PM-bound polycyclic aromatic compounds (PACs) in two large-scale petrochemical bases in South China: Spatial variations, sources, and risk assessment

Polycyclic aromatic compounds (PACs) are potential pollutants emitted from the petrochemical industry, whereas their occurrence and sources in petrochemical regions are still poorly known. The present study revealed the spatial variations, compositional profiles, sources and contributions, and health risks of PM-bound PACs in two large-scale petrochemical bases (GDPB and HNPB) in South China. The concentrations of parent polycyclic aromatic hydrocarbons (PAHs) were 7.14 ± 3.16 ng/m3 for ∑18PAHs and 0.608 ± 0.294 ng/m3 for the PAHs with molecular weight of 302 amu (MW302 PAHs) in the GDPB base and 2.55 ± 1.26 ng/m3 and 0.189 ± 0.088 ng/m3 in the HNPB base. Oxygenated PAHs (OPAHs) showed comparable concentrations to the parent PAHs in both the bases and nitrated PAHs (NPAHs) had the lowest mean levels (260 pg/m3 and 59.4 pg/m3 in the two regions). Coronene, 2,8-dinitrodibenzothiophene, and dibenzo[a,e]fluoranthene showed remarkably higher contributions to the PAC and can be PAC markers of the petrochemical industry source. Five sources of PACs were identified respectively in both petrochemical bases by the positive matrix factorization (PMF) model. The vehicle (and ship) traffic exhaust was the primary source of PACs (contributed 33% to the ∑PACs), and the sources related to the coking of coal and heavy petroleum and refinery exhaust were identified in both bases, with contributions of 10–20%. PACs in GDPB also contributed from secondary atmospheric reactions (17.3%) and the usage of sulfur-containing fuels (20.9%), while the aromatics industry made a significant contribution (20.1%) to the PACs in the HNPB region. The cumulative incremental lifetime cancer risks (ILCRs) induced by inhalation of PM-bound PACs in both petrochemical bases were low (10–8–10–6). For the sources related to the petrochemical industry, coking activities and the aromatic industry were the significant contributors to the ∑ILCRs in GDPB and HNPB, respectively. This research has implications for further source-targeted control and health risk reduction of PACs in petrochemical regions.


