Trace element mass fractions in road dust as a function of distance from road

ABSTRACT

Road dust particles play an important role in atmospheric pollution and are associated with adverse human health effects. Traffic emissions are a major source of particles in road dust. However, there has been limited information about the relationship between distance from road and traffic-related elements levels in road dust. We investigated the relationships between proximity to the nearest major roadway and trace element mass fractions in PM₁₀ and PM_2.5 re-suspended from the road surface, based on measurements at three different distance ranges. We found that mass fractions of Ba, Cu, Zr, Zn, Cl, Co, Cr, Ca, Ti in PM₁₀ road dust as well as Zr, Cu, Cl, Zn, Cr, Ti, Mn, Ca, Ni, and Fe in PM_2.5 road dust, significantly decreased with distance from major road. Most of these elements are associated with road traffic emissions, including both tailpipe and non-tailpipe emissions. The decrease rates differed among elements due to differences in local traffic contributions. The decreases for elements which are mainly associated with non-tailpipe traffic emissions (e.g., Ba, Zr) were more dramatic. Our results indicate that traffic emissions, especially non-tailpipe emissions, contribute substantially to road dust, suggesting the need for control strategies for non-tailpipe emissions.

Implications: We investigated the relationships between road proximity with trace element mass fractions in PM10 and PM2.5 re-suspended from the road surface. We observed significant decrease of traffic-related elements in PM10 and PM2.5 road dust with log distance from major road. We also found that the mass fractions for elements, which mainly come from traffic decrease more sharply compared to elements which come from both traffic and other sources. Our results indicate that traffic emissions contribute substantially to road dust, and imply that the distance to major road can be used as a proxy for ambient exposure.

Introduction

Road dust is a complex mixture of particles that contain various chemical components such as organic, inorganic, and trace elements (Nakajima and Aryal 2018). Road dust commonly contains high concentrations of trace contaminants, which originate from road traffic (including tailpipe emissions from combustion of fuel and lubricating oil additives and non-tailpipe emissions from brake wear, tire wear, road surface wear), crustal material (such as soil and sand), road salt, vegetation debris, and so on. (Gunawardana et al. 2012; Hwang et al. 2018). For example, a considerable proportion of Zn, Cu, Pb, Ni, Cr, and Cd partly originate from road traffic; Si, Ca, and Al originate from soil; Na and Cl mainly originate from road salt (Gunawardana et al. 2012; Ho et al. 2003; Johansson, Norman, and Burman 2009).

Particulate matter (PM) exposure has been proved to be associated with increases in morbidity and mortality (Gilmour et al. 2007). Particulate matter (PM) is defined as an air-suspended mixture of a variety of chemical species (e.g., Water-Soluble Inorganic Ions [WSIIs], Organic Carbon [OC], Elemental Carbon [EC], crustal, and trace elements) (Samek et al. 2017). Hundreds of epidemiological studies have shown that exposure to PM is associated with increases in morbidity and mortality (Gilmour et al. 2007). The toxicity of PM is influenced by its sizes and compositions (Li et al. 2019). Road dust is typically re-suspended into the air, due to motion of vehicles and wind. As an important source of atmospheric PM (Li et al. 2018), road dust also has been proved to be significantly associated with adverse health effects (Khan and Strand 2018). Exposures to road dust particles have adverse effects on the respiratory system (Barrett et al. 2010; Maher 2010; Saini-Eidukat and Triplett 2014; Weissman and Kiefer 2011). PM_2.5 in road dust have been found to be related to the increase of cardiovascular-related hospital admissions and mortalities (Kioumourtzoglou et al. 2014). Ba, Pb, and Cu in road dust are associated with non-carcinogenic health risks (Li et al. 2015). Ar in road dust has both carcinogenic and non-carcinogenic effects on humans (Tchounwou, Patlolla, and Centeno 2003).

According to existing studies, road traffic has significant influence on the resuspension of road dust, including PM_2.5 and PM₁₀ road dust (Amato et al. 2016; Birmili et al. 2006; Bukowiecki et al. 2010; Duong and Lee 2011; Harrison et al. 2012; Jeong et al. 2019; Ntziachristos et al. 2007). These studies indicated that traffic volume, vehicle speed, type of traffic emission (exhaust emission and non-exhaust emission) were influencing factors of contamination levels in road dust. Road proximity have also been recognized as an important predictor of ambient PM exposure (Dorans et al. 2016; Wilker et al. 2015). Studies also indicated that contaminants levels in road dust decreased with the increase of distances from major road (Hitchin et al. 2000; Pagotto et al. 2001).

