winter, spring, summer, and fall, respectively. This sampling protocol follows the established methodology for atmospheric research in South China [30,31,32,33]. And these four months could well reflect the meteorological characteristics of the four seasons. Temperature and relative humidity during the sampling periods were within ±10% deviation of the long-term seasonal means. We analyzed a total of 170 samples and 19 blanks in this study, about 4 to 5 samples at each site in each season.

Figure 1. Sampling sites in the PRD region, including 3 urban sites, Zhaoqing (ZQ), Guangzhou (GZ), and Dongguan (DG); 2 suburban sites, Zhuhai (ZH) and Nansha (NS); and 4 rural sites, Boluo (BL), Taishan (TS), Heshan (HS), and Tianhu (TH).

2.2. Emission Source Sampling

Five coal samples from different regions in China were collected, including Ping Ding Shan (PDS), Xuan Wei (XW), Lin Fen (LF), Qin Yuan (QY), and Jin Cheng (JC). Combustion experiments were conducted in a controlled environment chamber to simulate cooking and heating processes in a honeycomb briquette stove, a common household appliance in China. The combustion experiment followed this sequence: The first honeycomb briquette was ignited at the base of the stove using charcoal. After burning for 1 h, the second briquette was added, followed by the removal of the charcoal ash. The third briquette was introduced after the second had burned for an additional hour. The stove was then transferred to the combustion chamber for immediate sampling. During sampling, particulate matter was collected on a quartz fiber filter (Whatman). Details of coal samples and sampling information were described in a previous study [34].

The on-site open biomass burning experiments were conducted in a field in the rural area of Guangzhou. Six types of biomass commonly used for heating and cooking in rural areas of China were examined, including four types of straw (wheat straw, soybean straw, rice straw and corn straw) and two kinds of pine (pine needles and pine branches). Particulate matter was sampled using an Anderson sampling device positioned both upwind and downwind. All samples were collected on quartz fiber filters that were pre-baked at 450 °C for organic contamination removal. The sampling was conducted at a constant flow rate of 28.3 L/min. The net compound concentration was then calculated by subtracting the upwind concentration from the downwind concentration.

Samples of the PM_2.5 from typical small- and medium-sized biomass combustion boilers were collected at 12 factories in Guangzhou, China. The PM_2.5 collection was carried out by using smoke nozzles to extract the flue gas at a height of 1.5 to 7 m from the ground at each sampling point. Then, a dilution channel sampler was used to dilute the flue gas by a certain proportion, and PM_2.5 samples of the flue gas were collected. The sampling flow rate was 16.7 L min⁻¹.

All source and environmental samples were analyzed using identical chemical methods, as detailed in Section 2.3.

2.3. Chemical Analysis

The meteorological parameters (temperature and relative humidity), gaseous pollutants (SO₂, CO, NO₂, NO, and O₃), organic carbon (OC), elemental carbon (EC), biogenic SOA tracers, DHOPA (tracer of SOA_1-ring) and other major PM_2.5 components at the 9 sites were reported in our previous studies [27,30]. In this study, we measured two SOA_2-rings tracers (1,2-PhA and 4-MPhA). An isotopically labeled standard (phthalic acid-3,4,5,6-d₄) was added to the samples as an internal standard. Then, the samples were sonicated in mixed organic solvents of hexane/dichloride methane (1:1, v:v) and methanol/dichloride methane (1:1, v:v) for extraction. The extracts of each sample were combined, filtered, and concentrated to approximately 2 mL. Subsequently, the concentrated solution was used for methylation.

The methylated samples were analyzed through an Agilent gas chromatography/mass spectrometer (GC/MS, 7890/5975C), which was operated in the scan mode with a 30 m HP-5 MS column (i.d. 0.25 mm, film thickness 0.25 μm) with an electron impact (EI) ionization source. The initial GC temperature was set at 65 °C and held for 2 min, then increased to 290 °C at a rate of 5 °C min⁻¹, and maintained for 20 min. A more detailed analytical process of SOA tracers was previously described elsewhere [6]. 1,2-PhA and 4-MPhA were quantitatively determined by the internal standard approach. Table S2 summarizes the concentrations of the two compounds at each sampling site in the PRD region.

2.4. Quality Assurance and Quality Control

The data reported in this study were corrected by field blank values. We analyzed the spiked samples to evaluate the recoveries of the analytical procedures. The results indicated that the recoveries of the analytical method were 92 ± 6% for 1,2-PhA and 95 ± 8% for 4-MPhA. The relative standard deviations of repeated analyses for the target compounds in paired duplicate samples (n = 7) were lower than 15%. Considering the errors in field blanks, recovery and repeatability in analysis, the uncertainties of the four-acid analysis were estimated to be 16% and 14% for 1,2-PhA and 4-MPhA, respectively, using the method developed by Stone et al. [35]. The method detection limits (MDLs) were 0.01 ng m⁻³ for 1,2-PhA and 0.07 ng m⁻³ for 4-MPhA.

