Pollution Characterization, Source Identification, and Health Risks of Atmospheric-Particle-Bound Heavy Metals in PM10 and PM2.5 at Multiple Sites in an Emerging Megacity in the Central Region of China

A total of 588 daily PM filters were collected at five sites in Zhengzhou, and the mass concentrations and sources of the elements were analyzed. The health risks and source regions of the particles and toxic elements were also estimated. The results indicated severe PM2.5 and PM10 pollution, especially at traffic sites. Additionally, the PM10-bound As far exceeded the Chinese standards. Although the total elemental levels were relatively low at the rural site, they were high at the GY site. High levels of crustal elements were also observed at the SSQ and HKG sites. Seasonal-variation analysis revealed that the crustal elements, more abundant in the PM10, occurred at high levels in spring; the combustion-source elements, more abundant in the PM2.5, caused significant pollution in winter; and the elemental concentrations were low in summer. The coefficients of divergence for the PM2.5 were slightly higher than those for the PM10. Vehicles, industry, coal combustion, oil fuel, dust, and biomass burning were important sources of the PM-bound elements. Although the ZM site was characterized by low traffic and high contributions from biomass burning and dust emission, the HKG site featured a high proportion of emissions from traffic sources, and the SSQ site was also highly affected by vehicular pollution. Whereas elements in the PM2.5 largely originated in combustion sources, those in the PM10 received greater contributions from dust sources. The levels of As and Ni posed intolerable carcinogenic risks (CR) and, along with that of Pb, also demonstrated significant non-CR risks. Children were more sensitive than adults to these risks, and the daily intake pathway demonstrated the highest CR and hazard index (HI) values. Obvious differences in the CR and HI values were detected between the various sites, suggesting the necessity of multiple-site studies for health risk assessment. Jiyuan, Jiaozuo, Xuchang, and Zhoukou; Pingdingshan and Nanyang; and Jiyuan, Jiaozuo, Xinxiang, Anyang, and Kaifeng were the main potential sources of PM2.5, PM10, and As, respectively.


INTRODUCTION
Over the past three decades, the rapid economic development, urbanization, and industrialization in China have caused atmospheric pollution to become a common phenomenon in megacities, and the primary pollutant is particulate matter (PM) (Ji et al., 2014). PM 10 (aerodynamic diameter ≤ 10 µm) and PM 2.5 (aerodynamic diameter ≤ 2.5 µm) have been attracting considerable attention in related research fields because of their important impacts not only on global and regional climate changes (Bytnerowicz et al., 2007), as well as reduced visibility (Moosmüller et al., 2009), but also in human cardiovascular diseases and wheezing (Brunekreef and Forsberg, 2005;Pope and Dockery, 2006;Bell, 2012;Dunea et al., 2016).
Toxic elements, such as As, Pb, Ni, Cd, Cu, Zn, and Cr, are important trace components in PM 2.5 and PM 10 with considerably higher concentrations in urban areas than natural background levels due to anthropogenic activities Okuda et al., 2008;Fang et al., 2010;Charlesworth et al., 2011). As a result, for urban residents, relatively high potential risks to human health are caused by these elements in atmospheric particulates. For example, the multi-element risk of PM 2.5 through inhalation exposure Jiang et al.,Aerosol and Air Quality Research,xxxx 3   where j and k represent two sampling sites; x ij represents the mean concentration of element i at site j; p is the total number of elements. When CD jk approaches zero, the result suggests that the two sites are similar for the chemical species, whereas when CD jk approaches one, the two sites significantly differ. According to a previous study (Contini et al., 2012), CD lower than 0.2 is selected for representing a relatively similar chemical composition.

Source Identification
In order to qualitatively evaluate contributions of different sources, i.e., anthropogenic sources and crustal origins, EFs are calculated as follows (Mason, 1966;Chao and Wong, 2002): where C is the concentration of trace element. Al is selected as the reference element (Hsu et al., 2016). (C/Al) PM and (C/Al) Crust are the ratios of trace elements to Al in each PM sample and the upper continental crust, respectively, and the data of average crustal abundances are shown in a study on the element background values of Chinese soil (Wei, 1990). EF > 10 indicates that the element was generated from an anthropogenic source (e.g., coal combustion and vehicle emission); when EF is approximately equal to one, crustal origin is indicated (Nolting et al., 1999). Pearson's CA and cluster analysis, two of the most popular methods, were chosen to detect linear relationship (Robin et al., 2013) and source contribution without source profiles (Viana et al., 2008) of trace elements in PM 2.5 and PM 10 at multiple sites through IBM SPSS for Windows, Version 22.0. Moreover, according to the species from the same source with similar characteristics (Ny and Lee, 2011), PCA was conducted for source identification of elements in PM by SPSS 22.0.