In this study, our aim was to investigate the relationships between distance from the closest major road and the mass fractions of traffic-related trace elements (including Na, Mg, Al, Si, S, Cl, K, Ca, Ti, V, Cr, Mn, Fe, Co, Ni. Cu, Zn, Br, Sr, Zr, Sn, Sb, Ba, and Pb) in re-suspended road dust particles, including those with diameters that are below 10 µm (PM₁₀) and 2.5 µm (PM_2.5, fine particles). Our analysis was based on 78 PM₁₀ and 79 PM_2.5 re-suspended from the road surface, which were collected at different distance ranges from 20 major roads in the Greater Boston, Massachusetts area.

Materials and methods

We conducted a field study from June 2018 to January 2019 in the Greater Boston area to examine the relationships between road proximity and trace elements mass fractions in PM₁₀ and PM_2.5 road dust at different sites, including roadside (Road), near-road and background (Back) sites, with different distances from the closest major roads. The roadside site is at the side of the major road. The near-road and background sites are on small, low traffic roads, which were often residential but at different distance ranges from the major road. The distance ranges from the center of the nearest major road to the sampling sites were: ≤25 m for roadside site, 26–200 m for near-road site, and >300 m for background.

Road proximity

We put the coordinates in the “Google Earth Pro” software and measured the perpendicular distance (d in meters) from sampling site to center of the closest major road. The accuracy of this method is 1.7 m (Mohammed, Ghazi, and Mustafa 2013). We included three types of major roads from the U.S. Census Feature Class Codes: A1 road, a primary road with limited access via on ramps and exit ramps; A2 road, a primary road without limited access; and A3 road, a secondary or connecting road. For this study, we did not consider differences in elevation because they are small compared to distance variability (Huang et al. 2018).

Measurement of trace elements concentrations in PM₁₀ and PM_2.5 from road dust

We developed a novel Road Dust Aerosolizer (RDA) sampler to collect PM₁₀ and PM_2.5 re-suspended from surface road dust (PM₁₀ and PM_2.5 road dust). This sampler aerosolizes dust from the road surface, simulating ambient road dust resuspension, resulting in measured PM composition and size more appropriately (and less labor-intensive) than those obtained from studies using bulk road dust sieved and re-suspended in the laboratory. Our results showed that, for samples with elements that were well above the lower detection limit (LOD), the relative error for mass was 7% for PM₁₀ and 3% for PM_2.5 within pairs and ranged from 2 to 10% for elements, indicating that the RDA system has an excellent precision for mass and elements.

During this study, we collected samples of PM_2.5 and PM₁₀ dust re-suspended from the road surface at a series of three distances away from major road sites in the Greater Boston area: roadside, near-road, and local background (as described in the introduction). In order to avoid the impact of precipitation, we did not collect samples during rainy days. Major roadways sampled included multi-lane divided state and interstate highways (with<A1/> or without<A2/> limited access via onramps and exit ramps) and busy state secondary and connecting roads <A3/>. Near-road and background samples were collected at a location on an adjacent roadway within the target distance from the roadside site. Near-road samples were collected on roads in residential or commercial areas, while background samples were almost entirely on residential roads. Some locations were sampled more than once, during different seasons. The distribution of sampling sites shown in Figure 1. Details of each sampling day including the road ID, date, and number of sampling sites are shown in Table S1 in Supplementary Material. We had repeated measurements for 6 major roads. Finally, we had 27 sampling days for 20 major roads at 62 sites, including 78 PM₁₀ samples and 79 PM_2.5 samples. We used energy dispersive X-ray fluorescence (EDXRF) spectrometry to analyze samples for 23 elements, including Na, Mg, Al, Si, S, Cl, K, Ca, Ti, V, Cr, Mn, Fe, Co, Ni. Cu, Zn, Br, Sr, Zr, Sn, Sb, Ba, and Pb. Uncertainty was determined by repeating a measurement to arrive at an estimate of the standard deviation of the values. LOD is 3 times of uncertainty (Kang et al. 2014). In our data analysis, we excluded elements for which the uncertainty was greater than the measurement in more than 20% of samples. The concentration of each element was expressed as mass fraction, that is, the mass of an element (μg) divided by PM₁₀ or PM_2.5 road dust mass (mg).

