This study showcases the chemical fractionation and primary source identification of Cu, Zn and Cd in road-deposited sediment. Results show that Zn and Cd were identified as easily mobilised and biologically available metals that pose high risks to the receiving aquatic environments. However, Cu was released in substantial amounts only under relatively extreme conditions. With the assistance of principal component analysis and chemical fractionation, two primary contributors to heavy metals in road-deposited sediment were tentatively identified as vehicle-related sources (especially, auto brake pad erosion for Cu and tyre debris for Zn) and atmospheric deposition (especially for Cd).

INTRODUCTION

With the rapid development of the economy and society, the process of urbanisation has increasingly caused urban runoff pollution problems in many countries (Sartor & Boyd 1972). Natural and anthropogenic pollutants are washed off urban watersheds and transported through sewers or directly into lakes, rivers and seas, ultimately contaminating receiving waters (Hillenbrand et al. 2005; Egodawatta et al. 2013). In this context, the urban watershed has been recognised as a leading diffuse pollution source in urban areas worldwide (Sartor & Boyd 1972; Loganathan et al. 2013). Therefore, a mitigation strategy for stormwater pollution should be based on the in-depth understanding of initial pollutant build-up on urban watersheds (Egodawatta et al. 2013; Zhang et al. 2015).

Among the priority pollutants abundant in both the urban watershed and stormwater runoff, heavy metals are known to be acutely toxic, carcinogenic, and persistent trace elements (Sartor & Boyd 1972). The knowledge on the metals in road-deposited sediment (RDS) is essential to assess the potential adverse effects of RDS on the stormwater quality and the potential performance of an integrated stormwater pollution management in eliminating heavy metals.

It has been reported that heavy metal concentrations in RDS are commonly at least an order of magnitude higher than those in neighbouring soils (Loganathan et al. 2013). However, the environmental and health effects of metals are dependent, at least initially, on the mobility and availability of their chemical fractionation, e.g. exchangeable, reducible, oxidisable, and residual fractions determined by a sequential extraction protocol (Davidson et al. 1998; Sutherland et al. 2012). The chemical fractionation of heavy metals in solid-phase concentration (mass of metals per unit mass of dry RDS) has been reported (Sutherland et al. 2012). However, less attention has been paid to the mass of heavy metal chemical fractionation per unit surface area, referred to as surface load, which describes the accumulation ability of RDS adsorbed toxic substances on urban watersheds and the total mass contribution to receiving aquatic environments (Lau & Stenstrom 2005).

Therefore, the focuses of this study were to (i) determine the chemical fractionation of RDS adsorbed heavy metals in the form of surface load and (ii) identify the primary source contributors of these metals.

MATERIAL AND METHODS

Sample collection

RDS samples were obtained from three urban roads with various surrounding land use types and traffic loads in the city of Dresden (51°02′55″ N, 13°44′29″ E), Germany. The characteristics of the sampling sites are shown in Table 1. The sampling campaign was performed in July 2012. A professional vacuum sweeper (Puzzi 100 Super, Kärcher) with a water filtration system was employed for the sample collection. It has been proven to be more efficient in removing the fine materials within a typical pavement structure than mechanical broom sweepers and traditional domestic vacuum sweepers (Sartor & Boyd 1972; Egodawatta et al. 2013). The power requirement was provided by a generator (Honda EU30i, Rated power COP, 2.6 kW).

Table 1

The characteristics of sampling sites

     Average daily traffica Heavy traffic   
Site Abbr. Road class Pavement quality Lane number vehicle/day Vicinity land use 
Walpurgis str. Main road Good 12,600 3–4 City commercial centre 
Bannewitz Federal highway Average 15,900 Petrol and bus station 
Noethnitz Secondary road Good 50 <1 Rural residential area 
     Average daily traffica Heavy traffic   
Site Abbr. Road class Pavement quality Lane number vehicle/day Vicinity land use 
Walpurgis str. Main road Good 12,600 3–4 City commercial centre 
Bannewitz Federal highway Average 15,900 Petrol and bus station 
Noethnitz Secondary road Good 50 <1 Rural residential area 

aThe traffic loads were determined in one flow direction (www.dresden.de, accessed July 2012).

For each land use, six sampling plots of 6 m2 in close proximity were predefined. Since the majority of the sediments were assembled near roadside kerbs (Sartor & Boyd 1972), all the enclosed sampling plots were situated next to these areas. The RDS build-up process in each plot started at the same time period after the last substantial rainfall event. After the stipulated antecedent dry-weather period (ADP, the length of time which has elapsed since the last substantial rainfall event), i.e. 1, 2, 3, 4, 6, and 8 days, 13.9–72.7 g RDS were obtained from the respective sampling plots. During the sampling process, the vacuum nozzle was moved from the kerb to the bottom of the enclosed sampling plots at approximately 10 cm/s in parallel strips.

