Elemental Composition and Health Risk Assessment of PM 10 and PM 2 . 5 in the Roadside Microenvironment in Tianjin , China

To determine the elemental composition and health risk of particles in the roadside microenvironment, particulate matter samples (PM10 and PM2.5) were collected at the side of four roads in winter, spring, and summer of 2015. The total concentrations of crustal and trace elements and the average concentrations of most single elements followed the sequence of spring, winter, and summer. Crustal elements accounted for 18.9%, 13.2%, and 9.3% of the PM10 and 15.1%, 11.2%, and 18.8% of the PM2.5 in the three seasons, respectively. On average, Zn contributed the largest share, accounting for 34.0%, 32.2%, and 34.2% of the total trace elements in PM10 and for 35.2%, 28.3%, and 35.6% in the PM2.5 for winter, spring, and summer, respectively. The enrichment factor results showed that Mn, K, Na, Co, and Si originated in crustal sources; V, Ca, Mg, Mo, Cr, Fe, Ti, and Ba were derived from both anthropogenic and crustal sources; and Cd, Sb, Zn, Cu, Pb, As, and Ni were largely derived from anthropogenic sources. The results of principal component analysis explained 72.64% of the total variance for PM10 and 78.76% of the variance for PM2.5. Possible sources include resuspended road dust, vehicular exhaust, fugitive dust (road dust and soil dust), and tire/brake wear. The respective ratios of Fe/Al, Mn/V, Cu/Sb, Cu/Zn, Zn/Pb, V/Ni, and Cu/Pb in this study were 0.58, 2.87, 11.07, 0.28, 1.23, 1.34, and 4.45 for PM10 and 1.07, 0.83, 23.07, 0.44, 1.83, 1.60, and 4.18 for PM2.5. Accumulative non-cancerous toxic elements in winter and spring may pose risks to roadside workers via inhalation. The integrated cancer risks (CR) of As, Co, Cd, Cr, Ni, and Pb in PM2.5 and PM10 ranged from 1.58E-05 to 4.95E-05 CR.


INTRODUCTION
Due to rapid urbanization and industrialization, urban traffic has become a major source of atmospheric particles in Chinese megacities (Geng et al., 2013;Tao et al., 2013;Zhang et al., 2013;Huang et al., 2014;Wang et al., 2015;Hsu et al., 2016;Lu et al., 2016;Wang et al., 2016;Fang et al., 2017).Air quality in the roadside microenvironment, which is dominated by traffic-related pollution sources, is important for roadside workers, such as traffic policemen, urban sanitation workers, pedestrians, and drivers.
The main pollution sources of particulate matter (PM) at the roadside include exhaust, abrasion of clutch/brake pad, tires and road surfaces, and resuspension of road dust.These sources are rich in metallic elements which may affect the In order to fill the data gap on the elemental composition of PM 10 and PM 2.5 in the roadside microenvironment in Tianjin and to quantify the health risk of some heavy metals to roadside workers, PM 10 and PM 2.5 samples were collected at four roadside sampling sites in Tianjin, China.Twenty-three metallic elements including eight crustal elements (Na, Mg, Al, Si, K, Ca, Ti, and Fe) and 15 trace elements (V, Cr, Mn, Fe, Ni, Co, Cu, Zn, As, Zr, Mo, Cd, Sn, Sb, Ba, and Pb) were determined.The main objectives were as follows: (1) to determine the elemental composition and characteristic ratios of PM 10 and PM 2.5 in the roadside microenvironment; (2) to assess the possible sources of these elements; and (3) to calculate the health risks of heavy metals in PM 10 and PM 2.5 to roadside workers.

Description of Sampling Sites
Tianjin is located in Northern China (39°10'N, 117°10'E), and had a population of 15.47 million at the end of 2015.The climate in Tianjin is a largely warm and semi-humid continental monsoon climate with distinct seasonal variation.It has 16,110 km of expressways and class I to IV highways stretching around and through the city.The total number of civilian automobiles was 2.7362 million at the end of 2015 (NBSC, 2016).
PM 10 and PM 2.5 samples were collected at four roadside sampling sites, namely Baidi Rd. (BD, secondary road), Fukang Rd. (FK, major road), Jianyang Rd. (JY, expressway), and Waihuan West Rd.(WH, outer ring road) in municipal districts.The one-way traffic flow in the four roads was approximately 1,000, 2,000, 4,300, and 1,400 vehicles h -1 , respectively.Corresponding average vehicle speeds were 38, 54, 73, and 73 km h -1 .The locations of the four sampling sites are shown in Fig. 1.