Figure 1. Distribution of sampling sites within the greater Boston metropolitan area

Back represents samples taken at a distance > 300 m; Near-road represents samples taken at a distance range of 26 to 200 m; Road represents samples taken at a distance range of 0 to 25 m.

Other covariates

We examined the effects of season in our primary model. We considered June through September as the warm season and October through January as the cold season. In order to examine whether elemental fraction depends on road type, we examined the effect of road type as well as the interaction of road type and proximity. In order to increase statistical power, we combined A1 and A2 as primary roads (A12), as a result we developed statistical models for two road types (A12 and A3).

Statistical analysis

We used a mixed-effects regression model (package with “lme4” and “lmerTest” in R software, version 3.3.2, The R Project; https://www.r-project.org/) to examine the relationships between each of the trace element mass fractions in PM₁₀ and PM_2.5 with road proximity. We used linear mixed-effects models as we had repeated measurements for each major road (road, near-road, and background samples collected on the same day) on a given day defined as experiment i. We used the following model:

(1) $ln\left(C_{ij}\right)\hspace{0.17em}={\beta }_{0j}+{\gamma }_{ij}+{\beta }_{1j}\times \hspace{0.17em}ln{d_i}_j+{\beta }_{2j}\times season+{\beta }_{3j}\times r{t}_{ij}+e_j$(1)

where C_ij is the mass fraction of element j for experiment i; β_0j is the fixed intercept; γ_ij is the random intercept for experiment i; lnd_ij is the log value of perpendicular distance of the sampling site to the nearest major road (meters); rt is the road type (rt = 1 for A12 roads, while rt = 0 for A3 roads); β_1j is the slope of an element j for lnd_ij; β_2j is the slope for season (season = 1 for warm season, while season = 0 for cold season); β_3j is the slope for road type; and e_ij is the error term.

Before we determined our final model, we also considered meteorology factors including wind direction (including upwind, downwind, and parallel wind), wind speed during sample period, and daily temperature. For each road, wind speed and direction were measured using a Davis Vantage Vue Weather Station at the roadside location. The daily temperature is from National Centers for Environmental Information (NOAA) (https://www.ncdc.noaa.gov/). However, we found no significant relationship between wind direction or wind speed with the mass fraction of elements. Therefore, we did not include these factors in our statistic model.

We considered a relationship as significant when p < .05.

Results

Mass fractions of trace elements

Table S2 shows the mass fractions of selected trace elements in 78 PM₁₀ and 79 PM_2.5 samples of re-suspended road dust, separated by columns representing different distance ranges from the nearest major road. We conducted paired sample t-tests to compare the differences of elements fraction masses in PM₁₀ and PM_2.5. The p-values of Welch’s t-tests for Na, Mg, Al, Si, K, Cr, Fe, Ni, Cu, Zn, and Br are less than 0.05 as shown in Table S3, indicating the mass fractions for these elements in PM_2.5 were significantly higher than those in PM₁₀. The main reason for this is that the PM₁₀ has proportionally more soil particles (Hueglin et al. 2005), and therefore, the toxic elemental fractions are higher for the PM_2.5 samples. The mass fractions of selected trace elements in cold and warm seasons are shown in Tables S4 and S5, respectively, in the Supplementary Material. Figure 2 shows the mass fractions of 6 elements (i.e., Ba, Ca, Cu, Fe, Zn, and Zr) at different distance ranges.

Figure 2. Mass fractions of Ba, Ca, Cu, Fe, Zn and Zr at different distance ranges

SD represents the standard deviation: Back represents samples taken at a distance > 300 m; Near-road represents samples taken at a distance range of 26 to 200 m; Road represents samples taken at a distance range of 0 to 25 m.

Percentages of the elements samples above the uncertainty are shown in Table S6 in the Supplementary Material. Our model only included elements for which the element measurement was greater than sample uncertainty for at least 80% of the samples. Based on these criteria, we excluded Br and Sn in the PM₁₀ samples, and Br, Sn, and Ba in the PM_2.5 samples.

Distribution of road proximity

The range of distances from the nearest major road ranged from 2 to 1,250 m. About 1/3 of the samples were collected at the roadside, within 25 m of the nearest major road. The average (SD, standard deviation) values from sites to major roads were 282 m (352 m). The average (SD) values for road, near-road, and background distances were 10.4 m (6.5 m), 113 m (51.0 m), and 745 m (237 m), respectively.