Health Risk Assessment Exposure Assessment
Inhalation, ingestion, and dermal absorption are the main pathways of airborne toxic element exposure for local residents, i.e., children (< 15 years) and adults, divided by different respiration and behaviors (Hu et al., 2012). According to US EPA (1989,2004,2009), exposure concentration (EC; µg m -3 ), chemical daily intake [CDI; mg (kg d) -1 ], and dermal absorption dose [DAD; mg (kg d) -1 ] of toxic elements in PM were calculated as the following equations. All the relative parameters were obtained from the official website of US EPA, unless otherwise specified. where C is the upper bound of the 95% confidence limit of the toxic elements in PM, µg m -3 or mg kg -1 ; ET is the exposure time, 6 h d -1 ; EF is the exposure frequency, 350 d year -1 ; ED is exposure duration, 24 and 6 years for adults and children, respectively; AT 1 is the average time, AT 1 (for carcinogens) = lifetime years × 365 d year -1 × 24 h d -1 , AT 1 (for noncarcinogens) = ED year × 365 d year -1 × 24 h d -1 ; lifetime is 74 years, which is the life expectancy in Henan (National Health and Family Planning Commission of the People's Republic of China, 2013); IngR is the ingestion rate, 100 and 200 mg d -1 for adults and children, respectively; BW is the body weight, 59 National Bureau of Statistical of China, 2015) and 15 kg for adults and children, respectively; AT 2 is the average time, AT 2 (for carcinogens) = 74 years × 365 d year -1 , AT 2 (for noncarcinogens) = ED year × 365 d year -1 ; CF is the conversion factor, 10 -6 kg mg -1 ; SA is the surface area, 5700 and 2800 cm 2 for adults and children, respectively; AF is the adherence factor, 0.07 and 0.2 mg (cm 2 d) -1 for adults and children, respectively; and ABS is the absorption fraction, 0.001, 0.03, and 0.01 for Cd, As, and other toxic elements (Hu et al., 2012;US EPA, 2017), respectively.

Risk Assessment
CR and hazard quotient (HQ) of toxic elements in PM were calculated for carcinogenic and noncarcinogenic risk assessment, respectively. The relative formulas are as follows (US EPA, 1989, 2004, 2009: where all the upper parameter values for different elements were chosen according to US EPA (2017); IUR is the inhalation unit risk, (µg m -3 ) -1 ; SF O is the slope factor, [mg (kg d) -1 ] -1 ; GIABS is the gastrointestinal absorption factor; RfC i is the inhalation reference concentrations, mg m -3 ; RfDo is the oral reference dose, mg (kg d) -1 ; and HI is the hazard index for the three exposure pathways.

Trajectory Calculation
The hybrid single-particle Lagrangian Integrated Trajectory (HYSPLIT) model is a useful tool for assessing potential sources of pollutants. In this study, the model was used to obtain 48-h backward trajectories of air masses with an altitude of 500 m arriving at the urban site SSQ (34°46′24″N, 113°44′2″E) in Zhengzhou and with four trajectories each day (00:00, 06:00, 12:00, and 18:00). Then, the potential source contribution function (PSCF) model, based on the HYSPLIT results, was used to identify the potential source regions of PM 2.5 , PM 10 , and toxic elements. The study region is equally divided into i × j grid cells, and the PSCF value in the ij th cell is calculated by m ij /n ij (Ashbaugh et al., 1985;Jeong et al., 2011). n ij and m ij are the number of trajectories whose endpoints fall in the ij th cell and the concentrations higher than the pollution criterion, respectively. According to the previous studies (Hsu et al., 2003;Liao et al., 2017), 75, 150 µg m -3 , and the average concentration of As were chosen as the pollution criterion for PM 2.5 , PM 10 , and toxic elements, respectively. The details of these methods were described by Ashbaugh et al. (1985) and Hopke et al. (1995). Uncertainties exist due to the small values of n ij . Thus, to reduce the uncertainties, an empirical weight function, W ij , was defined in the following formula. Then, the PSCF results were multiplied by W ij (Polissar et al., 1999;Zhang et al., 2013
From a seasonal perspective, the mean concentrations of PM among the sampling sites in Zhengzhou are exhibited in Table 2. Obvious seasonal variations were observed, with the highest average concentrations in winter (PM 2.5 : 235 ± 126 µg m -3 ; PM 10 : 308 ± 162 µg m -3 ) and the lowest average values in summer (PM 2.5 : 48 ± 17 µg m -3 ; PM 10 : 62 ± 18 µg m -3 ). The reasons for these variations of PM include not only the different sources of emissions, but also the variant meteorological conditions in four seasons. For example, a great deal of coal is combusted for heating in the northern cities in China in winter. Moreover, frequent stagnant meteorological conditions occur in this season, causing high PM pollution. In summer, generally, the planetary boundary layer (PBL) development is enhanced by high temperature and higher PBL influences the vertical dispersion of the pollutants (Yang et al., 2015).