2.5. Statistical Analysis of Measurement Data

Prior to statistical analysis, all raw measurement data underwent quality control procedures. Invalid records due to instrument malfunction or maintenance were removed, and outliers were identified and reviewed. Descriptive statistics, including mean, standard deviation, minimum, and maximum values, were calculated for pollutant concentrations and meteorological parameters. To assess the statistical significance of differences in pollutant concentrations among multiple groups (e.g., different seasons or sampling sites), one-way analysis of variance was employed, with the significance level set at p < 0.05. The relationships between pollutant concentrations and meteorological variables were evaluated using Pearson correlation analysis. All statistical analyses and graphical presentations were performed using Microsoft Office 2019 (Excel).

3. Results and Discussion

3.1. Primary Emission Characteristics of 1,2-PhA and 4-MPhA

1,2-PhA and 4-MPhA have long been suggested to be secondary products of Nap and MN oxidation [12,16,36]. For instance, 1,2-PhA and 4-MPhA correlate well with SOA formation from Nap and MN in chamber experiments [16]. Up to now, 1,2-PhA and 4-MPhA were widely regarded as tracers for Nap- and MN-derived SOA [17,37]. However, some other studies have found that primary combustion sources can also directly emit 1,2-PhA and 4-MPhA [22,38], and their emission cannot be neglected under actual atmospheric conditions. For example, Huang et al. found that primary 1,2-PhA could account for 57–80% of the total 1,2-PhA concentration during haze episodes in Beijing’s winter [7]. Therefore, it will lead to an overestimation when using observational 1,2-PhA and 4-MPhA data to estimate the SOA_2-rings directly. In order to estimate a more accurate value of SOA generated from Nap and MN, we need to separate 1,2-PhA and 4-MPhA from the primary emission and the secondary formation.

Primary 1,2-PhA and 4-MPhA are mainly emitted from anthropogenic sources, including biomass burning, industrial processes, coal combustion, and vehicle exhaust [11,20,21]. In this study, we measured the emission characteristics of 1,2-PhA and 4-MPhA from biomass burning, industrial processes, and coal combustion sources (Table S3 and Figure S1).

For the biomass burning source, we determined the 1,2-PhA and 4-MPhA emission characteristics of four types of straws and two kinds of pines. The mass fractions of 1,2-PhA and 4-MPhA in emitted PM_2.5 were in the range of 0.014–0.27 ng/μg and 0.006–0.042 ng/μg. For industrial processes, we measured the 1,2-PhA and 4-MPhA emission characteristics of biomass burning boilers, which are widely used in the PRD region, and mass fractions of 1,2-PhA and 4-MPhA ranged from 0.014 to 1.47 ng/μg and 0.001 to 0.044 ng/μg. The emission characteristics of 1,2-PhA and 4-MPhA for coal combustion were determined by using five types of coal from different coal-producing regions in China. The mass fractions of 1,2-PhA and 4-MPhA were measured to be in the range of 0.068–0.368 and 0.005–0.195 ng/μg. In addition, we also systematically collected published data on 1,2-PhA and 4-MPhA primary emission characteristics [11,22,39], and they are summarized in Table S3 and Figure S1.

Zhang et al. investigated the emission characteristics of chemical species from three types of cereal straw under smoldering and flaming conditions in the laboratory [20], and the mass fractions of 1,2-PhA and 4-MPhA in PM_2.5 were measured to be 0.14 ± 0.06 ng/μg and 0.01 ± 0.026 ng/μg. The emission characteristics of chemical constituents in PM_2.5 from two different coal combustion systems, including industrial boilers and residential stoves burning anthracite, bituminous, and brown coal, have also been investigated [11]. The mass fractions of 1,2-PhA and 4-MPhA emitted by residential stoves were measured to be 0.12–0.34 ng/μg and 0.011–0.028 ng/μg, and the two mass fractions of 1,2-PhA and 4-MPhA emitted by industry boilers were 0.021–0.084 ng/μg and 0.0025–0.0045 ng/μg. Zeng et al. characterized the emission profiles of organic compounds in flue gases from eight categories of stationary sources, encompassing ultra-low emission power plants and typical industries [39], and the mass fractions of 1,2-PhA in PM_2.5 were 0.052–0.160 ng/μg. As for vehicle emissions, Cai et al. measured different types of gasoline and diesel vehicle emissions (1,2-PhA: 0.10–1.07 ng/μg; 4-MPhA: 0.0071–0.47 ng/μg) [21].