Associations between mass fractions of trace elements and road proximity

Table 1 shows the regression coefficients of ln(road distance) for each of the elements in PM₁₀ and PM_2.5 road dust. Eleven elements (Cu, Zr, Zn, Cl, Co, Cr, Pb, Ca, Ti, Fe, and Ba) in PM₁₀ road dust and 10 elements (Zr, Cu, Zn, Cl, Cr, Mn, Ti, Ca, Ni, and Fe) in the PM_2.5 road dust have significantly negative associations with the log distance from major roads (trace elements are presented from the highest to lowest |β_1i|). The decrease in mass fraction of these elements may reflect both a decreased contribution of re-suspended dust from the main roadway and the decreased traffic at the near-road and background sites compared to the major road.

Table 1. Regression coefficients of log road distance for each of the trace elements in PM10 and PM2.5 from road dust

Element	PM10		PM2.5
	β1i (95%CI)	p-value	β1i (95%CI)	p-value
Na	−0.017 (−0.081, 0.048)	0.617	−0.011 (−0.086, 0.063)	0.768
Mg	0.010 (−0.035, 0.055)	0.656	0.023 (−0.022, 0.068)	0.327
Al	0.029 (−0.002, 0.060)	0.070	0.024 (−0.001, 0.048)	0.059
Si	0.026 (−0.002, 0.055)	0.078	0.018 (−0.002, 0.039)	0.084
Cl	−0.089 (−0.157, −0.021)	0.013	−0.113 (−0.176, −0.050)	0.001
K	0.018 (−0.001, 0.037)	0.062	0.006 (−0.010, 0.022)	0.458
Ca	−0.035 (−0.068, −0.002)	0.040	−0.050 (−0.087, −0.012)	0.012
Ti	−0.031 (−0.054, −0.007)	0.012	−0.050 (−0.074, −0.026)	<0.001
V	−0.008 (−0.053, 0.038)	0.745	−0.016 (−0.076, 0.045)	0.619
Cr	−0.064 (−0.093, −0.034)	<0.001	−0.065 (−0.099, −0.032)	<0.001
Mn	−0.017 (−0.053, 0.019)	0.362	−0.052 (−0.082, −0.022)	0.001
Fe	−0.031 (−0.045, −0.016)	<0.001	−0.043 (−0.059, −0.027)	<0.001
Co	−0.071 (−0.119, −0.022)	0.006	−0.042 (−0.109, 0.025)	0.224
Ni	−0.024 (−0.058, 0.010)	0.171	−0.050 (−0.092, −0.008)	0.025
Cu	−0.126 (−0.166, −0.087)	<0.001	−0.141 (−0.187, −0.094)	<0.001
Zn	−0.094 (−0.127, −0.060)	<0.001	−0.109 (−0.143, −0.075)	<0.001
Sr	−0.010 (−0.048, 0.027)	0.588	0.014 (−0.031, 0.059)	0.555
Zr	−0.105 (−0.161, −0.049)	0.001	−0.178 (−0.266, −0.089)	<0.001
Ba	−0.138 (−0.213, −0.063)	0.001	-	-
Pb	−0.064 (−0.121, −0.007)	0.033	−0.063 (−0.132, 0.006)	0.079

Notes. “-“ represents that the element was excluded for mixed model regression analysis; we excluded elements for which the uncertainty was greater than the measurement in more than 20% of samples.