Elemental levels and Comparability among the Different Sites Elemental Levels among the Different Sites
The mean concentrations of 21 elements in PM in Zhengzhou are summarized in Figs. 3, S2 and S3. In general, among the sampling sites, these concentrations of elements in PM 2.5 and PM 10 vary substantially. Crustal elements, i.e., Ca (from 1653 ± 1229 ng m -3 to 4935 ± 6279 ng m -3 and from 4357 ± 2110 ng m -3 to 8949 ± 4670 ng m -3 , respectively), Si (from 1218 ± 949 ng m -3 to 2925 ± 3298 ng m -3 and from 3142 ± 1633 ng m -3 to 5138 ± 3420 ng m -3 , respectively), Fe (from 766 ± 502 ng m -3 to 2015 ± 2183 ng m -3 and from 1900 ± 972 ng m -3 to 3516 ± 2186 ng m -3 , respectively), K (from 1437 ± 1544 ng m -3 to 1728 ± 1577 ng m -3 and from 1712 ± 1143 ng m -3 to 2424 ± 1682 ng m -3 , respectively) and Al (from 523 ± 395 ng m -3 to 1408 ± 1240 ng m -3 and from 1378 ± 789 ng m -3 to 2326 ± 1210 ng m -3 , respectively) were the most abundant. According to the previous studies (Han et al., 2010;Li et al., 2017), these crustal element concentrations were considerably higher than those in Beijing, Shenyang, and Anshan in North China; meanwhile, the ratio of ∑crustal elements/PM (i.e., proportion of the five total crustal elements in PM) was approximately in the range of 7-10% and 9-13% for PM 2.5 and PM 10 , respectively, in Zhengzhou and also much higher than those in Beijing (3-4%), Shenyang and Anshan (both approximately 1%). The elements associated with combustion, i.e., S (from 3150 ± 2171 ng m -3 to 3965 ± 3205 ng m -3 and from 3502 ± 3365 ng m -3 to 5067 ± 4088 ng m -3 , respectively), and Cl (from 1051 ± 1268 ng m -3 to 2981 ± 2502 ng m -3 and from 1259 ± 1493 ng m -3 to 3160 ± 2681 ng m -3 , respectively) were also plentiful in PM. The concentrations of S associated with PM 2.5 in Zhengzhou were far beyond those in Beijing (from 1.1 ± 1.1 µg m -3 to 1.3 ± 1.4 µg m -3 in PM 2.5 ) , with the former ratio of S/PM 2.5 (approximately 3-4%) also higher than the latter (approximately 1-2%). These results suggested that dust and combustion sources played more important roles in elemental levels of PM in Zhengzhou. Na, Mg, Zn, Ti, Pb, Mn, and Ba were present in moderate quantities, and Cu (except at the GY site), Sb, As, Se, V, Cr, and Ni were sparse in PM, with average concentrations almost lower than 50 ng m -3 among the sites. Though the toxic elements accounted for only a small fraction of PM, they should also be paid enough attention for their adverse health effects, especially for As, with all mean concentrations in PM 10 (from 16 ± 15 ng m -3 to 28 ± 28 ng m -3 ) far exceeding the Chinese NAAQS (annual As: 6 ng m -3 ) among the sampling sites. Table 2 in PM 2.5 and PM 10 , respectively. Generally, the total elemental levels at the ZM site, defined as the rural site, are relatively low; in contrast to the variation tendency of PM, the total elemental concentrations at the GY site were high, especially Cl, Zn, Pb, and Cu, suggesting the important effects from combustion and vehicular sources Bhangare et al., 2011;Bozlaker et al., 2013;Bozlaker et al., 2014). High levels of crustal elements were observed at the SSQ and HKG sites, suggesting dust as one of main sources of elements in PM. Seasonal variations of element concentration were evident in Zhengzhou. Overall, among the sites, crustal elements, i.e., Ca, Si, Fe, K, and Al, demonstrated high levels in spring while elements from combustion presented relatively serious pollution in winter, and the elemental concentrations were low in summer. These phenomena are ascribed to the combined contribution of emissions and meteorological conditions. For example, lack of rain and high-speed winds are the typical spring climatic characteristics in North China  that result in increasing crustal element levels in PM. By comparing the elements with different diameters, crustal elements were found to be more abundant in PM 10 , with the sum concentration of Ca, Si, Fe, K, and Al accounting for 63-73% of the total elements among the five sites, which were higher than those in PM 2.5 (50-68%). However, elements from combustion, i.e., S, Cl, Pb, As, and Se were more gathered in PM 2.5 , with an average ratio (element in PM 2.5 /element in PM 10 ) from 77% (As) to 87% (Se).