In summary, we used the weighted mean mass fractions in PM_2.5 for 1,2-PhA and 4-MPhA as their primary emission parameters (1,2-PhA_pri and 4-MPhA_pri) for each source. Specifically, for biomass burning, we used the average 1,2-PhA and 4-MPhA emission characteristics from our measurements of six types of biomass burning and the data reported by Zhang et al. [20]. Industrial emission characteristics were derived by averaging the data reported in studies by Zhang et al. [11] and Zeng et al. [39] because industrial emissions in the PRD region were predominantly from coal-fired boilers and power plants during our sampling period. Coal combustion emissions were characterized using averaged data from emission characteristics of our measurements of five coal types combined with coal burning data from Zhang et al. [11]. And vehicle emission profiles were obtained by averaging gasoline and diesel vehicle emission characteristics from Cai et al.’s study [21].

As shown in Figure 2, among the four primary emission sources, vehicle exhaust demonstrated a significantly higher 1,2-PhA emission intensity (0.55 ± 0.13 ng/μg PM_2.5), with levels exceeding those from coal combustion (0.19 ± 0.12 ng/μg PM_2.5) by more than twofold and exceeding those from biomass burning and industrial emissions by more than fivefold. Compared to 1,2-PhA, 4-MPhA showed significantly lower emission intensity for all four types of sources. Vehicle exhaust exhibited the highest 4-MPhA emission intensity among the four sources, with the mass fractions being 0.24 ± 0.17 ng/μg PM_2.5, exceeding the other three sources by an order of magnitude. This suggests that vehicle exhaust may be an important source of primary 1,2-PhA and 4-MPhA in the atmosphere.

Figure 2. Primary emission characteristics of 1,2-PhA (a) and 4-MPhA (b) from the four sources. The blue columns represent biomass burning, the light green columns represent coal combustion, the dark green columns represent industrial emissions, and the red columns represent vehicular exhaust. Each colored dot indicates an emission characteristic in a source sample. The box with error bars represents 10th, 25th, 75th and 90th percentiles. The white or black horizontal lines represent the median, and the white or black circles represent the average value.

3.2. Distinction of Primary and Secondary Contributions for 1,2-PhA and 4-MPhA

In this study, we used the following method to separate the primary and secondary fractions of 1,2-PhA and 4-MPhA [7]. 1,2-PhA_pri could be estimated by Equation (1), where $f_{m_i}$ (ng PhA/μg PM_2.5) refers to the mass fraction of 1,2-PhA in PM_2.5 emitted from different emission sources. $f_{C_i}$ (%) is the contribution of each source to PM_2.5, which was determined by the source apportionment results of PM_2.5 in the PRD region during the same sampling period based on positive matrix factorization (PMF) modeling [27]. The PMF source apportionment results indicated that the major PM_2.5 sources in the PRD region were: coal combustion (31%), industrial emissions (21%), vehicle exhaust (13%), and biomass burning (7%).

$${1,2\text{-}\mathrm{PhA}}_{pri}={\sum }_i^n{f}_{m_i}\times f_{C_i}\times {\mathrm{PM}}_{2.5}$$

(1)

Similar to the estimation method of 1,2-PhA_pri, 4-MPhA_pri could be estimated by Equation (2), where $f_{m_j}$ refers to the mass fraction of 4-MPhA in PM_2.5 emitted from different emission sources and $f_{C_j}$ is the contribution of each source to PM_2.5 (%).

$${4\text{-}\mathrm{MPhA}}_{pri}={\sum }_{\mathrm{j}}^{\mathrm{n}}{f}_{m_j}\times f_{C_j}\times {\mathrm{PM}}_{2.5}$$

(2)

Based on the above estimated results, 1,2-PhA and 4-MPhA from secondary formation (1,2-PhA_sec and 4-MPhA_sec) could be calculated by Equations (3) and (4):

$${1,2\text{-}\mathrm{PhA}}_{sec}={1,2\text{-}\mathrm{PhA}}_{obv}-{1,2\text{-}\mathrm{PhA}}_{pri}$$

(3)

$${4\text{-}\mathrm{MPhA}}_{sec}={4\text{-}\mathrm{MPhA}}_{obv}-{4\text{-}\mathrm{MPhA}}_{pri}$$

(4)

where 1,2-PhA_obv and 4-MPhA_obv are the observed 1,2-PhA and 4-MPhA in this study.