Due to the water filtration system used by the vacuum sweeper, the collected material was suspended in deionised water, which was used as a filtration reagent. The suspended RDS was then deposited into pre-cleaned 1 L amber-brown glass bottles with Bakelite caps before being transported to the laboratory. During the sample collection, around 0.5–0.8 L/m2 deionised water was used. After sample collections at each site, about 6 L deionised water was passed through the sweeper three times to prevent cross-contamination. The vacuum nozzle and brush were cleaned with deionised water to remove all the remaining particles. Every effort was made to ensure consistent initial conditions for each RDS collection.

Laboratory analysis

RDS samples obtained from each site were fractionated into sub-samples and wet-sieved using nylon and stainless-steel sieves with mesh sizes of 1,000 μm, 400 μm, 100 μm and 63 μm in sequence. The particles passing through the 63 μm sieve were filtered by a 0.45 μm (classification diameter between soluble and particle) cellulose nitrite filter. The fractionated RDS samples were labelled, sealed and refrigerated at 4 °C in preparation for analysis.

A total of 68 bulk and size-fractionated RDS samples were tested for copper (Cu), zinc (Zn), and cadmium (Cd). The contents of metals were determined according to both a three-step sequential extraction protocol proposed by the Standards, Measurements and Testing Programme (SM&T, formerly BCR) of the European Commission, which has proved reproducible and gave good recoveries with respect to acid dissolution (Davidson et al. 1998) and a pseudo-total digestion protocol following the German Norm DIN 38406-8 (E8), DIN 38406-7 (E7), and DIN EN ISO 5961. The chemical extraction reagents and experimental protocol are given in Table 2. More details can be found elsewhere (Quevauviller et al. 1997).

Table 2

Three-step sequential extraction (SM&T) and pseudo-total digestion protocols

Step Fraction Nominal target heavy metals Chemical reagents and experimental protocols 
F1 Exchangeable Exchangeable, carbonate bound A total of 40 mL 0.11 mol/L acetic acid (CH3COOH, pH 2.85) was added to 1.0 g dry weight of RDS samples and shaken for 16 h. The mixture was centrifuged to separate the extract from the residue at 3,000 rpm for 20 min. 
F2 Reducible Iron, manganese oxide bound A total of 40 mL 0.1 mol/L hydroxylammonium chloride (NH2OH·HCl, pH 1.5) was added to the residue from step 1 and the extraction performed as above. 
F3 Oxidisable Organically, sulphide bound A total of 10 mL hydrogen peroxide (H2O2, 30%) was added to the residue from step 2 at room temperature for 1 h, and then evaporated at 85 ± 2 °C to dryness. Additional 50 mL 1 mol/L ammonium acetate (NH4CH3COO, pH 2) was added and shaken for 16 h. The extraction performed as above. 
F4 Residual* Remaining, non-silicate bound With microwave assistance, the remaining material after step 3 was digested with 5 mL nitric acid (HNO3), 5 mL ultrapure water and 1 mL hydrogen peroxide (H2O2, 30%). The digested samples were filtered through the cellulose nitrate membrane filter into a 50 mL Erlenmeyer flask and the leftover particles were removed. 
Step Fraction Nominal target heavy metals Chemical reagents and experimental protocols 
F1 Exchangeable Exchangeable, carbonate bound A total of 40 mL 0.11 mol/L acetic acid (CH3COOH, pH 2.85) was added to 1.0 g dry weight of RDS samples and shaken for 16 h. The mixture was centrifuged to separate the extract from the residue at 3,000 rpm for 20 min. 
F2 Reducible Iron, manganese oxide bound A total of 40 mL 0.1 mol/L hydroxylammonium chloride (NH2OH·HCl, pH 1.5) was added to the residue from step 1 and the extraction performed as above. 
F3 Oxidisable Organically, sulphide bound A total of 10 mL hydrogen peroxide (H2O2, 30%) was added to the residue from step 2 at room temperature for 1 h, and then evaporated at 85 ± 2 °C to dryness. Additional 50 mL 1 mol/L ammonium acetate (NH4CH3COO, pH 2) was added and shaken for 16 h. The extraction performed as above. 
F4 Residual* Remaining, non-silicate bound With microwave assistance, the remaining material after step 3 was digested with 5 mL nitric acid (HNO3), 5 mL ultrapure water and 1 mL hydrogen peroxide (H2O2, 30%). The digested samples were filtered through the cellulose nitrate membrane filter into a 50 mL Erlenmeyer flask and the leftover particles were removed. 

