Polycyclic Aromatic Hydrocarbons in Atmospheric PM2.5 during Winter Haze in Huang-gang, Central China: Characteristics, Sources and Health Risk Assessment

Levels, compositions, sources and health risk of PM2.5-bound PAHs were measured at a regional monitoring site of Huang-gang in the winter of 2018. Samples were collected every morning, afternoon, evening and late night (LN). The average concentrations of PM2.5 and PAHs were 110.4 ± 48.3 μg m–3 and 25.6 ± 12.0 ng m–3, respectively. The concentration of PM2.5 was in the order of evening > morning > afternoon > LN, while PAHs concentration was in the order of evening > LN > morning > afternoon. According to PM2.5 concentration and its changing trend, the haze can be divided into three stages: early (low concentration, January 13–15, EHZ), medium (high concentration, January 16–22, MHZ) and late (concentration decreased rapidly, January 23–24, LHZ). A positive correlation (P < 0.01) was observed between PM2.5 and ∑16PAHs concentration only in MHZ. Besides, PAHs concentration was positively related to NOx concentration, CO concentration and temperature but negatively correlated with RH, independent of O3 concentration. The ratios of OC/EC indicated that the secondary aerosol had a great influence on the formation of haze. Three sources of PAHs including traffic emission, coal combustion and biomass combustion were extracted and quantified, accounting for 54.4%, 13.3% and 32.3% of PAHs, respectively. Potential source contribution function (PSCF) and concentration weighted trajectory (CWT) indicated the transmission of PM2.5 from Wuhan and local pollutant emission were crucial to the formation of haze. BaPeq concentration combined with ILCR values was used to evaluate health risk and the results showed this haze had potential health risk to both adults and children.


INTRODUCTION
As one of the most serious atmospheric environmental problems, haze not only reduces visibility, affects people's daily life, but also has a negative impact on human health and global climate change (Chang et al., 2009;Tan et al., 2009). Aerodynamic equivalent diameter less than 2.5 µm (PM2.5) has been proved to be directly related to the formation of haze (Cao et al., 2012). Due to its small particle size, PM2.5 can be suspended in the air for a long time, and its chemical composition and distribution will constantly change in time and space (Kim et al., 2015). PM2.5 is mainly composed of organic aerosol (OA), heavy metals, water-soluble ions and so on. As the main component of PM2.5, OA accounts for 20-90% of PM2.5 mass (He et al., 2001;Dan et al., 2004). Organic carbon (OC) and elemental carbon (EC) are two main forms of OA. OC, is one of the main causes of visibility reduction, and its sources are primary organic carbon (POC) emitted directly from fuel combustion and secondary organic carbon (SOC) generated by chemical from the main road of this city. The sampling site is a representative sampling point and the surrounding environment of sampling site is shown in Fig. 1.
Samples were collected throughout the haze period (from January 13 to January 24). Four samples were obtained every day at 06: 00-11:00, 11:30-16:30, 17:00-22:00 and 22:30-05:30, and totally 48 samples were obtained. A high-volume sampler (Wuhan Tianhong Environmental protection Industry, TH-1000C) with a flow rate of 1.05 m 3 per minute was used to collect particulate matter and PAHs by quartz fiber filters. In order to reduce the experimental error, the quartz fiber filters were baked in a muffle oven at 550°C for 4 hours before using. Samples collected were sealed at -10°C for storage.