As shown in Figure 3, the annual average concentrations of 1,2-PhA_pri in the PRD region ranged from 0.62 to 32.9 ng m⁻³, with a mean concentration of 6.83 ± 1.57 ng m⁻³. The annual average concentrations of 4-MPhA_pri ranged from 0.10 to 9.22 ng m⁻³, with a mean concentration of 1.31 ± 0.46 ng m⁻³ (Figure S2). The concentration of 1,2-PhA_pri is higher in urban sites (8.45 ± 6.95 ng m⁻³) compared to suburban (5.97 ± 4.29 ng m⁻³) and rural (6.22 ± 4.26 ng m⁻³) sites. 4-MPhA_pri concentrations exhibit analogous trends for urban sites (1.83 ± 1.82 ng m⁻³) and have significantly higher concentrations compared to suburban (1.07 ± 0.76 ng m⁻³) and rural sites (1.10 ± 0.82 ng m⁻³). The highest concentrations of 1,2-PhA_pri and 4-MPhA_pri were both observed at the ZQ site, with means of 9.78 ± 8.98 ng m⁻³ and 2.34 ± 0.86 ng m⁻³, respectively. The lowest concentrations of 1,2-PhA_pri and 4-MPhA_pri were both observed at the TH site, with the means of 5.07 ± 2.25 ng m⁻³ and 0.74 ± 0.37 ng m⁻³, respectively. ZQ is an important industrial city in the PRD region, which consumes a large amount of energy and emits more anthropogenic pollutants than other cities, while TH is a rural site with lower anthropogenic emissions. In terms of seasonal variations, 1,2-PhA_pri and 4-MPhA_pri showed higher concentrations in fall and winter and lower concentrations in spring and summer at all sites (Figure 3 and Figure S2).

Figure 3. Spatial and seasonal variations in 1,2-PhA_pri and 1,2-PhA_sec at nine sites in the PRD. The bars in the central panel represent the annual average concentrations of 1,2-PhA_pri and 1,2-PhA_sec. The surrounding panels show the seasonal change at each site. The yellow part of the stacked column chart represents 1,2-PhA_pri, and the pink part represents 1,2-PhA_sec. The blue circular dot represents the mass fraction of 1,2-PhA_pri to 1,2-PhA.

A site-by-site correlation was performed among the nine sites for NO, levoglucosan, 1,2-PhA_pri, O_x (O₃ + NO₂), SO₄²⁻ and 1,2-PhA_sec (Figure 4). NO is a characteristic primary pollutant emitted from anthropogenic sources, which serves as an indicator for primary emission processes such as vehicles and power plants. Levoglucosan is predominantly emitted from biomass burning and is usually used for tracing biomass burning sources [40]. O_x and SO₄²⁻ represent two pivotal categories of secondary pollutants in the atmosphere; their mass concentrations are predominantly determined by atmospheric oxidative ability and precursor concentrations. For correlation analyses involving variables with different sampling frequencies, such as filter-based sample measurements (24 h sampling every 6 days, approximately 18–20 samples per site, total N = 170) and the gaseous pollutant concentrations (hourly automatic monitoring data), the gaseous pollutant concentrations were aggregated to 24 h averages. Only paired data points from the same dates and sites were included in the correlation calculations. As shown in Figure 4, NO and levoglucosan showed no significant correlations across most sites (r: −0.39~0.46, p > 0.05), which is consistent with their characteristics as primary emissions predominantly from local sources. However, O_x and SO₄²⁻ exhibited strong correlations across nearly all monitoring sites (r: 0.64~0.94, p < 0.05), demonstrating the significant regional transport characteristics of O_x and SO₄²⁻. As for 1,2-PhA_pri, significant correlations were only observed among a small fraction of monitoring sites. Compared to 1,2-PhA_pri, better correlations for 1,2-PhA_sec were observed among sites (e.g., TS and BL, NS and BL). In addition, good correlations for 1,2-PhA_sec were found between geographically proximate sites (e.g., ZH and NS, HS and TS, GZ and TH). These results demonstrated that the method used in this study is effective for separating the primary and secondary fractions of SOA_2-rings tracers.

Figure 4. Site-by-site correlation coefficients between each pair of the nine sites for NO, levoglucosan, 1,2-PhA_pri, O_x, SO₄²⁻, and 1,2-PhA_sec. The numerical values in the lower left grid cells represent correlation coefficients, while asterisks (*) in the upper right grid cells denote the statistical significance of these correlations (*: p ≤ 0.05, **: p ≤ = 0.01 and ***: p ≤ = 0.001).