Sampling Protocol
Polypropylene filters were prebaked in an oven at 60°C for 2 h, and then equilibrated under constant relative humidity (35% ± 1%) and temperature (22.0°C ± 1°C) for 48 h before gravimetric analysis.Samplers (TH-150AII, Wuhan Tianhong Co., Ltd, China) were operated at a sampling flow rate of 100 L min -1 with cutoff point diameters of 10 and 2.5 µm to collect PM 10 and PM 2.5 (particles smaller than 10 or 2.5 microns).
The sampling height was approximately 1.5 m above ground level and approximately 1.5 m away from the traffic road.There were no other obvious anthropogenic sources apart from vehicles at the sampling points.Three sampling campaigns were carried out at each site, in January (winter), April (spring), and July (summer) 2015.Sequential sampling was adopted at the four sites due to equipment limitations.Samples were collected during four periods on each day as follows: 07:00-11:00 (including the morning rush hour), 11:00-15:00 (including the lunch and rest hours), 15:00-19:00 (including the afternoon rush hour), and 19:00-23:00 (including trucks driven in the city center).Detailed information on the sampling campaigns is shown in Table 1.All samples were stored at 4°C until analysis.

Chemical Analyses and Statistical Analysis Methods
After 48 h of equilibration and weighing, Mg, Al, Si, Ca, Ti, Fe, Zr, and Ba were determined using inductively coupled plasma optical emission spectrometry (Vista MPX, Varian Inc., CA, USA).Part of the polypropylene filter was placed in a muffle furnace on a crucible for low temperature carbonization; the temperature was gradually increased until the filter was completely ashed.A few drops of absolute ethyl alcohol were added followed by solid NaOH (0.1-0.2 g) and then 5 mL hot water (approximately 90°C) was added when the sample had reached a molten state.The solution was then transferred into a plastic tube with 2 mL HCl (50% v/v) after it was boiled on a galvanothermy board.Finally, the samples were diluted to 10 mL with HCl (50% v/v).The other 15 elements (Na, K, V, Cr, Mn, Ni, Co, Cu, Zn, As, Mo, Cd, Sn, Sb, and Pb) were analyzed using inductively coupled plasma mass spectrometry (ICP-MS) (Agilent 7500a, Agilent Technologies Inc., USA) (see Kong et al., 2011, for details).For quality assurance and quality control (QA/QC), blank samples were processed simultaneously and analyzed using the same procedure.To guarantee high recovery rates (80%-120%), one standard sample was analyzed for every 10 samples.Descriptive statistical analysis and principal component analysis (PCA) were performed using SPSS.This is the most common multivariate statistical method for distinguishing pollution sources and their contributions to particles; further details are provided in a review by Liang and Duan (2016).

Enrichment Factor Analysis
An enrichment factor (EF) approach was used to assess the degree of pollution caused by toxic elements (Liu et al., 2014;Alghamdi et al., 2015;Keshavarzi et al., 2015;Li et al., 2015;Meena et al., 2016;Nayebare et al., 2016;Acciai et al., 2017;Cheng et al., 2017).EF is defined as: where C n /C ref is the ratio of concentrations between target elements and a reference element in the sample (PM) and the upper continental crust (UCC), respectively.The content of Si in the UCC was derived from Taylor and McLennan (1995) and the others were derived from the background values in Chinese soils (CNEMC, 1996).The coefficient of variation (CV) of the common reference elements (Al, Fe, Zr, Ca, Ti, and Si) were calculated.CV values of PM 2.5 and PM 10 in ascending order were as follows: Al (0.26), Ca (0.36), Fe (0.42), Zr (0.45), Ti (0.46), and Si (0.71), and were Al (0.35), Ca (0.36), Fe (0.42), Zr (0.45), Ti (0.46), and Si (0.71), respectively.As the CV values of Al in PM 2.5 and PM 10 were the lowest, Al was chosen as the reference element in this study.