Introduction
The petrochemical industry is one of China's essential pillar industries, accounting for 20% of the gross domestic product (GDP) . There are over 2500 petrochemical enterprises above the designated size in China in 2020, and the number is expected to continue to grow (SSB 2022). A variety of organic and inorganic contaminants and particulate matter may be emitted or leaked into the environment during petrochemical industrial production and pose potential risks to human health (Amoatey et al. 2019;Luo et al. 2018).
Polycyclic aromatic hydrocarbons (PAHs) are mutagenic, carcinogenic, and genotoxic substances, which formed predominantly from pyrolytic processes, such as solid fuel combustion and vehicle exhaust (Du et al. 2020;Lin et al. 2019). The petrochemical industry is also an important source of PAHs since it has large-scale pyrolytic activities (Kabyl et al. 2020;Nguyen et al. 2020). It has been found that the particulate matter (PM)-bound PAH concentrations in the air of petrochemical-industrialized areas were higher than those in urban areas, and the PAHs were dominated by four to six-ringed congeners (Thang et al. 2019;Turap et al. 2018). PAHs releasing pathways from the petrochemical industry mainly include petroleum leaking, coking, and thermal coal combustion, as found in previous research (Chen et al. 2016b;Zeng et al. 2021).
In addition to the high-priority PAHs designated by the US Environmental Protection Agency (US EPA), other polycyclic aromatic compounds (PACs) may also release from the petrochemical industry, such as large PAHs with a molecular weight of 302 amu (MW302 PAHs) and diverse PAH derivatives. MW302 PAHs have been found in environmental matrices, such as airborne particles and soil (Chen et al. 2016a;Chibwe et al. 2017). They are identified to be more persistent, mutagenic, and carcinogenic pollutants compared to the lower molecular weight (LMW) PAHs. Previous research found that MW302 PAHs contribute about 13-57% to the inhalation cancer risk for urban and rural residents in China because of the higher carcinogenicity Wang et al. 2012). The formation of these pollutants is mainly related to the thermal process (Ma et al. 2022). High concentrations of MW302 PAHs have been observed in petrochemical products (e.g., asphalt and coal tar) generated in the cracking and coking processes (Titaley et al. 2016). Oxygenated PAHs (OPAHs) and nitro-PAHs (NPAHs) are important PAH derivatives. Some of them are also more toxic than their parent compounds (Abbas et al. 2018). These derivatives can derive from primary sources (like PAHs) (Wang et al. 2022) and are formed through secondary atmospheric reactions with oxidants (such as hydroxyl radical, nitrate radical, and ozone) (Hu et al. 2019). Liu et al. (2019) measured several NPAHs in the air of a petrochemical-industrialized city and found their concentrations were 1-2 orders of magnitudes higher than those in other cities. Several characteristic pollutants, including nitrodibenzothiophene, nitrobiphenyl, and nitrobenzo(a)pyrene, were identified during oil sand mining, which implies potential pollution of NPAHs may be generated during petroleum processing (Vasiljevic et al. 2021). Nevertheless, the sources and their contributions to the atmospheric PAH derivatives from the petrochemical industry are still poorly known, hindering the source-targeted control of these toxic pollutants in this industry sector.
In this study, parent PAHs, including 18 PAHs (MW < 302 amu) and 22 MW302 PAHs, as well as 30 PAH derivatives, were measured in the total suspended particulates collected from two large-scale petrochemical bases in South China. The objectives are 1) to reveal the contamination and spatial variations of PAHs and their derivatives in the PM, 2) to resolve their atmospheric sources and contributions, and 3) to evaluate the health risks of PACs associated with particle inhalation exposure. This work provides important information on toxic PACs in the air released from the petrochemical industry.

Sample collection
The total suspended particulate samples were collected from two large-scale petrochemical bases in Guangdong Province (GDPB) and Hainan Province (HNPB) in South China, respectively ( Fig. 1). Oil refining, ethylene production, and coal conversion occupy an important position in the petrochemical industry in the GDPB region (SMPC 2022). The petrochemical industry in the HNPB region includes refining of petroleum and natural gas and various petrochemical production (e.g., paraxylene, terephthalic acid, polyethylene terephthalate, and aromatic hydrocarbon) (SHRC 2022). The HNPB is also an important port and the largest commercial petroleum reserve base of Hainan Province. Particulate matter samples were collected at eight sites (G1 to G8) in the GDPB region and 16 sites (H1 to H16) in the HNPB region. Sampling campaigns were conducted in March 2019 in the HNPB region and in December 2019 in the GDPB. PM was collected on a Whatman quartz fiber filter for 24 h using an active large-volume air sampler (TH-1000H, Wuhan Tianhong Instrument Co., China) at a flow rate of 1.0 m 3 /min. In total, 48 PM samples in HNPB and 56 in GDPB were obtained. The loaded filter was wrapped in aluminum foil, sealed in a small polyethylene zip bag, and stored at -20 °C until extraction.

Chemicals and reagents
Materials and methods information was given in detail in the Supplementary Material and was described briefly here. Target analytes (chemical analysis in Supplementary Material) and their authentic standards were purchased from Accu-Standard (New Haven, CT, US). Chromatography-grade solvents, including hexane, dichloromethane, acetone, and methanol, as well as silica gel solid phase extraction cartridges (CNWBOND ultraclean Si SPE, 1 g, 6 mL) were purchased from ANPEL Laboratory Technologies Inc. (Shanghai, China).