To further validate the effectiveness of the separation method, we conducted a correlation analysis between 1,2-PhA_pri and 4-MPhA_pri with the main primary emission source tracers, such as picene for the coal combustion source, hopanes and steranes for the vehicle exhaust source, levoglucosan for the biomass burning source, and heavy metal Zn for the industrial emission source. As shown in Table S4, 1,2-PhA_pri and 4-MPhA_pri showed significant positive correlations with the four primary emission source tracers at almost all sites in the PRD (p < 0.05), suggesting that the sources of 1,2-PhA_pri and 4-MPhA_pri were significantly associated with these primary emission sources. The results indicate that the separation method could distinguish the primary and secondary fractions of 1,2-PhA and 4-MPhA reliably.

The concentrations of 1,2-PhA_sec in the PRD region ranged from 2.37 to 291.4 ng m⁻³ with a mean concentration of 30.3 ± 10.6 ng m⁻³. The concentrations of 4-MPhA_sec ranged from 0.37 to 45.0 ng m⁻³ with a mean concentration of 4.96 ± 1.17 ng m⁻³. The highest concentrations of 1,2-PhA_sec and 4-MPhA_sec were both observed at the ZQ site (52.6 ± 76.8 ng m⁻³ and 6.67 ± 10.9 ng m⁻³), while the lowest concentrations were measured at the NS (13.5 ± 10.3 ng m⁻³) and TH (3.35 ± 2.48 ng m⁻³) sites, respectively. The concentrations of 1,2-PhA_sec and 4-MPhA_sec were higher in urban sites (6.02 ± 7.48 ng m⁻³), followed by suburban (4.53 ± 5.10 ng m⁻³) and rural (1.02 ± 0.72 ng m⁻³) sites. An IVOC emission inventory study revealed that the total atmospheric naphthalene emissions in the PRD region in 2015 were approximately 10,508 t/year [28]. Comparative analysis at the same six monitoring stations selected in this study demonstrated that the concentration of 1,2-PhA_sec showed a significant positive correlation with its precursor naphthalene emissions (p < 0.05), indicating that the content of a precursor has an important influence on the formation of 1,2-PhA_sec.

The concentrations of 1,2-PhA_sec and 4-MPhA_sec were higher in fall and winter and lower in spring and summer at most sites. Previous studies in the PRD region have also shown that the concentrations of Nap in fall and winter were obviously higher than those in spring and summer [8,29]. This might be attributed to the higher anthropogenic activities (industrial activity, coal-fired power generation, open burning of agricultural biomass, and residential burning of biomass for heating) in fall and winter in the PRD region, leading to an increase in precursor (Nap and MN) emissions, further contributing to the increase in 1,2-PhA_sec and 4-MPhA_sec concentrations. In addition, a significant positive correlation could be found between 1,2-PhA_sec and 4-MPhA_sec and O_x and SO₄²⁻ (Table S5, p < 0.05) at most sites in the PRD, demonstrating that atmospheric oxidation exerts an important influence on 1,2-PhA_sec and 4-MPhA_sec formation.

The proportions of 1,2-PhA_pri in 1,2-PhA (1,2-PhA_pri/1,2-PhA%) were in the range of 18.6% to 39.9%, with an average of 26.7%. As shown in Figure S3, 1,2-PhA_pri /1,2-PhA% was higher in suburban sites (35.7 ± 4.88%) compared to urban sites (25.0 ± 1.81%) and rural sites (23.9 ± 2.26%). In addition, 1,2-PhA_pri /1,2-PhA% was found to be higher in spring and fall than in summer and winter. The proportions of 4-MPhA_pri in 4-MPhA (4-MPhA_pri/4-MPhA%) were in the range of 20.7% to 34.5%, with an average of 29.2%. Similar to the trend of 1,2-PhA_pri/1,2-PhA%, 4-MPhA_pri/4-MPhA% was higher in suburban sites (34.0 ± 8.58%) compared to urban sites (30.3 ± 7.93%) and rural sites (26.4 ± 3.58%). 4-MPhA_pri/4-MPhA% was also found to be higher in spring and fall than in summer and winter. The non-negligible contributions from 1,2-PhA_pri and 4-MPhA_pri suggest that, although atmospheric 1,2-PhA and 4-MPhA were predominantly secondarily formed, their direct application for estimating SOA_2-rings would lead to an overestimation exceeding 30%.

3.3. Estimation of SOA_Nap and SOA_MN

Using the separated 1,2-PhA_sec and 4-MPhA_sec, we could calculate SOA_Nap and SOA_MN over the PRD region using the tracer-based method [41]. However, 1,2-PhA_sec was detected in both Nap and MN photooxidation products [16]. Therefore, 1,2-PhA_sec generated by Nap oxidation (1,2-PhA_sec(Nap)) should be estimated by subtraction of 1,2-PhA_sec generated by MN oxidation (1,2-PhA_sec(MN)) from 1,2-PhA_sec. 4-MPhA_sec was only detected in MN oxidation products [19]. The 1,2-PhA_sec(MN) and 4-MPhA_sec formation from MN photooxidation in chamber experiments was studied by Kleindienst et al. [16], and the ratio (r) of 1,2-PhA_sec(MN) to 4-MPhA_sec was 5.43 ± 2.25.