*Digestion of the residual material is not a speciation of the SM&T three-step sequential extraction protocol, and is referred to as pseudo-total digestion.

The metal compounds of each fraction were determined in sequence by an atomic absorption spectrophotometer equipped with graphite furnace atomisation (Varian, SpectrAA 220 Fast Sequential AAS-220 Z). The samples were processed in triplicate throughout all the analyses and the results reported were the average values. To maintain the laboratory quality assurance and quality control, Cu (1.014 g/cm³, 20 °C), Zn (1.02 g/cm³, 20 °C), Cd (1.013 g/cm³, 20 °C) and multi-element standard solutions (23 elements in diluted nitric acid, 1.09 g/cm³, 20 °C), for each at 1,000 mg/L, were prepared (Merck KGaA, Darmstadt, Germany). A certified reference material of dry sewage sludge (KS 2/2006) from a European Union reference laboratory (Sächsische Landesanstalt für Landwirtschaft. Fachbereich 8 – Landwirtschaftliches Untersuchungswesen, Leipzig, Germany) was used. The recovery (the summation of the four fractions to the pseudo-total metals concentration) were in the range 98–116% for Cu, 109–124% for Zn, and 92–115% for Cd. The limits of detection were 1.0 μg/L for Cu, 20 μg/L for Zn, and 0.1 μg/L for Cd. The limits of quantitation were 0.5 μg/L for Cu, 10 μg/L for Zn, and 0.05 μg/L for Cd.

Source apportionment

Principal component analysis (PCA) is one of the most classical source receptor approaches, which has been widely used to qualitatively identify the major sources of various environmental pollutants (Zhang et al. 2013). Its aim is to reduce a larger set of variables into a smaller set of artificial variables, referred to as principal components, which account for most of the variance in the original variables. In this study, a PCA receptor model with the assistance of chemical fractionation data was employed to qualitatively estimate the potential source apportionment of heavy metals.

RESULTS AND DISCUSSION

Heavy metal surface load in bulk RDS

As given in Table 3, the mean values of overall total surface load were: Cu = 0.85 ± 0.19 mg/m2 (mean ± standard error of the mean), Zn = 1.70 ± 0.31 mg/m2, and Cd = 1.16 ± 0.14 μg/m2. Compared with data from Queensland, Australia (Egodawatta et al. 2013), Cu overall mean surface load in the present study was high, and Zn overall mean surface load was comparable. Compared with data from Christchurch, New Zealand (Wicke et al. 2012), the presented Cu overall mean surface load was high, and the presented Zn mean overall surface load was low. All the presented metal contents were lower than the data from Santa Monica, California, USA (Lau & Stenstrom 2005).

Table 3

The present and reviewed heavy metal surface loads in bulk RDS

  Mean surface load
 
   
Study area Cu (mg/m2Zn (mg/m2Cd (μg/m2No.a Land use Year 
Present study 1.40 2.86 1.60 68 Commercial 2012 
0.88 1.64 0.68  Highway 
0.15 0.39 1.22  Rural 
0.85 1.70 1.16  Overall 
Queensland, Australia (Egodawatta et al. 20130.15 1.43 n/ab 17 Residential 2013 
Christchurch, New Zealand (Wicke et al. 20120.21 2.72 n/a 11 Coarse asphalt 2012 
0.15 2.90 n/a  Smooth asphalt 
0.18 2.81 n/a  Overall 
Santa Monica, California, USA (Lau & Stenstrom 20054.10 16.71 53 18 Commercial 2005 
0.63 3.80 18  Residential 
2.68 6.44 33  Industrial 
2.47 8.98 35  Overall 
  Mean surface load
 
   
Study area Cu (mg/m2Zn (mg/m2Cd (μg/m2No.a Land use Year 
Present study 1.40 2.86 1.60 68 Commercial 2012 
0.88 1.64 0.68  Highway 
0.15 0.39 1.22  Rural 
0.85 1.70 1.16  Overall 
Queensland, Australia (Egodawatta et al. 20130.15 1.43 n/ab 17 Residential 2013 
Christchurch, New Zealand (Wicke et al. 20120.21 2.72 n/a 11 Coarse asphalt 2012 
0.15 2.90 n/a  Smooth asphalt 
0.18 2.81 n/a  Overall 
Santa Monica, California, USA (Lau & Stenstrom 20054.10 16.71 53 18 Commercial 2005 
0.63 3.80 18  Residential 
2.68 6.44 33  Industrial 
2.47 8.98 35  Overall 

aNo.: number of observations.

bn/a: not available.