Most of the above elements are mostly or partially originated from road traffic emissions: Ca originates from soil as well as the combustion of motor oil additives; Mn is mainly due to aged road dust and gasoline additive (MMT) (anti-knock agent) (Cadle et al. 1997; Dongarra, Manno, and Varrica 2009; Garg et al. 2000; Gertler et al. 2002; Hjortenkrans et al. 2003; Kam et al. 2012; Lough et al. 2005); Fe is associated with brake, tire, engine wear (Cadle et al. 1997; Garg et al. 2000; Harrison et al. 2012; Kam et al. 2012; Lough et al. 2005); Zn is associated with brake wear, tire wear, and tailpipe emissions of motor oil (specifically from zinc dithiophosphate, an anti-wear and antioxidant compound) (Cadle et al. 1997; Councell et al. 2004; Garg et al. 2000; Hjortenkrans, Bergback, and Haggerud 2003; Huang, Olmez, and Aras 1994; Lough et al. 2005); Cu is associated with brake wear, brake lining, vaporization from hot brake surfaces and wear of bearings or other components, and motor/lube oil combustion (antioxidant additive) (Cheung et al. 2010; Kaiser 2003; Lytle, Smith, and McKinnon 1995; Sternbeck, Sjodin, and Andreasson 2002); Zr comes primarily from brake dust (Adachi and Tainosho 2004; Garg et al. 2000), and can be effectively bound to humic substances in road dust with a preference for complexation with low molecular weight compounds; Ba is mostly due to brake wear and lining (Garg et al. 2000). Pb originates from motor vehicle wheel weights, motor oil combustion, brake wear, and vaporization from hot brake surfaces (Cadle et al. 1997; Garg et al. 2000; Gertler et al. 2002; Kam et al. 2012; Lough et al. 2005; Ning et al. 2008; Root 2000); Cl mostly originates from sea salt and road salt (deicing salt in winter season) (Ayers et al. 1999; Jang et al. 2007), and is also related to Motor vehicle emission (Bourotte et al. 2011). Ni comes mainly from oil combustion, and partially originates from automotive batteries (Jang et al. 2007; Kelly, Dai, and Wang 2020).

Different values of regression coefficients (β_1j in Equation (1)(1) $ln\left(C_{ij}\right)\hspace{0.17em}={\beta }_{0j}+{\gamma }_{ij}+{\beta }_{1j}\times \hspace{0.17em}ln{d_i}_j+{\beta }_{2j}\times season+{\beta }_{3j}\times r{t}_{ij}+e_j$(1) ) as shown in Table 1 indicate that the decreases of mass fractions for different elements in PM_2.5 or PM₁₀ with distance from major road are different. The mass fractions of elements which mainly come from traffic (e.g., Cu and Zn) decrease more sharply with distance from the major road compared to elements (e.g., Fe and Ti) which come from both traffic and additional source (i.e., soil), both in PM₁₀ and PM_2.5 road dust. Figure 3 and 4 show the relationship between log value of relative mass fraction decreases for Fe, Ti, Cu, Zn, Zr, and Ba in PM₁₀ and Fe, Ti, Cu, Zn, Zr in PM_2.5 road dust with the perpendicular distance from the nearest A12 road in warm season. Figure 1 and 2 show the relationship between log value of relative mass fraction decreases with the perpendicular distance from the nearest A12 road in cold season.

Figure 3. The relationship between log value of relative mass fraction decreases for elements in PM₁₀from road dust and the perpendicular distance from the nearest A12 road in warm season. (a) Fe, (b) Ti, (c) Cu, (d) Zn, (e) Zr, (f) Ba

The dash-dot line represents 95% confidence interval.

Figure 4. The relationship between log value of relative mass fraction decreases for elements in PM_2.5from road dust and the perpendicular distance from the nearest A12 road in warm season. (a) Fe, (b) Ti, (c) Cu, (d) Zn, (e) Zr

The dash-dot line represents 95% confidence interval.

No significant relationships were found for elements mostly associated with other sources with distance from major road, in either PM₁₀ or PM_2.5. These elements include: Na (mostly from soil, and road and sea salt), Mg (from soil, ocean, and combustion of motor oil additives: detergent additives), Si (soil), K (from soil and wood burning), V (from oil combustion), and Sr (from crustal) (Cadle et al. 1997; Garg et al. 2000; Gertler et al. 2002; Huang et al. 2018; Kam et al. 2012; Lough et al. 2005; Ning et al. 2008; Root 2000).

Effect of season

The coefficients by season (β_2j in Equation (1)(1) $ln\left(C_{ij}\right)\hspace{0.17em}={\beta }_{0j}+{\gamma }_{ij}+{\beta }_{1j}\times \hspace{0.17em}ln{d_i}_j+{\beta }_{2j}\times season+{\beta }_{3j}\times r{t}_{ij}+e_j$(1) ) are shown in Table S7 in the Supplementary Material. These results show that road dust mass fractions of Ti, Cr, Fe, Ni, Cu, Zn, Zr, and Pb in PM₁₀ and Ti, Cr, Fe, Co, Ni, Cu, and Zr in PM_2.5 were significantly higher in warm season than those in cold season. We found the same results for ambient PM elements and EC/OC. It rains less during warm season. During our experiment period, the average precipitation is 12.3 cm in cold season and 10.9 cm in warm season, according to data from National Centers for Environmental Information (NOAA) (https://www.ncdc.noaa.gov/). Therefore, the elements associated with traffic accumulate faster than during the cold season.