Thus, 1,2-PhA generated by Nap oxidation (1,2-PhA_sec(Nap)) could be estimated by Equation (5):

$${1,2\text{-}\mathrm{PhA}}_{sec\left(Nap\right)}={1,2\text{-}\mathrm{PhA}}_{sec}-r\times {4\text{-}\mathrm{MPhA}}_{sec}$$

(5)

Finally, in this study, we applied 1,2-PhA_sec(Nap) and 4-MPhA_sec to estimate SOA_Nap and SOA_MN and to check the contribution of SOA_2-rings to PM_2.5.

In ambient air, SOA_Nap and SOA_MN can be estimated by dividing measured 1,2-PhA_sec(Nap) and 4-MPhA_sec by f_SOANap and f_SOAMN, with the assumption that the chamber-generated relationship between tracers and SOA is stable in the atmosphere (Equations (6) and (7)):

$${\mathrm{SOA}}_{Nap}={\displaystyle \frac{{1,2\text{-}\mathrm{PhA}}_{sec \left(Nap\right)}}{f_{SOANap}}}$$

(6)

$${\mathrm{SOA}}_{Nap}={\displaystyle \frac{{4\text{-}\mathrm{M}\mathrm{PhA}}_{sec}}{f_{SOAMN}}}$$

(7)

The mass fractions of 1,2-PhA and 4-MPhA in SOA_Nap and SOA_MN (f_SOANap and f_SOAMN) have been measured in chamber experiments [16,36]. However, the value of f_SOA varied under different NO_x conditions. The chamber experiments carried out by Kleindienst et al. demonstrated that f_SOANap is almost stable at 2.0% when NO_x is both present and absent [16]. The f_SOAMN is 0.18% under NO_x-absent conditions and lower than 0.64% under NO_x-present conditions. Al-Naiema et al. measured f_SOANap values of 2.4% and 5.5% under NO_x-present and NO_x-absent conditions [37], respectively. NO_x levels are relatively more abundant in the PRD region during the sampling period [30]. After considering the combined effects of NO_x on f_SOA and actual atmospheric conditions, we chose f_SOANap and f_SOAMN under NO_x-present conditions, which were 2.0% and 0.64%, as estimated by Kleindienst et al. [16], respectively (Table S6).

As shown in Figure 5, the annual average concentrations of SOA_Nap and SOA_MN were 0.28 μg m⁻³ and 0.76 μg m⁻³, respectively. Both SOA_Nap and SOA_MN exhibit similar seasonal variations, with higher levels in fall and winter, which is consistent with the seasonal pattern of 1,2-PhA_sec and 4-MPhA_sec. The average contribution of MN to SOA was three times that of Nap in the PRD. The annual averages of estimated SOA_2-rings ranged from 0.38 to 1.93 μg m⁻³ among the nine sites, with a mean of 0.98 ± 0.44 μg m⁻³. The mean values in winter, spring, summer and fall were 2.03 ± 1.20, 0.52 ± 0.29, 0.69 ± 0.32 and 0.84 ± 0.27 μg m⁻³, respectively.

Figure 5. Spatial and seasonal variations in SOA_Nap (the green bar) and SOA_MN (the yellow bar) at nine sites in the PRD. The central figure presents the annual average of SOA_Nap and SOA_MN at each site. The surrounding figures show the seasonal change in SOA_Nap and SOA_MN at each site. The error bar represents one standard deviation of the sum concentration of SOA_Nap or SOA_MN in each season.

We also summarized the relative contributions of SOA_1-ring and SOA_2-rings to total aromatic SOA (ASOA, sum of SOA_Nap, SOA_MN, and SOA_1-ring) estimated in this and previous studies (Figure 6). In this study, the contributions of SOA_1-ring and SOA_2-rings were comparable, accounting for approximately 47% and 53% of ASOA, respectively (Bar 1 in Figure 6). However, in the study by He et al. in Dongguan, Guangzhou, and Nanhai, which also used the tracer-based method (Bar 2–4 in Figure 6), SOA_2-rings was overwhelmingly dominant, and SOA_1-ring contributions to ASOA were typically less than 20% [42]. This discrepancy is likely because they estimated SOA_2-rings directly based on the measured concentrations of the SOA tracer without explicitly differentiating the primary and secondary contributions. The relative contribution of SOA_2-rings to ASOA estimated by the precursor consumption-based method (Bars 5–7) [7,8,14] and the model-based method (Bar 8) [13] was lower than that in our study. The underestimation of the SOA_2-rings contribution via the precursor consumption method compared to our study might be attributed to the negatively biased SOA yields from uncorrected vapor wall losses in smog chamber experiments [43]. And the underestimation of SOA_2-rings in the model-based method might be attributed to the omission of IVOC precursors in emission inventories [12,43].