Chemical analysis and quality control
Sample preparation was conducted in accordance with our previous study . Briefly, the PM sample was spiked with deuterated surrogate standards and was Soxhlet extracted with a 200 mL mixture of hexane and acetone (v/v = 1:1) for 48 h. The extract was concentrated to about 1 mL, solvent-exchanged to hexane, and then purified with a solid-phase extraction cartridge. Internal standards were added before instrumental analysis. Target compounds were analyzed using a gas chromatograph-mass spectrometer (GCMS-QP2020 XN Shimadzu, Japan). GC separation for PAH 18 , MW302 PAHs, and PAH derivatives was accomplished by a DB-5 MS (30 m × 0.25 mm × 0.25 μm, Agilent), DB-17 MS (60 m × 0.25 mm × 0.25 μm, Agilent), and DB-17 MS (15 m × 0.25 mm × 0.25 μm, Agilent) capillary columns, respectively. MS was operated in the electron ionization (EI) model for parent PAHs and the negative chemical ionization (NCI) model for PAH derivatives. Procedural (n = 14) and field blanks (n = 4) were run with samples, and only trace amounts (< 7.6% of their amounts in the sample extracts) of analytes were found in the blanks. The recoveries of surrogate standards (mean ± standard deviation) for parent PAHs ranged from 52.1 ± 9.82% to 115 ± 14.4%, and from 103 ± 26.0% to 89.1 ± 24.0% for PAH derivatives. Reported concentrations were blank-corrected accordingly. The method detection limits (MDLs) were assigned as the average method blank mass plus three times the standard deviation or ten times the signal-to-noise ratio for compounds that were not found in the blanks, which ranged from 2.11 to 69.6 pg/m 3 for parent PAHs and 0.004 to 67.6 pg/m 3 for PAH derivatives.

Source appointment
The positive matrix factorization (PMF) receptor model was used to identify and quantify the sources of PACs in GDPB and HNPB. PMF is a well know source apportionment analysis model which considers the error estimates in data to solve a restricted weighted linear least-square estimation model (Ren et al. 2021). This model does not need the source spectrum and applies a non-negative constraint to the data to make the source contributions always positive. Given these advantages, this model was recommended by the US EPA as a source appointment model. Considering the detection rates of PAC compounds and their signal/noise ratio, the model conducted in GDPB and HNPB involved 56 samples with 34 species and 48 samples with 24 species, respectively.

Risk assessment method
Risk assessment models developed by the US EPA were adopted to evaluate the cancer risks induced by PACs. The incremental lifetime cancer risk (ILCR) of inhalation exposure to PACs was evaluated in the GDPB and HNPB regions, respectively. Monte Carlo simulation was used to quantify the uncertainty derived from the risk assessment.