PM2.5 Measurement and PAHs Analysis
After baking (550°C for 4 hours), the quartz fiber filters were kept equilibrium at a temperature of 23 ± 1°C and a relative humidity of 40% ± 5%. Afterthat, the quartz fiber filters were weighed by an electronic microbalance at least three times before sampling, and the weighing error did not exceed 0.05 mg. After that, the sampled quartz fiber filters were weighed under the same process. Thus, the mass difference of the filters before and after sampling is the mass of PM2.5 adsorbed on the filter membrane.
Concentrations of OC, EC were measured by an Organic/Elemental Carbon Analyzer (DRI model 2001, Atmoslytic Inc. USA) based on the Interagency Monitoring of Protected Visual Environments (IMPROVE) thermal/optical reflectance (TOR) protocol for pyrolysis correction. The analysis procedure of PAHs was the same as described in published paper (Zhang et al., 2019b). Briefly, a half of the circular quartz fiber filter was put into the Soxhlet extractor for extraction. About 140 mL of dichloromethane (DCM) was added immediately after adding five deuterated PAHs (naphthalene-d8, acenaphthene-d10, phenanthrene-d10, chrysene-d12, perylene-d10, o2si smart solutions, USA). The liquid was heated in a water bath at 45°C for 24 hours and then was concentrated to 5 mL on a rotating evaporator (Heidolph Laborota 4000, Germany). The glass column with silica gel and alumina (V:V = 6 mL:3 mL) was used to remove impurities in the extract with approximate 25 mL mixture of DCM and n-hexane (V:V = 2:3). The eluent was concentrated to about 1.5 mL on the rotating evaporator, and then transferred to a cell bottle for nitrogen blowing. The eluent was concentrated by nitrogen (purity ≥ 99.999%) to 200 µL. Finally, 5 µL internal standard substance hexamethylbenzene (HMB, 200 mg L -1 ) was added before instrumental

Sampling site
Sampling site analysis. Silica gel (100-200 mesh) and alumina (100-200 mesh) need to be baked at 180°C and 240°C for 24 hours to activate before experiment. After that, 3% pure water was added to silica gel and alumina to remove their activity.
Sixteen U.S. EPA priority PAHs (Table 1) were detected by gas chromatography-mass spectrometry (GC-MS, Agilent 6890N-5975). In order to ensure the accuracy of data and ensure samples are not contaminated during sampling and analysis, field blanks, process blanks and samples were detected together. The results showed that the concentration of blank PAHs was much lower than the detection limit of the method. The surrogate recoveries were in the range of 70%-120%, and the final results of PAHs concentration showed in Table 1.

Description of PCA-MLR
PCA-MLR has strong source analysis ability in the absence of source component spectrum. The main purpose of PCA is to explain most of the variation information in data with as few variables as possible. It is aimed to eliminate as much collinearity as possible through dimensionality reduction. A large amount of data information is simplified into a few representative influence factors by PCA. According to the load of the extracted factors on each component of PAHs, the sources of PAHs can be distinguished. Based on this, MLR is used to further determine the contribution rate of pollution sources. The equation for MLR analysis is In the formula, dependent variable Y denotes ∑16PAHs, "n" denotes the number of factors, Xi denotes the factor score variable and mi represents the coefficient of corresponding factor, the constant "b" denotes the unexplained residual variable information. Standardized the variable results and operate linear regression analysis again, the following equation is obtained 13.4 ± 4.79 PM2.5 concentration: μg m -3 . PAH concentration: ng m -3 . EHZ, MHZ, LHZ represent early (January 13-15), middle (January 16-22) and late (January 23-24) stage of haze, respectively.
In the formula, Ci represents MLR coefficient. The average contribution rate (Ni) of pollution source factor "i" can be expressed by equation

PSCF and CWT
HYSPLIT is a comprehensive model system developed by the US National Oceanic and Atmospheric Administration (NOAA) Air Resources Laboratory (ARL) and the Australian Meteorological Agency to analyze trajectories of airflow, diffusion, and sedimentation. It can be used to diagnose the source and transmission velocity of air mass, but it cannot determine the distribution of potential pollution sources and the level of pollution. PSCF and CWT are grid-based statistical methods based on the HYSPLIT model. They can be used in the TrajStat model to obtain the distribution of potential source regions and pollution source concentrations semi-quantitatively.
The source of PM2.5 and PAHs during the haze period can be distinguished spatially by PSCF. PSCF analysis method is a probability function that infers the spatial distribution of the possible source area by backward trajectory. The mesh covering more than 90% of the track area is established by software, and the end point of the track segment terminated in each cell (0.5° × 0.5°) is calculated to get the PSCF value of the mesh in the study area (Wei et al., 2019). The PSCF value is defined as: where i and j are on behalf of the latitude and longitude, respectively; nij is defined as the total number of endpoints that fall in the ij-th cell, and mij is the number of endpoints in the same cell that exceeded the threshold criterion. For PM2.5, the threshold criterion is the second-grade level of National Ambient Air Quality Standards of China (NAAQS, GB 3095-2012, 75 µg m -3 ). The higher PSCF values, the greater probability that the concentration is higher. When the average number of each grid of the trajectory end point (nave) is three times that of nij, the uncertainty of each cell can be reduced by introducing a weight function W(nij): The PSCF value can only reflect the probability that the concentration of the pollutant is higher than the threshold criterion, and it is difficult to distinguish the pollution levels of different potential regions. The CWT analysis method refers to assigning a weighted concentration to each grid by averaging the sample concentrations having trajectories associated with the grid cells . CWT is expressed as follows: where Cij represents the average weight concentrations in the ij-th cell, cl is the measured PAHs concentration observed on the arrival of trajectory l, and τijl is the number of trajectory end points in the ij-th cell associated with the cl sample. The weighting function described in Eq. (6) can also be used in the CWT analyses.