Figure 6. The relative contributions of SOA_1-ring (orange columns with green boxes), SOA_Nap (light red columns) and SOA_MN (dark red columns) to ASOA in this study (Bar 1, PRD 9 sites) and other studies by different methods (Bars 2–8). The blue boxes represent SOA_2-rings. a: The tracer-based method data in Dongguan, Guangzhou and Nanhai were obtained from He et al. [42]. b: The precursor consumption-based method data of Guangzhou was from Fang et al. [8]. c: The precursor consumption-based method data of Beijing was from Huang et al. [7]. d: The precursor consumption-based method data of Taizhou was from Yu et al. [14]. e: The model-based method data of Taizhou was from Ye et al. [13].

Regarding the composition of SOA_2-rings, our study identified a dominant contribution of SOA_MN. SOA_MN accounted for approximately 40% of ASOA and 75% of SOA_2-rings, significantly higher than the contribution of SOA_Nap. Our results exhibit a strong similarity to the precursor-based study conducted in Taizhou [14] (Bar 7 in Figure 6), in which SOA_MN accounted for approximately 33% of the ASOA and > 85% of the SOA_2-rings. It revealed that MN is the main precursor of SOA_2-rings formation. In the study utilizing the precursor consumption-based method conducted in Beijing [7] (Bar 6 in Figure 6), the SOA_2-rings fraction in Beijing was dominated by SOA_Nap, which may be attributed to the higher SOA yield of naphthalene relative to methylnaphthalene used in the study.

The main precursors of SOA include isoprene, monoterpene, and β-caryophyllene, which were emitted from biogenic sources, and monoaromatics, as well as Nap and MN from anthropogenic sources. The concentrations of SOA tracers of BVOCs and monoaromatics during the same period and at the same sites were measured in our previous studies [30], and the f_SOA values we used are summarized in Table S6. Utilizing the tracer-based method, we calculated the SOA derived from both biogenic and anthropogenic sources, abbreviated as SOA_tra. The total SOA_tra concentrations (sum of SOA_Nap, SOA_MN, SOA_1-ring and BSOA) at the nine sites in the PRD were estimated [44,45]. The annual average SOA_tra was estimated to be 3.15 ± 2.26 μg m⁻³ at the nine sites. The highest concentration occurred at the ZQ site (5.14 ± 5.10 μg m⁻³), and the lowest concentration was observed at the NS site (1.81 ± 1.66 μg m⁻³).

To evaluate the contribution of SOA estimated by the tracer-based method (SOA_tra) to the total SOA, we calculated SOC using the EC-tracer method at the nine sites, and further used 1.6 as the ratio of SOA to SOC (SOA/SOC) to estimate the SOA_EC to represent the total SOA [44,46].

Since EC is emitted solely from primary combustion sources (e.g., fossil fuel and biomass burning), it is widely used as a tracer for primary organic carbon (POC). Based on this assumption, POC is calculated by multiplying EC by a representative primary OC/EC ratio. Secondary organic carbon (SOC) is then derived as the difference between total OC and estimated POC. SOC can be estimated by Equations (8) and (9):

POC = (OC/EC)_pri × EC

(8)

SOC = OC − POC

(9)

where OC and EC can be obtained from observations; (OC/EC)_pri is the ratio of aerosol OC and EC from fresh primary emissions. We used the minimum R-squared (MRS) method to estimate (OC/EC)_pri [47,48]. The MRS method is premised on the assumption that EC is essentially unrelated to SOC, and thus, (OC/EC)_pri should be the value that minimizes the correlation coefficient between estimated SOC and measured EC.

SOA_EC (SOA estimated by the EC-tracer method) and SOA_tra exhibit a strong positive correlation (r = 0.624, p < 0.001), indicating the estimated SOA_EC and SOA_tra values were reliable (Figure S4). As shown in Figure 7a, the annual average SOA_EC over the PRD was 4.09 ± 1.40 μg m⁻³; the lowest level was observed at NS (2.57 ± 1.57 μg m⁻³), and the highest concentration occurred at ZQ (6.55 ± 3.09 μg m⁻³).

Figure 7. (a) Proportions of 6 types of SOA_tra in the total SOA at each site. (b) Average proportions of 6 types of SOA_tra in the total SOA at the 9 sites.