Comparability among the Different Sites
The CD values of elements in PM 2.5 and PM 10 of the different sampling sites in this study are shown in Fig. 4. Among the sampling sites, the CD values obviously differed and ranged from 0.11 to 0.37 for PM 2.5 and 0.09 to 0.41 for PM 10 during four seasons, which were almost higher than the annual CD values. On the whole, the CD values for PM 2.5 are slightly higher than those for PM 10 from a seasonal perspective (except autumn), which may be due to the more complicated sources for fine particles than those for coarse particles. The CD values are not only related to discrepant spatial distribution of emission sources at the five sites but also attributed to different meteorological conditions in four seasons. Moreover, variant emission levels of elements have an effect on CD values. For example, a series of measures, including factories closed and production suspended, were conducted in the summer by the local government (Zhengzhou Municipal Environmental Protection Bureau, 2016). Measures were carried out to cooperate with the regional supervision centers, the organization set up by the Ministry of Environmental Protection of the People's Republic of China to monitor the implementation of environmental regulations.

Sources of Elements
The indicatory elements for various major sources, i.e., dust, coal combustion, traffic emission, industrial emission, oil fuel, and biomass burning, in previous studies are listed in Table 3. EFs, Pearson's CA, cluster analysis, and PCA were conducted to identify the main sources of elements in PM.

Enrichment Factors of Elements
The calculated EFs of elements in PM among the sampling sites in this work are displayed in Fig. 5. Na, Sb, Pb, Zn, Cu, and As were significantly enriched at all sites with high EFs (PM 2.5 : 60-12,125; PM 10 : 44-8,032), indicating anthropogenic sources. In the previous study, Na is mainly derived from coke-making, and cold-forming and hot-forming processes in iron and steel industries (Tsai et al., 2007). Sb is emitted from industries, e.g., the steel or petrochemical industry , and also from vehicular emissions (Charlesworth et al., 2011). Smelting and coal combustion processes are considered to be the primary sources of Pb and As Bhangare et al., 2011). Zn and Cu are probably generated from abrasion of brake linings and tire tread wear (Bozlaker et al., 2014). Therefore, industries, vehicles, and coal combustion are the main anthropogenic sources of elements in PM in Zhengzhou. The EFs of Si, Mg, and Ti are close to 1, indicating that crustal origin is also an important source of elements in aerosol (Nolting et al., 1999). The EFs of other elements, i.e., K, Ca, Mn, Cr, Ba, V, Fe, and Ni, are between 1 and 10, suggesting the emission from anthropogenic and natural sources. For example, K is associated with dust (Jiang et al., 2018a) and biomass burning sources (Silva et al., 1999).