Carcinogenic Risk Assessment
As semi-volatile organic pollutants, PAHs encounter gas-particle phase distribution after being discharged into the atmosphere. Both the gas and particle phase PAHs can reach the alveolar epithelium through inhalation exposure and enter the circulatory system rapidly, and cause harm to human (Liu et al., 2019b). Because of the lack of gas-PAHs concentration data, the G/P distribution model is introduced to estimate PAHs in gas phase to assess the carcinogenic risk more accurately. The distribution model has been described in detail in previous study (Ma et al., 2019), and the equation is as following: where Kp is G/P partitioning coefficient of each PAH, CP and CG are concentrations of each PAH in gas and particle phase (ng m -3 ), PM is the concentration of particle matter (µg m -3 ). Individual PAH has different carcinogenic risk, and Bap is the most toxic and representative. The total carcinogenic concentration of PAHs can be assessed with BaP equivalent concentrations (BaPeq) (Gao et al., 2015). Here, BaPeq can be calculated with the formula: where PAHi is the total PAH concentration (G+P phase), TEFi is the toxic equivalency factor (TEF) for each PAH (Nisbet and Lagoy, 1992):0.001 for Nap, Acy, Ace, Flu, Ant, Fla, Pyr, 0.01 for Phe, BghiP, 0.1 for BaA, Chr, BbF, BkF, IcdP, 1 for BaP, DBA. The incremental lifetime cancer risks (ILCRs) model was used to calculate the exposure risks of PAHs, and the pathways of exposure risks include inhalation, ingestion and dermal contact. Parameters of ILCRs are showed in (Table 2) and the equations for estimating ILCRs are as following (Wang et al., 2011;Gong et al., 2019;Ma et al., 2020;Xing et al., 2020):