Health Risk Assessment Model
In this study, the model used to estimate exposure of roadside workers to heavy metals was developed by the US EPA (USEPA, 1989).This model has been widely applied in the health risk assessment of heavy metals in China (Sun et al., 2014;Li et al., 2016b).
Human exposure to the toxic effects of PM 10 and PM 2.5 mainly takes place via inhalation.Therefore, only the exposure concentration (EC, µg•m -3 ) via inhalation was calculated using Eq. ( 2) (USEPA, 1996): where the contaminant concentration of heavy metals in air (C, µg m -3 ) is considered to yield an estimate of the ''reasonable maximum exposure'' of the 95% upper confidence limit (95% UCL), which were calculated using SPSS software; EF is exposure frequency (350 day/year); ED is exposure duration (24 years); and ET is exposure time (8 hour/day for roadside workers); AT n = ED × 365 days/year × 24 hours/day, when evaluating cancer risk, AT n = 78.89× 365 days/year × 24 hours/day.Lifetime values were cited from the Exposure Factors Handbook of Chinese Population (adults) (MEPC, 2014) and others were cited from the User's Guide/Technical Background Document for US EPA Region 9's RSL (Regional Screening Levels) Tables (USEPA, 2013).The Hazard Quotient (HQ) was used to estimate the noncancer effects of a single element via inhalation.The Hazard Index (HI) is the sum of HQ for multiple chemicals.HQ and cancer risk (CR) were calculated using the following equations: where R f C i is the inhalation reference concentration (µg m -3 ) and IUR is the inhalation unit risk (µg m -3 ) −1 .If HQ or HI < 1, it can be considered that there is little or no non-cancer risk to humans, but when HQ or HI ≥ 1, adverse health effects in humans may occur and special attention is required.CR is the probability of cancer, and shows the number of individuals with cancer among a certain number of people.The USEPA has indicated that a one in a million chance of additional human cancer is the acceptable level of risk (1.0E-06) and a one in ten thousand chance or greater is too serious to gain the priority of attention, which means that the tolerable risk for the public is between 1.0E-06 and 1.0E-04.Non-cancer and cancer health risks are evaluated based on a number of assumptions and estimates.