Pearson's Correlation and Cluster Analysis
Correlation of different elements is an important basis of source identification that can help to confirm and obtain interpretation of PCA results. Table 4 shows the Pearson's CA of elements in PM among the sampling sites in Zhengzhou. A significant correlation was observed among Na, Mg, Al, Si, Ca, Fe, and Mn in PM 2.5 and PM 10 during the sampling period at the five sampling sites, suggesting that these elements were probably emitted from a common source. As and Se are highly correlated (0.51-0.88 for PM 2.5 and 0.53-0.88 for PM 10 ), indicating a similar source. High correlation coefficients were observed not only between Zn and Cu but also between Ni and V, demonstrating similar sources.
The cluster analysis results of elements in PM are shown in Figs. S4 and S5 in Supplemental Materials. Three clusters of elements in PM 2.5 and PM 10 at the sampling sites, except for PM 2.5 at HKG and XM sites (four clusters), were chosen. Generally, Cluster 1 is characterized by Ni, Cr, V, Se, As, Cu, Sb, Mn, Na, Pb, Ti, Mg, Ba, and Zn;  Tsai et al., 2007;Zhang et al., 2009;AEA, 2011;Bhangare et al., 2011;Taiwo et al., 2014 S, Cl, As, Se and Pb Smelting and coal combustion Zhang et al., 2009;Bhangare et al., 2011 Zn, Cu and Ba Brake linings and tire tread wear Garg et al., 2000;Bozlaker et al., 2013;Bozlaker et al., 2014 Cu, Zn, Sb, V, Ni, Cr, Pb, Cd andCl Vehicular emission Cadle et al., 1999;Sternbeck et al., 2002;Fang et al., 2006;Viana et al., 2006;Pan et al., 2013 Ni, V and Cr Oil fuel AEA, 2011 K, Ba and Cl Biomass burning Argyropoulos et al., 2013;Wang et al., 2016 V, Ni, Sb, Cr Aircraft Cluster 2 contains partly or entirely crustal elements, i.e., Al, Fe, Si, Ca, K; other elements are distributed to Clusters 3 and 4 (if any). Similar clusters of elements exhibit good correlations, suggesting common sources . The sources can be identified by fingerprint elements (Table 3). The results suggested that vehicular emission, industrial emission, coal combustion, and oil fuel were probable important sources of PM-bound elements in Zhengzhou. In addition, dust and biomass burning also affected elemental levels in PM.

Principal Component Analysis
In this study, PCA with varimax rotation was conducted for identifying the sources of elements in PM by using the software (SPSS 22.0), and the retention of principal components with eigenvalues greater than 1.0 was used to identify the main pollutant sources (Winner and Cass, 2001). Tables 5 and 6 present the PCA results of elements in PM 2.5 and PM 10 in each sampling site, respectively. Elemental loadings higher than 0.6 are in bold and considered to be important.            At ZM sampling site (rural site), for PM 2.5 , Factor 1 accounts for 31% of the total variance of the data, with high loadings of Mg, Al, Si, Ca, Ti, Cr, Mn, and Fe, representing contributions from dust (Jiang et al., 2018a) and industrial emission (Taiwo et al., 2014); Factor 2 accounts for 27% of the total variance in the dataset, with high loadings of Cl, K, Cu, Sb, Ba, and Pb, which are representative of biomass burning and vehicular emission  (Fang et al., 2006;Pan et al., 2013); Factor 3 (16%) is coal combustion (Bhangare et al., 2011), with a high content of S, As, and Se; Factor 4 (9%) is dominated by Zn, which is reported to originate primarily from the industrial emission (AEA, 2011). For PM 10 , Factor 1 accounts for 31% of the total variance, with high loadings of Mg, Al, Si, Ca, Ti, Ni, Cr, Mn, and Fe that came from dust and industrial emission (Taiwo et al., 2014;Jiang et al., 2018a); Factor 2 accounts for 28% of the total variance in the data and has high loadings of Na, Mg, Al, Cl, K, V, Cu, Sb, and Ba, suggesting that dust (Jiang et al., 2018a) and vehicular emission (Charlesworth et al., 2011) are the major contributors; high loadings of S, Cl, As, Se, and Pb are shown on Factor 3 (21%), considered to originate from coal combustion (Bhangare et al., 2011).
At the HKG sampling site (traffic site), for PM 2.5 , 49% of the total variance was observed in Factor 1, contributed by dust (Jiang et al., 2018a) and traffic emission, i.e., aircraft and vehicle sources (Charlesworth et al., 2011;Ren et al., 2012), because of the high loadings of Na, Mg, Al, Si, Ca, Ti, V, Ni, Sb, Cr, Mn, and Fe; Factor 2, with a high content of S, Cl, As, Se, and Pb, represents the contribution of coal combustion (Bhangare et al., 2011); Factor 3 (12%), with high loadings of K and Ba, represents biomass burning (Argyropoulos et al., 2013) in view of Cl loading; and Factor 4 (8%), with a high content of Cu and Zn, represents vehicular emission . For PM 10 , Factor 1 (37%) is commonly related to dust (Jiang et al., 2018a) and industrial emission (Taiwo et al., 2014), characterized by high loads of Na, Mg, Al, Si, K, Ca, Ti, V, Ni, Sb, Ba, Mn, and Fe; Factor 2 (29%) is identified as coal combustion (Bhangare et al., 2011) and industrial emission (AEA, 2011), with high loadings of S, Cl, Cu, As, Se, and Pb; and Factor 3 (17%), with a high content of Ni and Cr, represents oil fuel (AEA, 2011).
Given limited space, the details of PCA results in the other two urban sites are not shown with similar indicators. At the XM sampling site, for PM 2.5 , Factor 1 (dust and vehicular emissions), Factor 2 (coal combustion and industrial emission), Factor 3 (biomass burning and vehicular emission), and Factor 4 (industrial emission) contribute 41%, 22%, 21%, and 7% of the variance, respectively; for PM 10 , Factor 1 (dust and vehicular emissions), Factor 2 (coal combustion and industrial emission), and Factor 3 (biomass burning and industrial emission) contribute 43%, 26%, and 18% of the variance, respectively. At the SSQ sampling site, for PM 2.5 , Factor 1 (dust and vehicular emission) and Factor 2 (coal combustion and industrial emission) account for 56% and 26% of the variance, respectively; for PM 10 , Factor 1 (dust and vehicular emission), Factor 2 (coal combustion and industrial emission), and Factor 3 (biomass burning) contribute 44%, 24%, and 14% of the variance, respectively.
In the comparison of the source identification results among the five sites, the source contributions were found to be consistent with the category of sites. For example, ZM site is the rural site with low traffic effect. The main results included relatively high contributions of biomass burning, which is attributed to the use of straw as fuel for cooking and heating in winter, dust emission, and poor dust control measures in the country. HKG site is located near Xinzheng International Airport, and therefore, its traffic sources (i.e., aircraft and vehicles) contribute high proportion. SSQ site is an urban site with high traffic and is also affected highly by vehicle emission. Moreover, for comparison of PM 2.5 and PM 10 sources, in brief, combustion sources, including coal combustion, vehicular emission, and biomass burning, played more important roles in elements in PM 2.5 , whereas dust source contributed more to PM 10 -bound elements.