Concentrations and spatial distributions
The PM concentrations during the sampling period in the HNPB region (58.4-236 μg/m 3 ) were significantly lower than those in the GDPB region (80.1-341 μg/m 3 ) (p < 0.001) (Fig. S1). The mean concentrations in the HNPB region (117 μg/m 3 ) and the GDPB region (168 μg/m 3 ) were close to or higher than the national ambient air quality standard of China (120 μg/m 3 ). The sites of G5 and G6 in the GDPB region, which are nearby the petrochemical industrial zone, showed the highest PM concentrations. The PM concentrations among the sites in the HNPB region were comparable (112 ± 14.3 μg/m 3 ), except for the H9 site (188 μg/ m 3 ). The PM concentrations in the study areas were higher than those in a petrochemical area of Brazil (21-57 μg/m 3 ) (Caumo et al. 2022) and were comparable to those recently reported in a petrochemical area of South Korea (around 100 μg/m 3 ) (Thang et al. 2019) and in a few Chinese cities, such as Guangzhou (78.8-153 μg/m 3 ) and Hangzhou (mean = 149 μg/m 3 ) Jin et al. 2021).
The concentrations of parent PAHs were 2.15-18.7 ng/ m 3 for ∑ 18 PAHs and 0.123-2.49 ng/m 3 for ∑MW302 PAHs in the GDPB region (Table 1). The G1 site, located in the downtown area, showed the highest concentrations of these parent PAHs (Fig. 2). The concentrations were 1.4-to 3.3fold (for ∑ 18 PAHs) and 1.5-to 3.6-fold (for ∑MW302 PAHs) higher than other sites near the petrochemical enterprises in the GDPB region. We speculate the elevated levels of parent PAHs in the populous urban area may be associated with the strong contribution of sources such as motor vehicle emissions and food cooking. The concentration of parent PAHs G7 site (4.54 ng/m 3 ), downwind of the petrochemical industrial zone, was the lowest. However, the concentration at the G4 site (10.3 ng/m 3 , downwind of a large petroleum refinery) showed a higher concentration of parent PAHs than the two sites upwind of the plant (7.02 and 4.91 and ng/m 3 ). The result indicated that the petrochemical sector exerted a moderate impact on the occurrence of the parent PAHs in the GDPB region. In the HNPB region, the PAH 18 and MW302 PAH concentrations were approximately three times lower than those in the GDPB region (Table 1). Their concentrations also showed a similar spatial distribution, which peaked at the H4 site (4.00 ng/m 3 for PAH 18 and 0.29 ng/m 3 for MW302 PAHs) ( Fig. 2 and Fig. S2). The H4 site is in a medium-size industrial park, housing enterprises including (but not limited to) three petrochemical-related corporations, two concrete corporations, and a packaging corporation. A comparable concentration of parent PAHs was found at a residential site (H16), with values of 3.60 and 0.23 ng/m 3 . Lower concentrations (1.65 ± 0.02 ng/m 3 for ∑ 18 PAH and 0.14 ± 0.02 ng/m 3 for ∑MW302 PAHs) were present at H1 (a control site far away from the industrial area), H15 (a commercial site), and H6 (a paper mill). The concentrations of parent PAHs in PM at other sites in this region were generally comparable (2.51 ± 0.56 and 0.18 ± 0.03 ng/m 3 ). One explanation for the low concentrations of PAHs in the HNPB region compared to the GDBP region may be the low emissions from the industrial sectors. The HNPB region is a coastal town. Therefore, the clean marine air can reduce the overall air levels of PAHs.  