Effects of road type

The coefficients by road type (β_3j in Equation (1)(1) $ln\left(C_{ij}\right)\hspace{0.17em}={\beta }_{0j}+{\gamma }_{ij}+{\beta }_{1j}\times \hspace{0.17em}ln{d_i}_j+{\beta }_{2j}\times season+{\beta }_{3j}\times r{t}_{ij}+e_j$(1) ) are summarized in Table S8 in the Supplementary Material. The coefficients of V, Fe, Co, Cu, and Zn in PM₁₀ road dust and Cu and Zn in PM_2.5 road dust are significant and positive, indicating that the mass fractions of these elements at the site close to A12 roads are higher than those close to A3 roads. The main reason is that these elements are associated with traffic, and A12 roads have larger traffic density compared to A3 roads. However, it shows no significant difference between A12 and A3 roads for ambient PM elements.

Discussion

In this study, we investigated the influence of road proximity on the mass fractions of 20 elements in 78 PM₁₀ and 79 PM_2.5 samples of road dust, which were collected at 62 sites with different distances from 20 major roadways in the greater Boston area. We found that mass fractions of Ba, Cu, Zr, Zn, Cl, Co, Cr, Pb, Ca, Ti, and Fe in the PM₁₀ road dust; and, Zr, Cu, Cl, Zn, Cr, Ti, Mn, Ca, Ni, and Fe in the PM_2.5 road dust decreased significantly with increasing distance from major road. Most of these elements are associated with road traffic emissions.

The mass fractions of elements decrease differently with distance from major road for elements from different sources. According to our results, the mass fractions of elements which mainly come from traffic (e.g., Cu and Zn), decrease more sharply with distance from the major road compared to elements which come from both traffic and soil (e.g., Fe and Ti). Elements mainly associated with other sources, such as Na, Mg, Si, K, V, and Sr are not significantly related to the distance from the major road. These results were similar to those reported for samples of ambient air coarse and fine PM collected during this study. Our results verify that, as expected, Cu and Zn are the most reliable elemental tracers of road traffic indirect emissions.

We also found that season affect the mass fractions of elements in PM₁₀ and PM_2.5 road dust. Ti, Cr, Fe, Ni, Cu, Zn, Zr, and Pb in PM₁₀ and Ti, Cr, Fe, Co, Ni, Cu, and Zr in PM_2.5 were significantly higher in warm season. The possible explanation is that the increased wintertime precipitation reduces the resuspension of the road dust (Amato et al. 2012), and the compressed atmospheric mixing height leads to increases in their concentrations (Wagner and Schäfer 2017). Opposite trends have been observed in other more arid areas like Los Angeles (Pakbin et al. 2009).

Furthermore, our results show that for V, Fe, Co, Cu, and Zn in PM₁₀ road dust and Cu and Zn in PM_2.5 road dust, the mass fractions at the site closer to A12 roads are significantly higher than for those closer to A3 roads. This is expected since the traffic emissions are generally higher on A12 roads than A3 roads, which typically feature higher traffic speed and density.

Elements primarily associated with soil and crustal materials (Al, Si, K) do not show a distance-related decrease in mass fraction in either the PM_2.5 or PM₁₀ road dust. In fact, they show a slight borderline-significant increase, suggesting an increased contribution of soil and a decreased contribution of traffic-derived road dust in the road surface dust sampled at the near-road and background sites compared to the major road. Furthermore, other elements mostly associated with non-traffic related sources (Na, Mg, Si, K, V, and Sr) were not significantly associated with distance from road in either PM_2.5 or PM₁₀ (Cadle et al. 1997; Garg et al. 2000; Gertler et al. 2002; Huang et al. 2018; Kam et al. 2012; Lough et al. 2005; Ning et al. 2008; Root 2000).