Health Risks Posed by Toxic Elements
The health risk values, including CR, HQ, and HI, of toxic elements in PM 2.5 and PM 10 through the inhalation, dermal absorption, and daily intake pathways are calculated in this study. For CR, 1 × 10 -6 < CR < 1 × 10 -4 indicates a tolerable risk for regulatory purposes, and CR > 1 × 10 -4 indicates intolerable risk (US EPA, 1989). Moreover, for noncarcinogenic risk, HQ > 1 and HI > 1 suggests a significant risk of a single element and total toxic elements (US EPA, 1989). The relative detailed data of health risk evaluation from toxic elements in PM through the three exposure pathways are presented in Tables S1-S3 in Supplemental Materials, and the carcinogenic and noncarcinogenic risks among the sampling sites in Zhengzhou are exhibited in Tables 7, S4 and S5. Meanwhile, the total CR and HQ of toxic elements in PM 2.5 and PM 10 are summarized in Fig. 6.
Overall, the CR values of As, Pb, and Ni were all higher than 1 × 10 -6 , especially for As (PM 2.5 : 1.2 × 10 -4 -2.9 × 10 -4 ; PM 10 : 1.0 × 10 -4 -3.2 × 10 -4 ) and Ni (PM 2.5 : 6.8 × 10 -5 -1.9 × 10 -4 ; PM 10 : 6.3 × 10 -5 -1.8 × 10 -4 ), with CR values exceeding 1 × 10 -4 , respectively. This result suggested that As and Ni caused intolerable risks and Pb showed tolerable risk in PM 2.5 and PM 10 in Zhengzhou. According to the previous study (AEA, 2011), metal production, and public electricity and heat production are the major sources of As; combustion of heavy fuel oil is the major source of Ni. Therefore, the local regulatory  Among the three exposure pathways, the daily intake is the limiting one. Generally, the highest CR of As, Pb, and Ni existed through the daily intake pathway, especially for As, with CR values ranging from 9.8 × 10 -5 to 2.7 × 10 -4 in PM 2.5 and from 8.6 × 10 -5 to 3.0 × 10 -4 in PM 10 . The noncarcinogenic risk of all toxic elements, i.e., HI, was also the highest through the daily intake pathway, with the values ranging from 0.9 to 21.4 in PM 2.5 and from 1.9 to 18.1 in PM 10 , which are far beyond the limit (HI = 1). Moreover, in this pathway, Sb, As, and Pb exhibited high risks, with HQ values all exceeding 1 for children, thereby also indicating significant risks. By contrast, CR and HQ values from toxic elements in PM through inhalation and dermal absorption exposure were relatively low. However, significant noncarcinogenic risks still existed because of HI > 1, especially for children.
For comparison among the five sites, the health risks differed with various elemental concentrations. In general, the total CR values from As, Pb, and Ni in PM 2.5 ranged from 3.5 × 10 -4 (HKG) to 4.9 × 10 -4 (GY) and 2.0 × 10 -4 (HKG) to 2.8 × 10 -4 (GY) for children and adults, respectively; the CR values in PM 10 varied from 3.2 × 10 -4 (HKG) to 5.0 × 10 -4 (SSQ) and 1.9 × 10 -4 (HKG) to 2.8 × 10 -4 (SSQ) for children and adults, respectively. These results demonstrated intolerable CR from the three elements in PM at all the sites in Zhengzhou and thus should be given more attention. Moreover, the relatively obvious differences of CR at the sites, with values, i.e., the average ratio of difference between the lowest and highest risk, of 29% and 35% for PM 2.5 and PM 10 . For noncarcinogenic risks, all total HI (i.e., ∑HI) of the sites were higher than 1, suggesting significant risks. For PM 2.5 , the total HI values ranged from 16.6 (ZM) to 24.9 (GY) and 2.0 (ZM) to 4.1 (GY) for children and adults, with difference ratios of 34% and 51%, respectively. For PM 10 , the total HI values were in the range of 18.2 (HKG)-25.4 (SSQ) and 3.3 (HKG)-4.5 (SSQ) for children and adults, respectively, both with difference ratios of 28%. Therefore, data at multiple sites are necessary for health risk assessment in the study area, especially for a large region. Fig. 7 shows the 48-h back trajectories of air masses arriving in Zhengzhou during the sampling periods, with 1140 transport trajectories. Obviously, four types of air mass clusters influenced the pollution levels of the sampling site. Two northwestern airflows with long trajectories occurred. The first one was mainly from Kazakhstan, Xinjiang, or Gansu, passing through the northern part of Qinghai, Shaanxi, Shanxi, and Northwestern Henan. The second airflow came from Russia, Mongolia, or Inner Mongolia, then by way of the Beijing-Tianjin-Hebei region and northern Henan. These areas, i.e., Mongolia, Inner Mongolia, Xinjiang, Gansu, and Shaanxi, were commonly covered by the Gobi Desert and grasslands. The long-range transport brought the mineral aerosols, mixing with the local emission and aggravating PM, especially crustal materials and pollution levels, and changing the ratio of Ca to Al (Zhang et al., 2003), used to determine source area of dust, in the PM samples in this study area. Beijing-Tianjin-Hebei, characterized by huge consumption of coal and coal combustion as the predominant source of aerosols (Yao et al., 2009), is always one of the most deteriorated regions in China (Tao et al., 2016). Therefore, the second direction airflow probably increased the PM and elements relative to coal burning, i.e., S, Cl, As, Se, and Pb (Bhangare et al., 2011) concentrations, especially in winter, i.e., high emission season because of extra coal consumption for central heating, including the local emission in Zhengzhou and the emission from the airflow passing regions. The medium to long distance transport covered the eastern regions of Zhengzhou, i.e., the Yellow Sea, Jiangsu, Northern Anhui, Southern Shandong, and Eastern Henan; these inland areas featured flourishing agriculture (Kang et al., 2016). Hence, these trajectories from the east likely influenced the pollution level of PM, Na, Cl, and elements emitted from biomass burning, i.e., K and Ba (Argyropoulos et al., 2013;Wang et al., 2016), especially in harvest season. The shortdistance transport from the southern areas of Zhengzhou covered Hunan, Hubei, and southern Henan. Generally, these regions present mitigating PM levels (Tao et al., 2016), suggesting possible decreasing PM and elemental values in this study region due to dilution function.