OPAHs showed comparable concentrations to their parents in GDPB, with levels of 2.24-17.0 ng/m 3 , but their concentrations in HNPB were substantially lower than the parent PAHs (0.216-1.84 ng/m 3 ) ( Table 1). The spatial distribution of OPAHs was very similar to the parent PAHs in the GDPB region, and their concentrations were found to be significantly positively correlated with PAH 18 (r = 0.920, p < 0.001) and MW302 PAHs (r = 0.872, p < 0.001). This implied that OPAHs in GDPB might have similar sources to their parents. However, the correlations between the OPAH and parent PAH concentrations were moderate (r = 0.566 p < 0.001) in the HNPB region, indicating differences in their sources. Higher levels of OPAHs were also observed at the H4 (1.33 ng/m 3 ) and H16 (1.20 ng/m 3 ) sites, but the concentrations at other sites (with a mean of 0.71 ± 0.13 ng/ m 3 ) varied lesserly compared to those of the parent PAHs. Of these PACs, NPAHs had the lowest concentrations. The concentrations (ranging from 47.0 to 743 pg/m 3 , a median of 246 pg/m 3 ) of NPAHs in the GDPB region showed small spatial variations, although the highest concentrations were also found at the G1 site. The NPAH concentrations in the HNPB region (12.0-117 pg/m 3 , median of 63.0 pg/m 3 ) were also low compared to those in the GDPB region. The concentrations at most sites were also comparable. Less spatial variations were expected as NPAHs in the ambient air are primarily formed via atmospheric reaction with radicals. Table S1 lists the PAC concentrations in PM or PM 2.5 in other cities in China. In general, the concentrations of ∑ 18 PAHs in the present study (5.17 ± 3.91 ng/m 3 ) were much lower than those in cities in northern China  Table S2). The concentrations of other PACs were generally lower than those in other Chinese cities (Ei zein et al. 2019;Ma et al. 2019;Niu et al. 2017). This result indicated that the pillar petrochemical industry in the two cities did not lead to serious air contamination of these PAHs. Figure S3 shows the daily variations of these PACs, which provides information on their sources. First of all, the variations of parent PAHs and OPAHs for a specific site in most cases were very similar during the same periods. The daily variations of NPAHs were only partly similar to those of PAHs and OPAHs. However, site differences in the daily variations were observed. At the G2, G3, and G4 sites, around the petroleum refinery, the daily variations at G2 and G3 sites (upwind of the plant) were identical; while the daily variations at the G4 site (downwind of the plant) were somewhat different. The residential G1 site, which had the highest PAC concentrations, obviously had daily variations differing from the G2-G4 sites. The G5, G6, and G7 sites, which were around the petrochemical industrial park, showed very similar daily variations, although the sampling period for the G6 site was somewhat different. The G8 site, downwind of several enterprises, had different daily variations, although it is only 600 m from the petrochemical industrial park. Moreover, the daily variations at each of the four sites (G5-G8) were even similar for NPAHs and other PACs, which was not observed at the G1-G4 sites. The results clearly reveal the primary emission sources of these PACs at the study sites in this city, including the downtown, the large industrial park, the refinery, and other enterprises. In the HNPB region, the daily variations of concentrations of PAH 18 and MW302 PAHs at each site during the same sampling period were generally similar (Fig. S3). However, there are appreciable differences in the daily variations between parent PAHs and OPAHs at many of the sites, indicating their different sources in the air. This differed from the observation for them in the GDBP region. NPAHs and OPAHs displayed similar daily variations at a few sites. Interestingly, only at the H4 site, with the highest PAC concentrations, the daily variations were similar for MW302 PAHs, OPAHs, and NPAHs. Similarities in the daily variations among the sites (during the same period) were observed for parent PAHs. This was also observed for OPAHs and NPAHs, though to a lesser extent. Specifically, the lowest concentrations were frequently present on the second sampling day for OPAHs and NPAHs but not for the parent PAHs. Presumably, secondary formation or meteorological conditions made a more contribution to the OPAHs because of the low concentrations in this region.