Seasonal Variations an PMs and Trace Elements
Table 2 provides descriptive statistics of PM 10 and PM 2.5 and element concentrations during the sampling period.The concentrations of PM 10 and PM 2.5 ranged from 40.09 to 746 µg m -3 and from 13.03 to 309 µg m -3 , with average concentrations of 78.9 and 186 µg m -3 , respectively.Both PM 2.5 and PM 10 concentrations were more than twice the annual second-level concentration limits of the Chinese national ambient air quality standard (70 and 35 µg m -3 ).
The average concentrations of PM 10 and PM 2.5 were highest in winter (264 ± 147 and 112 ± 76 µg m -3 ), followed by spring (234 ± 118 and 89 ± 57 µg m -3 ) and summer (112 ± 51 and 51 ± 34 µg m -3 ).The average ratios of PM 2.5 /PM 10 were similar in winter (0.43 ± 0.19), spring (0.41 ± 0.22) and summer (0.42 ± 0.13).This was due to the traffic-related sources being dominated by the emission of particles and the constituents of PM 2.5 and PM 10 were not significantly affected by other sources (such as coal burning for heating in winter).
With regard to annual average concentrations, the sequence of crustal elements in PM 10 and PM 2.5 was as follows: Si > Al > Ca > Fe > Mg > K > Na > Ti, and Si > Fe > Al > Ca > Mg > K > Na > Ti, respectively, and accounted for 20.2% and 19.5% of the PM 10 and PM 2.5 .
The most abundant trace elements in PM 10 were Zn, Ba, and Mn and were Zn, Cu, and Ba in PM 2.5 , accounting for 61.6% and 56.0% of the total trace element concentrations in PM 10 and PM 2.5 , with average concentrations ranging from 59.22 to 159.1 ng m -3 and from 22.35 to 80.54 ng m -3 , respectively.Less abundant trace elements were Cu, Pb, Zr, V, Cr, and Ni in PM 10 and were Pb, Zr, V, Mn, and Ni in PM 2.5 , with average concentrations ranging from 15.41 to 44.05 ng m -3 and from 10.78 to 19.75 ng m -3 , respectively.
With regard to seasonal variation in the total concentrations of crustal elements in PM 10 and PM 2.5 , these were higher in spring (44,300 ng m -3 and 13,432 ng m -3 ), but lower in winter (20,887 ng m -3 and 12,638 ng m -3 ) and summer (21,703 ng m -3 and 9,634 ng m -3 ) and accounted for 18.9%, 13.2%, and 9.3% of PM 10 and accounted for 15.1%, 11.2%, and 18.8% of PM 2.5 , respectively.In PM 10 , the concentration of each crustal element in descending order was as follows: Si, Al, Ca, Fe, Mg, K, Na, and Ti.Although the concentration of crustal elements in PM 2.5 showed a different order, Si, Fe, and Al were the most abundant elements in the three seasons.
Similar to the total concentration of crustal elements, the average total concentration of the 15 trace elements in PM 10 and PM 2.5 followed the order spring > winter > summer.Twelve of 15 trace elements in PM 10 and nine of 15 in PM 2.5 showed the same seasonal variation in the total concentration of trace elements, which were Cr, Mn, Ni, Co, Cu, Zn, Mo, Cd, Sn, Sb, Ba, and Pb in PM 10 and were Cr, Mn, Co, Zn, Mo, Cd, Sn, Sb, and Pb in PM 2.5 , respectively (Fig. 2).The average concentration of Zn was highest in the winter, spring, and summer PM 10 and PM 2.5 samples, and accounted for 34.2%, 34.0%, and 32.2% of total trace elements in PM 10 and accounted for 35.6%, 35.2%, and 28.3% of total trace elements in PM 2.5 , respectively.This suggested that Zn is a significant pollution source all year round in addition to emission due to coal burning for heating, brake/tire wear, and steel smelting (Tian et al., 2012;Pant et al., 2013).The average concentrations of Cr, Sb, Ba, and Pb in spring PM 10 samples were similar to those in winter samples, but higher than those in summer samples, indicating that the main contributor of these elements is the same in the three seasons and that wet deposion in summer significantly reduced pollution.
The ratio of trace element concentrations in PM 2.5 and PM 10 (Fig. 3) showed that the ratios of Mn and Co were lower than the others, which indicated that Mn and Co tended to settle onto coarse particles.The differences in the ratios of V, Ni, Cu, Zn, As, Zr, Sb, and Ba were large in the winter, spring, and summer samples.All ratios of these eight elements were higher in summer samples except Sb, which showed a higher ratio in winter samples.

Source Identification Enrichment Factor
Fig. 4 shows the seasonal average EFs of the measured elements.In the three sampling seasons, Mn, K, Na, Co, and Si had mean EF values below 1 for PM 10 and PM 2.5 , indicating that they originated from crustal sources.V, Ca, Mg, Mo, Cr, Fe, Ti, and Ba had average EFs between 1 and 5 and were considered to be derived from both anthropogenic and crustal sources.Elements with EF values higher than 5 (Cd, Sb, Zn, Cu, Pb, As, and Ni) were identified as having anthropogenic origins (Hsu et al., 2010;Schleicher et al., 2011;Hsu et al., 2016;Li et al., 2016).