Source Regions of PM and Toxic Elements
In this study, source regions of PM 10 , PM 2.5 , and As, with the highest CR, were analyzed by PSCF model (Fig. 8). Generally, the potential source areas of the pollutants are almost distributed in the range of the surrounding regions of Zhengzhou in Henan Province. For PM 2.5 , the northwestern and southeastern regions of Zhengzhou, i.e., Jiyuan, Jiaozuo, Xuchang, and Zhoukou, were likely potential source areas, with WPSCF values higher than 0.6. Moreover, the spatial source distributions of PM 2.5 -bound As were confined in Jiyuan, Jiaozuo, Xinxiang, Anyang, and Kaifeng, which are the main industrial cities in Henan, with WPSCF values over 0.4. According to previous studies (AEA, 2011;Bhangare et al., 2011), industrial production (e.g., metal production as well as public electricity and heat production) and coal combustion are the major sources of As. Therefore, the relative activities in the five industrial cities influenced the pollution levels of PM 2.5 -bound As in Zhengzhou. For PM 10 , the potential source area was in the southwestern areas of Zhengzhou, mainly including Pingdingshan and Nanyang (WPSCF values above 0.4). Pingdingshan is the largest coal-producing city in Henan, with raw coal products amounting to 34.0 Mt in 2016 (Bureau of Statistics of Pingdingshan, 2017). Coal mining processes emitted huge amounts of coarse particles and accompanied with south air trajectories, the air mass carried a large amount of particles, thereby influencing the PM 10 concentration in Zhengzhou. Nanyang is a large agricultural city in Henan, with grain, oil-bearing crops, flue-cured tobacco, vegetables and edible fungus, and fruit outputs of 6.4, 1.4, 0.1, 10.5, and 1.6 Mt in 2016 (Bureau of Statistics of Henan Province, 2017). Agricultural activities and bare soil increased PM 10 levels in Zhengzhou when the south winds form. The WPSCF values of PM 10bound As, which are generally less than those of PM 2.5bound As, showed similar spatial distributions in fine particles due to the industries located in the surrounding cities. For comparison, fine particles with higher WPSCF values are more easily transmitted than coarse particles.