The composition of PACs
In the GDBP region, the overall compositions of PAH 18 at the sampling sites were very similar (Fig. S4). The PAH 18 were dominated by five and six-ring analogues, such as BbF, BkF, BeP, IcdP, and BghiP, together accounting for 68.3 ± 3.97% of the PAH 18 . An exception was the residential G1 site, in which the contributions of three-and four-ring compounds were even lower. The average composition profiles of MW302 PAHs (dominated by DBalP, DBaeF, and DBaeP) and OPAHs (1,2-ACQ, 1.4-CHR, and 6H-BcdP) were very similar among the sites. The average composition profile of NPAHs (all dominated by 2,8-DNDB and 2N-FLA) at G7 and G1 sites were somewhat different from other sites. The daily composition profiles were also very similar in most cases at a specific site; daily variations in the compositions were observed at several sites (G4, G5, G7, and G8). The daily profiles of MW302 PAHs and OPAHs at a specific site varied less. However, apparent differences in the daily profiles of NPAHs were observed, possibly due to the low concentrations and secondary formation in the atmosphere.
In the HNPB region, ∑PAH 18 accounted for 70% of the PACs on average, higher than that in the GDBP region (46.3 ± 4.5%). Five-and six-ring analogues were also the dominant compounds (with a total contribution of 62.5 ± 3.73% of the ∑ 18 PAH) (Fig. S4). The main difference is BkF, which contributed 11.6 ± 2.03% of the ∑PAH 18 on average in the GDBP region, but only 4.71 ± 0.33% in the HNPB region. Nevertheless, there were clear differences in the compositions of other PACs between the two regions. For instance, The contributions of 1,4-CQ (43.0 ± 4.92%), DBalP (39.5 ± 2.51%), and 2,8-DNDB (38.7 ± 8.79%) were all remarkably higher in the GDBP than those in the HNPB (2.20 ± 3.19% for 1,4-CQ and 4.09 ± 2.74% for DBalP, 2,8-DNDB was not detected); while DBaeF (31.5 ± 4.76%), 1,8-NANH (22.4 ± 5.96%), 2N-FLA (24.5 ± 5.86%), and 9N-PHE (20.3 ± 6.19%) were more abundant in the HNPB region. 1,4-CQ is one of the quinone compounds primarily generated from fossil combustion and/or secondary reactions of PAH precursors (Chu et al. 2017;Wnorowski and Charland 2017) and has been found to be a main OPAH species in the atmosphere. Nevertheless, DBalP, a probable carcinogen for humans with a high potency factor identified by the International Agency for Research on Cancer (Organization 2019), is a minor species in various combustion sources (Chen et al. 2016b) and in the ambient air (Polachova et al. 2020;Zhuo et al. 2017). The proportions of 2,8-DNDB in the GDPB region were significantly higher than those usually reported in urban locations (~ 2%) (Zhang et al. 2021a, b). Nitrodibenzothiophene compounds are mainly derived from incomplete combustion of sulfur-containing fossil fuel and natural gas (Zhang et al. 2021b) and have been found to be major NPAHs in oil sand mining regions and coal-driven heavy industrial cities (Manzano et al. 2017;Vasiljevic et al. 2021;Zhang et al. 2021b). This finding implied that more sulfur-containing fuels were used in the petrochemical industry in the GDBP region. 1,8-NANH and 6H-BcdP have also been found to be abundant OPAH species in the atmosphere in previous research (Alves et al. 2017;Niu et al. 2017). DBaeF makes up 25.1% and 31.5% of ∑MW302 PAHs in GDPB and HNPB, respectively, and was generally higher than that in the PM of other cities (5.45%-17.24%) (Chen et al. 2016a;Long et al. 2011). The differences in the compositions of PACs between the two regions may be attributed to the different petrochemical processes or fuels used in the petrochemical industry. Alternatively, other sources, especially the diverse industries in the HNPB regions, influenced the compositions. COR is a significant by-product formed in petroleum refining (Panda et al. 2018). It is noteworthy that COR accounted for 6.90%-16.6% (10.6% on average) of ∑ 18 PAHs in both regions, which was remarkably higher than those in other cities (~ 5%) (Nan et al. 2021). The PAC profiles indicated that COR, 2,8-DNDB, and DBaeF can be markers of PACs from petrochemical industry sources.
The site-specific average profiles of parent PAHs were also very similar in HNPB, with the standard deviations for the proportions of PAH and MW302 PAH compounds among sampling sites ranging from 0.32% to 2.71% and 0.08% to 0.55%, respectively (Fig. S4). However, there were apparent differences in the profiles of the PAH derivatives among the sites compared to those of parent PAHs and those in the GDBP region. For instance, the two sites with the highest PAC concentrations (H4 and H16) showed different profiles of NPAHs or OPAHs from other sites (Fig. S4). OPAHs at the H4 site have a relatively higher proportion of 6H-BcdP (33.5%) compared to those at other sites (11.2-21.8%) and the proportions of OPAH congeners at the H16 site were comparable. For the NPAHs, 1N-PYR was not detected at the H4 site compared with other sites, with proportions ranging from 5.80 to 22.8%. The highest proportion of 9N-ANT (13.1%) and 3N-PHE (24.2%) were observed at H4 and H16 sites, respectively. Likewise, the daily composition profiles were similar for parent PAHs and somewhat different for the derivatives (Fig. S6). The results indicated that the PAC compositions based on short-term measurements (e.g., 24 h) may vary significantly compared to long-term (e.g., 72 h or one week) measurements. In the study regions, despite the differences in their sources (as suggested above), the differences in compositions of PACs were relatively small, possibly due to air diffusion and mixing.