Principal Component Analysis (PCA)
PCA was carried out to determine the source factors of PM 10 and PM 2.5 .Metallic elements with eigenvalues above 1 in PCA were used in grouping (Table 3).Previous studies included Zn, Fe, Ba, Cu, Sb, Zr, Pb, Cr, Sn, Mo, Fe, Ni, and Cd as the key tracers for brake wear; Zn, Ca, K, Fe, Ti, Cr, Mo, and Co as the key tracers for tire wear; and Al, Si, Ti, Fe, Mn, As, Ca, Sb, Sn, Cu, Zn, and V as the key tracers for resuspended road dust (summarized by Pant et al., 2013).Road dust can be considered a receptor of PM, and has a variety of sources, such as brake/tire/clutch wear, road surface abrasion, soil dust, and dustfall.Therefore, the elemental composition of road dust is diverse.Ni and V have also been reported to be present in emissions due to oil combustion, and Ba, Sb, and V were considered potential tracers for gasoline and liquefied petroleum gas (LPG) engines (Cheng et al., 2010;Pey et al., 2010).For PM 10 , six factors were identified and 72.64% of the total variance was explained.High factor loading for Cr, Cu, Cd, and Pb in Factor 1 and high factor loading for Mn and Co in Factor 2 were likely attributed to the emission of road dust resuspension.In Factor 3 and 4, high factor loadings for Al, Si, Ti, and Fe were observed.This probably originated from fugitive dust (Thorpe and Harrison, 2008;Kong et al., 2010;Shen et al., 2016).High loadings for V and As were likely related to vehicular exhaust or coal burning (Xie et al., 2006;Pey et al., 2010).Factor 6 was represented by Cu, Sn, and Sb with 7.56% of the total variance and was attributed to vehicle brake wear (Pio et al., 2013).For PM 2.5 , five factors were identified which explained 78.76% of the total variance.Factor 1 had the highest variance (36.49%) with an eigenvalue of 8.39 and high loadings for Na, K, Mn, Co, Zn, Mo, Sn, Sb, and Pb (> 0.7) and moderate loadings for Cr, Cd, and Ba (> 0.5).The high loading factors for these elements were possibly due to vehicular exhaust emission and tire/brake wear.Factors 2 and 3 explained 26.04% of the total variance and high loadings for Al, Si, Fe, and Ti, and moderate loadings for Mg, Ca, and Zr, may be attributed to fugitive dust (Lin et al., 2005;Yu et al., 2013).Factors 4 and 5 explained 15.8% of the total variance and high loadings for Ni, Zn, Cd, Cu, and As, that may be attributed to coal combustion (As) and oil combustion (Ni) (Tian et al., 2012).

Characteristic Ratios of Specific Elements in Roadside PM
Elemental ratios were used to estimate the possible sources of particles.The ratios of Al/Fe, Mn/V, Cu/Sb, Cu/Zn, Cu/Pb, V/Ni, and Zn/Pb in PM 10 and PM 2.5 in the roadside microenvironments and other pollution sources are listed in Table 4.The Fe/Al ratio in roadside PM 10 was higher than the ratio in soil dust, possibly due to the emission of Fe from traffic-related sources (e.g., the Fe/Al ratio in braking dust can reach 13.35).A Cu/Sb ratio of 4.6 ± 2.3 was proposed as a typical value for brake wear particles (Sternbeck et al., 2002).However, higher Cu/Sb ratios in  Zhao et al., 2013; c Ni et al., 2013; d Shen et al., 2016; e Bi et al., 2007; f Hulskotte et al., 2014;  g Wu et al., 2016; h Chiang et al., 2012; I Pulles et al., 2012.PM 2.5 and PM 10 were found in the present study.Thus, the Cu/Sb ratio can vary across geographical regions based on the type of brake or the elemental content of the brakes, which differ depending on the manufacturer (Pant and Harrison, 2013;Hulskotte et al., 2014); the Cu concentration was relatively high in this study.The ratios of Cu/Zn and Cu/Pb in PM 2.5 (Cu/Zn: 0.44, Cu/Pb: 1.83) in this study were higher than those in the ambient atmosphere (Cu/Zn: 0.19, Cu/Pb: 0.64), road dust (Cu/Zn: 0.17, Cu/Pb: 0.96), and light-duty diesel exhaust (Cu/Zn: 0.25, Cu/Pb: 0.59).In addition, these ratios in PM 10 (Cu/Zn: 0.28, Cu/Pb: 1.23) were higher than those in ambient atmosphere (Cu/Zn: 0.15, Cu/Pb: 0.65), fugitive dust (Cu/Zn: 0.13-0.17,Cu/Pb: 0-1.00), and light-duty diesel exhaust (Cu/Zn: 0.14, Cu/Pb: 0.29), but were much lower than those in brake dust (Cu/Zn: 2.38, Cu/Pb: 85.82), which indicated that Cu, Zn, and Pb were derived from sources other than brake/tire wear and exhaust.The ratios of V/Ni in PM 10 and PM 2.5 were 1.06-2.04and 1.06-1.92with average values of 1.34 and 1.60, and were higher than the values in vehicle exhaust, suggesting that there were other pollution sources of V and Ni.Values of the Zn/Pb ratio (4.45 for PM 10 and 4.18 for PM 2.5 ) were similar to the ratios in ambient PM (3.41 for PM 2.5 and 4.40 for PM 10 ) and road dust (5.5 for PM 2.5 ), indicating that they may have the same pollution sources.