CONCLUSION
PM 2.5 and PM 10 filters were collected at five sites in Zhengzhou, and the concentrations, source apportionment, health risks, and source regions of the toxic elements were analyzed. The results indicated severe PM 2.5 and PM 10 pollution, with annual average concentrations that were considerably higher than the Chinese NAAQS. The rural site (ZM) exhibited only slight pollution, but the traffic site (HKG) and the urban site with high traffic (SSQ) showed relatively high PM 2.5 levels, and the highest PM 10 level was also observed at HKG. The highest and lowest mean levels were observed in winter and summer, respectively, for both PM 2.5 and PM 10 .
Overall, high levels of PM-bound crustal elements and plentiful Cl indicated that dust and combustion sources played major roles. Furthermore, the PM 10 -bound As greatly exceeded the Chinese NAAQS, posing a high potential risk. Generally, the total elemental levels were relatively low at ZM and high at GY, with high individual concentrations of Cl, Zn, Pb, and Cu, in particular, at the latter site. High levels of crustal elements were observed at SSQ and HKG, suggesting a significant influence from dust. High levels of crustal elements, which were more abundant in the PM 10 , were observed in spring, whereas combustion-source elements, which were more abundant in the PM 2.5 , displayed elevated levels in winter. The elemental concentrations were low in summer. In general, the CD values for the PM 2.5 were slightly higher than those for the PM 10 . These results are not only related to discrepant spatial distributions of emission sources at the five sites but also attributable to different meteorological conditions across the four seasons.
The Na, Sb, Pb, Zn, Cu, and As were emitted from anthropogenic sources, whereas the Si, Mg, and Ti were crustal in origin. Pearson's CA, cluster analysis, and PCA indicated that vehicles, industry, coal combustion, oil fuel, dust, and biomass burning were probably the main sources of PM-bound elements in Zhengzhou. The ZM site was characterized by low traffic and high contributions from biomass burning and dust emission, whereas the HKG site demonstrated high pollution from traffic sources, and the SSQ site was also highly affected by pollution from vehicles. Whereas elements in the PM 2.5 largely originated in combustion sources, those in the PM 10 , by comparison, received greater contributions from dust sources.
The PM-bound As and Ni posed both intolerable carcinogenic risks and, in addition to Pb, significant non-CRs. In general, children, as shown by higher CR and HQ values, were more sensitive than adults to these risks. The daily intake pathway exhibited the highest CR and HI values. Obvious differences in the CR and HI values were detected between the various sites; hence, data from multiple sites in a study area are necessary for accurate health risk assessment, especially for a large region. Analysis of the source regions identified Jiyuan, Jiaozuo, Xuchang, and Zhoukou; Pingdingshan and Nanyang; and Jiyuan, Jiaozuo, Xinxiang, Anyang, and Kaifeng as the main potential source areas of PM 2.5 , PM 10 , and As, respectively. Furthermore, fine particles with higher WPSCF values were more easily transmitted than coarse ones.