PM2.5 Levels and Composition
According to PM2.5 concentration and its variety trend (Fig. 2), the whole haze episode (WHZ) can be divided into three stages: early (low concentration, January 13-15, EHZ), medium (high concentration, January 16-22, MHZ) and late (concentration decreased rapidly, January 23-24, LHZ). PM2.5 concentration in MHZ was much higher than that in EHZ (about 2.33 times) and LHZ (around 1.75 times). Here, the sampling time period for obtaining four samples per day was specified as morning (06:00-11:00), afternoon (11:30-16:30), evening (17:00-22:00) and late night (LN,. The concentration of PM2.5 ranged from 30.7 µg m -3 to 196 µg m -3 , with an average concentration of 110.4 ± 48.3 µg m -3 in WHZ, which was 11 times of the World Health Organization (WHO) safe level (10 µg m -3 ), and 1.47 times the severe pollution level (75 µg m -3 , NAAQS, GB 3095-2012), indicating that Huang-gang was suffering from relatively serious air pollution. Compared with other cities, the concentration of PM2.5 was lower than those reported in northern China, such as Beijing , Liaocheng in Shandong Province (Liu et al., 2019a), whereas, higher than those in southern cities, like Guangzhou (Tan et al., 2011), Wuhan (Zhang et al., 2019b). The highest concentration of PM2.5 occurred in MHZ evening (16 January and 20 January). From the perspective of diurnal variation, the concentration of PM2.5 presented as evening > morning > afternoon > LN, and PM2.5 concentration at night was slightly higher than daytime, generally. Heavy traffic and coal combustion in the morning and evening might be the reasons for the high PM2.5 concentration (Yu et al., 2009). In addition, the lower boundary layer at night made it difficult for pollutants to diffuse, thus further increasing the concentration of PM2.5 . However, with the decrease of human activities at LN, the emission of pollutants would decrease, so the concentration of PM2.5 was lower in LN. The concentrations of OC and EC during the sampling period varied from 3.17 to 23.5 µg m -3 and 1.70 to 10.5 µg m -3 with the mean of 14.3 ± 4.95 µg m -3 and 5.63 ± 1.74 µg m -3 , respectively. The concentrations of OC and EC were much lower than that in Xi'an, Taiyuan, Chongqing (Ni et al., 2019) and were close to Tianjin (Wang et al., 2019). OC and EC averagely accounted for 12.9% and 5.10% of total measured PM2.5 concentration. Moreover, OC contributed 71.7% to total carbon (TC, sum of OC and EC), which indicated that OC was the predominant contributor to carbonaceous aerosols than EC. SOC can be calculated by the following formula (Grivas et al., 2012): where (OC/EC)min was the minimum ratio of OC to EC in the entire sampling period. The OC/EC ratios can identify the contribution of primary and secondary sources to carbonaceous aerosols, and the formation of secondary aerosols can be characterized by higher OC/EC ratios (greater than 2) (Peng et al., 2018;Wang et al., 2019). During the haze period, a total of 85.4% of samples OC/EC exceeded this value, while 50.0% of samples exceeded this value in LHZ. The proportion of SOC in OC decreased significantly in LHZ (Table 3), which indicated that secondary aerosols played a key role to the formation of haze in MHZ. From the perspective of intraday changes, higher SOC concentrations and higher SOC/OC values occurred at afternoon and evening, which was related to the fact that more residential cooking and heavy traffic can promote the formation of SOC (Stone et al., 2008;Yu et al., 2009;Xu et al., 2018). The average concentration of ∑16PAHs during haze episode was 25.6 ± 12.0 ng m -3 , which was similar to the annual average concentration of ∑16PAHs in Wuhan in 2014 (Zhang et al., 2019a), and was much lower than those in northern China, such as Liaocheng (Liu et al., 2019a), Jinan , higher than southern China cities, like Huangshi (Hu et al., 2018), Shanghai (Wei et al., 2019), Nanjing (Kong et al., 2015a), Nanchang . Furthermore, the concentration of carcinogenic PAHs (CARPAHs), including BaA, BbF, BkF, BaP, DBA, IcdP and BghiP, ranged from 1.59 to 38.4 ng m -3 , with a mean value of 12.8 ng m -3 , which was much higher than that in Shanghai (Wei et al., 2019). The average concentration of PAHs in EHZ was slightly higher than that in MHZ, and both of which were about twice than that in LHZ. A positive correlation (P < 0.05, R = 0.56) was observed between PM2.5 concentration and ∑16PAHs concentration only in MHZ, which suggested that PAHs and PM2.5 might come from the same source in MHZ. The concentration of PAHs at night was about 1.20 times higher than that at daytime and both the highest and lowest concentrations of ∑16PAHs occurred at LN. Additionally, the concentrations of ∑16PAHs presented as follows: evening > LN > morning > afternoon, which was different from PM2.5. The concentration of PM2.5 and PAHs can be affected by the amount of pollutants, the way of emission and meteorological conditions. In addition, as one of the components of PM2.5, PAHs accounts for only 0.2‰ of PM2.5 mass in this study, so PM2.5 is not the main factor causing the change of PAHs concentration. Moreover, PAHs concentration can be greatly affected by meteorological conditions. Wind speed, temperature and relative humidity can affect the gassolid distribution of PAHs. Among the CARPAHs, BaP was the representative compound of PAHs, and was considered as a good index for the whole PAHs carcinogenicity . The concentration of BaP was ranged from 0.09 to 5.18 ng m -3 , with a mean value of 0.96 ng m -3 . In all of the 48 samples, only 3 samples exceeded the daily limit (2.5 ng m -3 , NAAQS, GB 3095-2012). According to the number of aromatic rings, the concentration of ∑16PAHs presented as 4-ring > 5-ring > 6-ring > 3-ring > 2-ring in each time periods (Fig. 3(a)), which suggested ∑16PAHs was mainly concentrated in the middle molecular weight (MMW, 4-ring) and heavy molecular weight (HMW, 5-ring and 6-ring). MMW and HMW of PAHs accounted for 36.5% and 45.6% of total PAHs concentration respectively, which suggested the haze was mainly composed of PAHs with high molecular weight and low volatility. A significant positive correlation (P < 0.01) was also perceived between the concentration of ∑16PAHs and 4,5,6-rings PAHs during monitoring period, which suggested 4,5,6-rings PAHs contributed to the accumulation of PAHs. The highest PAHs concentration occurred at night, and was mainly concentrated in HMW. Heavy traffic was the main cause of this phenomenon, because HMW-PAHs were significantly derived from the emission of vehicle exhaust (Kong et al., 2015b;. Compared with other time periods, the proportion of 4-ring PAHs increased at afternoon, which might be related to the combustion of coal and biomass during lunch (Lin et al., 2015;Ren et al., 2017). The concentrations of PAHs in the morning and LN were higher, because the lower temperature (5.50°C for morning and LN, 7.80°C for afternoon) was favorable for the transition of PAHs from the vapour phase to the particle phase  and weaker solar radiation inhibited the photochemical decomposition of PAHs at LN (Duan et al., 2007). Compared to EHZ, the concentration of PM2.5 increased sharply in MHZ, but the concentration of PAHs was basically the same as that in EHZ, which was manifested in each ring of PAHs ( Fig. 3(b)). This may be due to that when PM2.5 increased to a light pollution level (PM2.5 > 75 µg m -3 ), organic matter fractions decreased probably during the transition and pollution (Zhang et al., 2019b). Therefore, more factors shuold be studied further.