The sources of PACs
Five sources of PACs in the GDBP region were resolved by the PMF receptor model (Fig. 3). Factor 1 was recognized as a source related to vehicle exhaust due to the heavy loading of five and six-ring PAHs (33.8%-55.1%), of which IcdP and BghiP have been suggested as a marker for vehicle exhaust (Donateo et al. 2014). ANTQ, BaAQ, and 6H-BcdP, the OPAHs with heavy loadings in this factor, are also abundant species in diesel exhaust (Kamiya et al. 2017). Factor 2 was largely characterized by COR (30.9%) and MW 302 PAHs (15.6%-30.5%) and was designated as a source related to the high-temperature coking of coal and heavy petroleum. COR is a significant by-product formed in the hydrocracking process of heavy petroleum (Panda et al. 2018). MW302 PAHs in the environment have been primarily linked to the production or usage of coal-derived products as well as coal combustion (Titaley et al. 2016). High concentrations of dibenzopyrenes and naphthopyrenes have also been found in coal tar (Magee and Forsberg 2016;Wang et al. 2013;Wilson et al. 2017). Factor 3 was predominately weighted by three and four-ringed PAHs (26.3%-40.5%), followed by five and six-ringed PAHs (12.2%-31.6%). This factor was identified as refinery exhaust emitted from the gas flare due to incomplete combustion. Refinery exhaust mainly originates from the direct emission of crude oil, in which low molecular weight (LMW) PAHs dominate (Zhao et al. 2014). Several OPAHs and NPAHs also moderately load in this factor which may be generated from the combustion of refinery exhaust. These three factors explained 33.6%, 11.3%, and 16.9% of the ∑PACs in the GDPB region, respectively. Factor 4 was strongly loaded by NPAHs (19.0%-80.7%) and moderately weighted by OPAHs (5.8%-46.4%) and was assigned as secondary atmospheric reactions. Parent PAHs reaction with hydroxyl and nitrate radicals was the most important pathway for NPAHs formation in the atmosphere (Degrendele et al. 2021;Wang et al. 2011). Factor 5 in GDPB was mainly characterized by 2,8-DNDB (63.5%) and was identified as a source related to the usage of sulfur-containing fuels, like coal and natural gas (Tang et al. 2018). 1,8-NANH, 9-FLO, 1,2-ACQ, and ANTQ are also moderately weighted (6.2%-20.6%) in this factor, which were found to be the major homologs of OPAHs in the coal-derived soot particle Wang et al. 2016). The contributions of refinery exhaust and the usage of sulfur-containing fuels were 17.3% and 20.9% to the ∑PACs in the GDPB region, respectively.
Similar source profiles of PACs were also observed in the HNPB region (Fig. 4). Factor 1 was strongly loaded by five to six-ring PAH 18 (28.5%-65.3%) and was designated as the vehicle and ship traffic exhaust source. Previous research also found that port activities contributed about 29%-87% to the ∑PAHs in a port city, and four to sixringed congeners were the dominant PAHs emitted from ship exhaust (Donateo et al. 2014;Zhao et al. 2019). This source contributed 32.3% to the ∑PACs in this region, higher than that in the GDPB region. Factor 2 was also assigned as a source related to the high-temperature coking of coal and heavy petroleum, which was predominately loaded by COR (54.8%) and MW302 PAHs (9.4%-47.0%). This factor contributed to 19.0% of ∑PACs, higher than that in the GDPB region. Factor 3 can be linked to the refinery exhaust source owing to the heavy loading of LMW PAHs (41.5%-63.5%), and this factor contributed 15.2% of ∑PACs, comparable to that in the GDPB region.
Factor 4 and factor 5 were the sources mainly related to OPAHs. OPAHs are not only pollutants emitted from combustion sources, but also important aromatic products produced by the petrochemical industry. Factor 4 is mainly weighted by 1-PyCHO (71.7%), 6H-BcdP (64.0%), BEZO (50.9%), and ANTQ (39.3%), the important aromatic products that are widely used in dyeing, coating, and organic synthesis industries (Joshi et al. 1986;Kajiwara et al. 2014;. Therefore, it was designated as a source related to the aromatic industry, and contributed about 20.1% of ∑PACs. Factor 5 was highly weighted of 1,2-ACQ (55.2%) and 1,8-NANH (44.4%) and was moderately loaded by BEZO (20.9%) and 6H-BcdP (18.1%). However, no characteristic species are found in this factor; thus, this source was defined as the other PAC source. This factor explained 13.4% of ∑PACs.