Health Risk Assessment of Heavy Metals
Non-cancer risks and cancer risks due to toxic elements in roadside PM 2.5 and PM 10 via inhalation are shown in Fig. 5 (the R f C i , IUR, 95% UCL and EC of each toxic element are listed in Table S1).The HQ values for As, Ba, Co, Cd, Cr, Mn, Ni, Sb, and V in PM 2.5 and PM 10 via inhalation for roadside workers were lower than 1, indicating no non-cancer risks via the inhalation exposure pathway for each element.The HI (total HQ value of all studied elements) values of PM 10 and PM 2.5 in spring were higher than those in winter and summer.HI values of PM 10 in spring and winter were higher than 1, which indicated that there were accumulative non-cancer risks in roadside workers via inhalation in spring and winter.
The cancer risks of PM 2.5 via inhalation of a single element (Cd, Co, Ni, and Pb) in roadside workers were below the acceptable level (1.0E-06), while the CR values of As and Cr were higher than 1.0E-06 (ranging from 4.04E-06 to 1.59E-05).The integrated risks of these six elements (2.12E-05 in winter, 2.19E-05 in spring, and 1.58E-05 in summer) were within the tolerance level (1.0E-04).The CR value of each carcinogenic element in PM 10 was higher than that in PM 2.5 , and CR values of As, Co (values in winter and spring) and Cr were higher than 1.0E-06 and the CR value ranged from 1.79E-06 to 4.05E-05.The accumulative CR value of PM 10 was 4.95E-05 in winter, 4.84E-05 in spring, and 2.58E-05 in summer.Although the CR values were lower than 1.0E-04, attention should still be paid to the the health of roadside workers and measures should be taken to reduce heavy metal contamination.
EC may be overestimated or underestimamted due to individual differences in exposure frequency (EF) in different age groups and occupations.Although there is still uncertainty, the results of health risk assessment include a reference value for evaluating non-cancer and cancer risks in people working in roadside environments.

CONCLUSIONS
Seasonal concentrations of PM (PM 2.5 and PM 10 ) and 23 metallic elements were characterized at four roadside sampling sites in Tianjin to identify the levels and the major sources of pollution.The concentrations of most measured metallic elements were influenced by human activities.Among all the detected elements, V, Ca, Mg, Mo, Cr, Fe, Ti, and Ba involved both anthropogenic and crustal sources; however, Cd, Sb, Zn, Cu, Pb, As, and Ni, with average EFs that were higher than 5, were dominated by anthropogenic sources.
Using PCA, six factors were extracted for PM 10 and five factors for PM 2.5 , with possible sources including resuspended road dust, vehicular exhaust, fugitive dust (road dust and soil dust), and tire/brake wear.Resuspended road dust was the main source of elements in PM 10 , whereas the main sources of elements in PM 2.5 were vehicular exhaust and tire/brake wear.Comparing the characteristic ratios in roadside PM 2.5 and PM 10 with same size particles from other sources, the Fe/Al ratio was higher in PM 10 than in soil dust, the Cu/Sb ratios were higher in PM 2.5 and PM 10 than in particles from brake wear, and the Cu/Zn and Cu/Pb Fig. 5. HQ, HI, and CR (cancer risk) values for toxic elements in PM 10 and PM 2.5 .ratios were higher in PM 2.5 than in road dust, resuspended dust, and light-duty diesel exhaust.Additionally, the V/Ni ratio was higher in PM 2.5 and PM 10 than in vehicle exhaust, whereas the Zn/Pb ratio was similar to that in road dust.
Based on the annual concentrations of heavy metals in PM, the accumulative non-cancer risks due to PM 10 in winter and spring exceeded safe levels via inhalation for roadside workers.The integrated cancer risk due to As and Cr in PM 2.5 and PM 10 was within tolerable levels.

Fig. 4 .
Fig. 4. Average EF values for elements in spring, summer, and winter.

Table 1 .
Details of sampling campaigns.

Table 3 .
Factor loadings and eigenvalues after varimax rotated normalization of elements in PM 10 and PM 2.5 in roadside microenvironments.

Table 4 .
Comparison of elemental ratios in roadside PM and related sources.