Relative to Meteorological Condition and Other Air Pollutants
During the WHZ, meteorological conditions and other air pollutants, including relative humidity (RH), wind speed (WS), carbon monoxide (CO), nitrogen oxide (NOx) and so on, had a significant impact on PM2.5 and PAHs concentration (Table 4). Generally speaking, WS had a significant negative correlation (P < 0.01) with PM2.5 concentration and PAHs concentration, while RH had a significant positive correlation (P < 0.01) with PM2.5 concentration. Intensive winds can lead to less inversion situation and better dilution of pollutant, while fine particles combined with water in the high humidity air result in particles suspended in the air and not easy to dissipate. Different from the whole monitoring period (Fig. 4), PM2.5 concentration showed a significant negative correlation (P < 0.05, R = -0.73) with RH in LHZ, while the relation was not obvious in MHZ. When RH is low, the PM2.5 concentration will increase with the rise of RH. But when RH increased to a high level (rainfall exceeded 1 mm per hour), PM2.5 can be effectively removed from the air (Pu et al., 2011). There was also a significant positive correlation between NOx and CO and PAHs concentration, which indicated that PAHs mainly come from primary emission sources, such as   combustion source and traffic emission source, since NOx and CO were basically generated by primary emission sources (Ohara et al., 2007). PAHs concentration showed no obvious correlation with O3 during the whole monitoring period. Furthermore, PAHs concentration was positively correlated with T (P < 0.01) and negatively correlated with RH (P < 0.05) in LHZ. With the increasing of relative humidity (> 80%) and rainfall in LHZ, PAHs were easier to deposit together with PM2.5. In addition, the low temperature (0°C) caused by strong wind made it difficult for PAHs in the atmosphere to be adsorbed by PM2.5.

Source of PAHs
In order to identify the sources of PAHs during haze episode, isomer diagnosis ratio was used firstly to explain sources qualitatively. BaA/(BaA+Chr), Fla/(Fla+Pyr), BaP/BghiP and IcdP/(IcdP+BghiP) can be used to preliminarily determine the possible origin of PAHs (Yunker et al., 2002;Li et al., 2006). In the WHZ, PAHs bound to PM2.5 in ambient air mainly come from petrogenic sources, traffic emissions (e.g., vehicular exhaust), tar/coal combustion and biomass (e.g., wood) combustion (Fig. 5). The sources of pollutants in EHZ and MHZ were relatively more complex than those in LHZ. BaA was considered to be more readily degradable than its isomers (Chr) during atmospheric transport, so the ratio of BaA to BaA+Chr can be used to determine whether the air mass was locally emitted or regionally transported (Ding et al., 2007). The ratio of BaA to BaA+Chr greater than 0.50 indicated typical local emissions (Khalili et al., 1995). The ratio of BaA to BaA+Chr was concentrated in the range of less than 0.50 in this study, especially in LHZ, indicating that regional transport has an effect on haze episode.
Because of the uncertainty of isomer diagnosis ratio, PCA-MLR was used to analyze the types of PAHs pollution sources and their contribution rates more accurately (Yunker et al., 2002;Zhang et al., 2019a). PCA factors for EHZ, MHZ, LHZ and WHZ are presented with four, four, two and three factors, totally explaining 90.7%, 89.3%, 89.4% and 82.2% of the data variance, respectively (Table 5).