Our study has some limitations. Our sample size was smaller than desired due to several factors, including difficulty in scheduling use of road sites where permission was required for access (especially in commercial areas), weather conditions (the roads had to be dry to collect dust from the road surface), and costs. Some collected road dust samples were too heavily loaded to be analyzed by EDXRF, and thus were not included in the data analysis. The resulting small sample size limited statistical power to investigate other possible influencing factors, such as other traffic or roadway characteristics, land use, and meteorological variables. However, even with a small sample size, we were able to observe significant relationships between distance from major road and traffic-related element mass fractions in PM₁₀ and PM_2.5 road dust. In addition, our results were based on experiments in Boston area, which may not be generalizable to other geographical areas with different weather.

Our study also has several strengths. We have developed a new method to collect PM₁₀ and PM_2.5 re-suspended from surface road dust, which results in a more realistic measurement of PM composition and size. To the best of our knowledge, this is the first study to include a wide range of traffic-related elements in both PM₁₀ and PM_2.5 road dust collected at different distances from major roads. Even with a small dataset, we observed significantly negative associations between mass fractions of traffic-related elements in PM_2.5 and PM₁₀ road dust with the log distance from major road.

Conclusion

We collected 78 PM₁₀ and 79 PM_2.5 samples of road dust collected at 62 sites with different distances from 20 major roads in the Boston area. Using a linear mixed effects model, we observed significant decrease of traffic-related elements in PM₁₀ and PM_2.5 road dust with log distance from major road. We also found that the mass fraction for elements, which mainly come from traffic, including the tailpipe emission and non-tailpipe emission, decrease more sharply with distance from the major road compared to elements which come from both traffic and other sources. In addition, elements associated with soil and crustal material do not decrease with distance to major roads; in fact, the elements associated with crustal materials show a small and borderline significant increase in their contribution to PM_2.5 and PM₁₀ road dust. Our results indicate that traffic emissions contribute substantially to road dust. As road dust plays an important role in atmospheric pollution, our results imply that the distance to major road can be used as a proxy for ambient exposure, and therefore, it is a reasonable exposure metric in epidemiological studies. Over the last decades, strict tailpipe emission controls from governmental agencies and technological upgrades have been pushing the automotive industry to a significant reduction of exhaust emissions (Allen et al. 2006). Our results implicated that government and future research should also pay particular attention to non-tailpipe emissions.

Acknowledgment

Research described in this article was conducted under contract to the Health Effects Institute (HEI), an organization jointly funded by the United States Environmental Protection Agency (U.S. EPA) (Assistance Award No. CR-83467701) and certain motor vehicle and engine manufacturers. The contents of this article do not necessarily reflect the views of HEI, or its sponsors, nor do they necessarily reflect the views and policies of the U.S. EPA or motor vehicle and engine manufacturers. This publication was made possible by U.S. EPA grant RD-83587201. Its contents are solely the responsibility of the grantee and do not necessarily represent the official views of the U.S. EPA. Further, U.S. EPA does not endorse the purchase of any commercial products or services mentioned in the publication.

Disclosure statement

The authors do not have any competing financial interests, including relevant financial interests, activities, relationships, and affiliation.

Supplementary material

Supplemental data for this article can be accessed on the publisher’s website.

Additional information

Funding

This work was supported by the Health Effects Institute [CR-83467701]; U.S. Environmental Protection Agency [RD-83587201].

Notes on contributors

Shaodan Huang

Shaodan Huang is Ph.D., Research fellow, Department of Environmental Health, Harvard T.H. Chan School of Public Health, Boston, MA, USA. Research area: Indoor Air Quality, Ambient Air Pollution, Health Effect of Air Pollution.

Pablo Taddei

Pablo Taddei is Master student, Department of Environmental Health, Harvard T.H. Chan School of Public Health, Boston, MA, USA. Research area: Population Health Science.

Joy Lawrence

Joy Lawrence is Research associate, Department of Environmental Health, Harvard T.H. Chan School of Public Health, Boston, MA, USA. Research area: Air Pollution, Ambient Environment.

Marco A.G. Martins

Marco A.G. Martins is Ph.D., Research associate, Department of Environmental Health, Harvard T.H. Chan School of Public Health, Boston, MA, USA. Research area: Air Pollution, Ambient Environment.

Jing Li

Jing Li is Ph.D., Research fellow, Department of Environmental Health, Harvard T.H. Chan School of Public Health, Boston, MA, USA. Research area: Ambient Air Pollution, Atmospheric Chemistry.

Petros Koutrakis

Petros Koutrakis is Ph.D., Professor, Department of Environmental Health, Harvard T.H. Chan School of Public Health, Boston, MA, USA. Research area: Air Pollution and its Health Effects.