Risk assessment
The average total benzo[a]pyrene equivalent concentration (∑BaP eq ) in PM was 9.09 ± 6.96 ng/m 3 in the GDPB region, which was about 20-fold higher than that in the HNPB region (0.484 ± 0.325 ng/m 3 ). MW302 PAHs contributed 77.7% of the ∑BaP eq in the GDPB region, though they made up only 3.86% of ∑PACs in the mass. DBalP was the primary contributor to the ∑BaP eq in the GDPB region. OPAHs and the 16 priority PAHs contributed to 13.0% and 9.20% of the ∑BaP eq in the GDPB region, respectively. For the HNPB region, the ∑BaP eq contributions mainly contributed by priority PAHs and MW302 PAHs (66.5% and 31.5%, respectively).
The cumulative ILCRs due to respiratory exposure to PMbound PACs were assessed (Fig. S7). Generally, the estimated cumulative ILCR values (1.70 × 10 -6 -6.03 × 10 -6 ) in the GDPB region were between the acceptable limits (10 -6 -10 -4 ). Spatially, higher ILCR values in the GDPB region were found at the G1 and G4 sites corresponding to the residential area and refinery. For the HNPB region, the ILCRs (7.44 × 10 -8 -2.37 × 10 -7 ) were far below the acceptable limits, and high ILCR values were mainly observed at the H4, H5, and H7 sites, indicating relatively high emission of carcinogenic PAHs in the medium-size industrial park (Fig. S8). The risks at residential sites (H13-H16) were comparable and relatively lower, although some of them (H13 and H16) have higher PAC concentrations.
Source-specific ILCRs in GDPB are depicted in Fig. S7. Overall, PACs emitted from vehicle exhaust contributed 32.7% of the cumulative ILCRs, due to the high contribution (33.6%) of this source to the PACs. Coking of coal and heavy petroleum source made up 25.9% of total ILCRs, although this source is the least contributor to the ∑PACs (11.3%). DBalP, a pollutant with high carcinogenicity generated from this source, is mainly responsible for the risks. Sulfur-containing fuels, refinery exhaust, and secondary atmospheric reactions contributed 16.6%, 11.5%, and 13.3% of the total ILCRs, respectively. In the HNPB region, vehicle and ship exhaust sources also contributed 50% of the ILCRs, higher than their contribution to the PACs (Fig. S8). Obviously, vehicle or ship exhausts are the most robust PAC source for the cancer risk in these two petrochemical bases. It is also the primary factor that impacts public health in the city of Spain (Callén et al. 2014). The aromatic industry, refinery exhaust, and coking (coal and heavy petroleum) contributed 17.8%, 10.8%, and 11.3% of the total ILCRs, respectively, consistent with the contributions of these sources to the PACs. The contribution of the other PAC source accounted for 9.1% of total ILCRs. Therefore, coking activities and the aromatic industry were the most harmful emission sources related to the petrochemical industry in the GDPB and HNPB regions, respectively.

Conclusions
The present research revealed the spatial variations, sources and contributions, and health risks of PM-bound PACs in two large-scale petrochemical bases in South China. PAC concentrations in GDPB were about three times higher than those in HNPB, but their concentrations were generally lower or comparable to those in most cities in China. PAH 18 and OPAHs have similar contributions to the PM-bound PACs in GDPB, which differed from HNPB in that PACs were predominant by PAH 18 . This composition difference may be attributed to the different industrial processes or fuels used in these two petrochemical bases, and COR, 2,8-DNDB, and DBaeF can be PAC markers of the petrochemical industry source. PMF receptor model revealed that vehicle (and ship) traffic exhaust was the primary source of PACs in both petrochemical bases. Several PAC sources related to the petrochemical industry were also identified, including the coking of coal and heavy petroleum, refinery exhaust, the usage of sulfur-containing fuels, and the aromatics industry. The cumulative ILCRs induced by PM-bound PACs in both petrochemical bases were low and were mainly contributed by the vehicle (and ship) traffic exhaust source.

Author contributions
Data availability All data generated or analyzed during this study are included in this published article and its supplementary information files.

Declarations
Ethics approval and consent to participate Not applicable.

Consent for publication Not applicable.
Competing interests The authors declare that they have no competing interests.