Fig. 5.
Four sources of PAHs, including petrogenic sources, traffic emissions, tar/coal combustion and biomass combustion, were identified using isomers diagnostic ratios. BkF were emitted from diesel exhaust (Rogge et al., 1993). Therefore, PC1 can be considered as a source of traffic emissions. Principal component 2 (PC2) explained 26.6% of the data variance with high loadings of Phe, Ant, which were markers for coal combustion source . Ant had been identified in coal combustion and coke production (Sofowote et al., 2008). Thus, PC2 was identified as coal combustion source. Principal component 3 (PC3) explained 10.1% of the data variance with high loadings of Acy and Ace. These LMW-PAHs were mainly generated at a relatively low combustion temperature of biomass, such as wood, grass (Yunker et al., 2002). Principal component 4 (PC4) explains 6.40% of the data variance with high loadings of DBA, which was related to thermal power plant that using coal as fuel (Fang et al., 2004). In addition, MLR was used to analyze the contribution rate of each principal component. It suggested that the main sources of PAHs in EHZ were traffic emissions (68.1%), coal combustion (18.5%) and biomass combustion (13.4%).
Furthermore, only two principal components were extracted at LHZ, PC1 and PC2 explain 61.0% and 28.4% of the data variance, respectively. PC1, with high loadings of Phe, Ant, Chr, BbF, BkF and BghiP, suggested that it was mainly from mixed sources of traffic emissions and coal combustion (Li et al., 2012). Therefore, the main sources of PAHs in LHZ were traffic emissions/coal combustion (80.0%) and biomass combustion (20.0%).
As can be seen from the above analysis, traffic emissions (TE), coal combustion (CC) and biomass combustion (BB) were the main sources of PAHs bound to PM2.5 throughout the haze period. PCA-MLR analysis of PAHs sources during the whole haze period also showed that TE (54.4%), CC (13.3%) and BB (32.3%) were the main sources. Compared with the EHZ, the contribution rate of TE decreased in MHZ, while the contribution rate of CC increased significantly. Therefore, the use of coal was one of the main factors for the aggravation of haze.

Geographic Origin
The mesoscale transport of pollutants in the atmospheric boundary layer is generally within 1000 meters, and the transport time is 2-3 days. Therefore, the HYSPLIT model was used to simulate the movement of air mass at an altitude of 500-meter in the past 72 hours (Xing et al., 2020), and each trajectory was obtained every 6 hours. TrajStat model was used to cluster the air mass trajectories to track the regional sources of air pollutants during haze. Cluster analysis of air mass trajectories during WHZ showed that four types of trajectories were obtained, of which 22.9% came from Kazakhstan and northwest China, 18.8% came from southwest China and 58.3% came from Hubei Province. Air masses trajectories in different periods of haze were also clustered (Fig. 6). The results showed that, on the one hand, the long-distance transport air masses from northwest China occupied 41.7% in EHZ, and decreased significantly in MHZ (7.14%), raised to 25.0% in LHZ, which indicated that the air masses from Northwest China had an obvious dilution effect on pollutants. On the other hand, the proportion of air masses from south Hubei Province was more than 60.0%, and the contribution rate of air masses from southwest China was more than 30.0% in MHZ, which indicated that air mass from south Hubei Province and southwest China had an aggravating effect on haze.
PSCF and CWT analysis results were shown in Fig. 7. PSCF analysis results of PM2.5 and PAHs showed that PSCF values were low in northwest and southwest China, which were far away from the sampling point, indicating that these regions were less likely to contribute to the haze. The higher PSCF values appeared in southern Henan Province, eastern Hubei Province and the junction of Hunan Province and Jiangxi Province, showing a north-south belt distribution. This phenomenon indicated that air masses from the south and the north played a key role in the formation and dissipation of haze in this region. Likewise, CWT analysis results were highly consistent with PSCF results. The concentration of PM2.5 and PAHs showed strip-shaped distribution in the region. The maximum value of CWT corresponding to PM2.5 appeared in Huanggang City and Wuhan City, indicating that the transmission of PM2.5 from Wuhan and the emission of local pollutants had a great effect on the formation of haze.