Figure 7b shows the proportion of six different types of SOA_tra in the total SOA (SOA_EC). Based on the whole year dataset, ASOA (2.02 ± 0.71 μg m⁻³) exhibited a dominant proportion over bsoa (1.13 ± 0.33 μg m⁻³, 36.2%) over the nine sites, accounting for 46.9% of SOA_EC. The contributions of SOA_2-rings and SOA_1-ring were comparable, accounting for 25% and 21% of SOA_EC, respectively. In addition, SOA_2-rings was the largest contributor to SOA_EC (~44%) in winter, and SOA_1-ring contributed ~ one third of SOA in winter. In summer, monoterpenes instead of SOA_2-rings became the largest contributor to SOA_EC (20%) (Figure S5). This suggests that two-ring PAHs may make a crucial contribution to the formation of SOA in the PRD region, especially in winter.

The concentrations and compositions of estimated SOC in this study and the literature are displayed in Figure S6. In terms of biogenic SOC (BSOC), our study found that it was dominated by monoterpenes (~20%), which is similar to the results of Cheng et al. [49] in Hong Kong. However, in the studies by He et al. [42] in the PRD region (Dongguan, Guangzhou and Nansha) and by Lewandowski et al. [50] in the U.S. (in Atlanta, Birmingham, Centreville and Pensacola), BSOC was dominated by isoprene-derived SOC.

The relative importance of two-ring PAH SOC (SOC_2-rings) versus monocyclic aromatic SOC (SOC_1-ring)varies significantly across different studies. In U.S. studies, SOC_1-ring consistently outweighs SOC_2-rings [50]. Conversely, in the two previous PRD studies by Cheng et al. [49] and He et al. [42], a dominant contribution of SOC_2-rings and a negligible contribution of SOC_1-ring were found. In our study, we found a similar proportion of SOC_1-ring (8.9%) and SOC_2-rings (10.4%).

As similar tracers were used in the above studies, the different compositions of SOC in the studies might potentially be due to the various tracer concentrations over different sites, the choice of f_SOC, and the decision to distinguish the primary and secondary fractions of the SOA tracers. The results highlight that even though the concentration and composition of SOC remain inconsistent across studies, the contribution of two-ring PAHs is equally critical and can even exceed that of traditional aromatics.

4. Conclusions

Two-ring PAHs, especially Nap and MN, are important precursors of SOA. 1,2-PhA and 4-MPhA are commonly used as tracers to estimate SOA derived from Nap and MN. However, some studies have found that the two tracers have primary sources, and using the tracers without any modification would lead to a gross overestimation of the contributions of two-ring PAHs to SOA or SOC.

In this study, we measured the emission characteristics of SOA_2-rings tracers (1,2-PhA and 4-MPhA) for biomass burning (0.067 ± 0.082 ng/μg PM_2.5 and 0.013 ± 0.012 ng/μg PM_2.5) and coal combustion (0.19 ± 0.12 ng/μg PM_2.5 and 0.018 ± 0.010 ng/μg PM_2.5) sources and summarized the emission characteristics of SOA_2-rings tracers from previous studies for industrial processes (0.078 ± 0.051 ng/μg PM_2.5 and 0.0035 ± 0.0014 ng/μg PM_2.5) and vehicle exhaust (0.55 ± 0.13 ng/μg PM_2.5 and 0.24 ± 0.17 ng/μg PM_2.5) sources.

Then, we conducted a year-long observation at nine sites in the PRD region and distinguished the primary and secondary fractions of 1,2-PhA and 4-MPhA. The site-by-site correlations were bad for NO, levoglucosan and 1,2-PhA_pri and were good for O_x, SO₄²⁻ and 1,2-PhA_sec, demonstrating that the method used in this study is effective in separating the primary and secondary fractions of SOA_2-rings tracers. The average proportion of 1,2-PhA_pri and 4-MPhA_pri in 1,2-PhA and 4-MPhA were 26.7% and 29.2%, respectively. This suggested that the direct application of 1,2-PhA and 4-MPhA for estimating SOA_2-rings would lead to an overestimation exceeding 30%.

Using the separated 1,2-PhA_sec and 4-MPhA_sec, we calculated SOA_2-rings over the PRD region using the tracer-based method. The annual average concentrations of SOA_Nap, SOA_MN and SOA_2-rings were 0.28 μg m⁻³, 0.76 μg m⁻³ and 0.98 μg m⁻³, respectively. When we combined the SOA_1-ring and BSOA tracers into the estimation, the contributions of SOA_2-rings (25%) and SOA_1-ring (21%) were comparable to SOA_EC. And SOA_2-rings was the largest contributor to SOA_EC (~44%) in winter. This suggests that two-ring PAHs may make a crucial contribution to the formation of SOA in the PRD region, especially in winter.