Health Risk Assessment
Gas-phase PAHs were estimated to assess the carcinogenic risk more accurately, and the BaPeq concentration determined by TETs was showed in Fig. 8(a). The average concentration of BaPeq was4.08 ± 2.22 ng m -3 , which was higher than the daily limit (2.5 ng m -3 , GB3095-2012). The maximum BaPeq concentration (10.2 ng m -3 ) appeared at evening, while the minimum (1.16 ng m -3 ) appeared at LN, and about 77.0% of the samples exceeded the daily limit. From the perspective of time series, BaPeq concentration followed descending order: evening > LN > morning > afternoon in the WHZ, whereas, the lowest temperature led to the lowest concentration of gasphase PAHs at evening. On the whole, BaPeq concentration was at a relative risk level.
The ILCR analysis result was shown in Fig. 8(b). The average TILCR value of adult was 3.69 ± 2.01 × 10 -5 and 15.0 ± 8.16 × 10 -6 for children, which indicated potential carcinogenic risk according to USEPA. From the perspective of time series, ILCR results were shown as evening > LN > morning > afternoon for both adults and children, and adults have a higher cancer risk during haze. Besides, as compared with inhalation, the carcinogenic risk of exposure to PM2.5 mainly came from ingestion.

CONCLUSIONS
In this study, 48 samples, collected during haze in Huang-gang, were analyzed to explain levels, compositions, sources and cancer risk of PM2.5-bound PAHs. The average concentrations of PM2.5, OC, EC and PAHs were 110 ± 48.3 µg m -3 , 14.3 ± 4.95 µg m -3 , 5.63 ± 1.74 µg m -3 and 25.6 ± 12.0 ng m -3 , indicating that Huang-gang was suffering from relatively serious air pollution. The concentration of PM2.5 presented as: evening > morning > afternoon > LN, while PAHs concentration showed evening > LN > morning > afternoon. A positive correlation (P < 0.05) was observed between PM2.5 concentration and ∑16PAHs concentration only in MHZ, which suggested PAHs and PM2.5 had the same sources in MHZ. The SOC concentration was higher at afternoon and evening, indicated that more residential cooking and heavy traffic can promote the formation of haze. The proportion of SOC in OC was significantly lower in LHZ than in EHZ and MHZ, which indicated that secondary aerosols had greatly influence to the formation of haze. Furthermore, MMW and HMW of PAHs accounted for 36.5% and 45.6% of total PAHs concentration respectively, which suggested that the haze was mainly composed of PAHs with high molecular weight and low volatility. PM2.5 concentration was negatively correlated with WS, and was positively correlated with RH in EHZ, negatively in LHZ. Additionally, PAHs concentration was positively correlated with NOx, CO and T, negatively correlated with RH, but showed no correlation with O3. Besides this, the main sources of PAHs in WHZ were traffic emission (54.4%), coal combustion (13.3%) and biomass combustion (32.3%), also the increase in the use of coal had an aggravating effect on haze in MHZ. Air masses from the south and the north played a key role in the formation and dissipation of haze in this region. Moreover, the transmission of PM2.5 from Wuhan and the emission of local pollutants had a great effect on the formation of haze. Gasphase PAHs were estimated to assess the carcinogenic risk more accurately, and the ILCR assessment showed this haze had potential health risks to both adults and children, the carcinogenic risk of exposure to PM2.5 mainly came from ingestion.

ACKNOWLEDGMENTS
This study was financially supported by the National Key Research and Development Program of China (NO.2017YFC0212603). We also thank Huang-gang Environmental Protection Bureau for their helps in sample collection.

DISCLAIMER
Reference to any companies or specific commercial products does not constitute an interest.