With regard to land use, the urban commercial centre of site W had the highest mean surface loads, which was consistent with the findings of Lau & Stenstrom (2005) that the commercial area had higher metal surface loads than most of the other land uses. A plausible explanation is the prevailing anthropogenic and traffic activities at site W. In contrast, the rural area of site N which was only slightly influenced by traffic or anthropogenic activities had a very low value of mean surface loads.

Heavy metal surface load in size-fractionated RDS

As given in Table 4, the mean percentages of size fraction apportioned to total surface load were: Cu = 63–0.45 μm (39.8%), > 400–100 μm (32.6%) > 1,000–400 μm (15.6%) > 100–63 μm (12.0%); Zn = 63–0.45 μm (46.1%) > 400–100 μm (29.6%) > 1,000–400 μm (12.9%) > 100–63 μm (11.3%); and Cd = 63–0.45 μm (53.8%) > 400–100 μm (23.9%) > 100–63 μm (12.0%) > 1,000–400 μm (10.3%). It is obvious that the highest mean surface load dominated in the finest fraction of 63–0.45 μm for each metal, which is consistent with the common surface loads apportioning to size fractions (Lau & Stenstrom 2005; Sutherland et al. 2012; Loganathan et al. 2013). A possible explanation could be that the parameters favourable for metal adsorption to RDS, such as specific surface area, organic carbon content, effective cation exchange capacity and clay-forming mineral content, decrease with the increase in particle size. Therefore, special concern should be paid to the finer sediments, which can stay in suspension longer, be transported further by runoff than can larger sediments, and eventually result in a large amount of pollutant contribution to stormwater runoff.

Table 4

Descriptive statistic of heavy metal surface loads in size-dependent RDS

  Cu (mg/m2)
 
Zn (mg/m2)
 
Cd (μg/m2)
 
Study area No.a Mean SDb Mean SD Mean SD 
Commercial W 
1,000–400 μm 0.16 0.13 0.32 0.14 0.19 0.12 
400–100 μm 0.57 0.27 1.07 0.52 0.47 0.27 
100–63 μm 0.19 0.073 0.33 0.12 0.21 0.14 
63–0.45 μm 0.47 0.23 1.14 0.47 0.72 0.28 
Highway B 
1,000–400 μm 0.34 0.56 0.31 0.34 0.071 0.039 
400–100 μm 0.24 0.23 0.54 0.30 0.18 0.11 
100–63 μm 0.066 0.024 0.18 0.041 0.077 0.022 
63–0.45 μm 0.23 0.11 0.61 0.26 0.35 0.12 
Rural N 
1,000–400 μm 0.017 0.0083 0.049 0.018 0.12 0.037 
400–100 μm 0.046 0.038 0.077 0.029 0.22 0.049 
100–63 μm 0.014 0.006 0.035 0.0066 0.13 0.019 
63–0.45 μm 0.070 0.032 0.23 0.078 0.76 0.085 
Overallc 
1,000–400 μm 17 0.18 0.35 0.24 0.24 0.13 0.092 
400–100 μm 17 0.30 0.30 0.59 0.53 0.29 0.21 
100–63 μm 17 0.096 0.089 0.19 0.14 0.14 0.10 
63–0.45 μm 17 0.27 0.22 0.68 0.49 0.60 0.063 
  Cu (mg/m2)
 
Zn (mg/m2)
 
Cd (μg/m2)
 
Study area No.a Mean SDb Mean SD Mean SD 
Commercial W 
1,000–400 μm 0.16 0.13 0.32 0.14 0.19 0.12 
400–100 μm 0.57 0.27 1.07 0.52 0.47 0.27 
100–63 μm 0.19 0.073 0.33 0.12 0.21 0.14 
63–0.45 μm 0.47 0.23 1.14 0.47 0.72 0.28 
Highway B 
1,000–400 μm 0.34 0.56 0.31 0.34 0.071 0.039 
400–100 μm 0.24 0.23 0.54 0.30 0.18 0.11 
100–63 μm 0.066 0.024 0.18 0.041 0.077 0.022 
63–0.45 μm 0.23 0.11 0.61 0.26 0.35 0.12 
Rural N 
1,000–400 μm 0.017 0.0083 0.049 0.018 0.12 0.037 
400–100 μm 0.046 0.038 0.077 0.029 0.22 0.049 
100–63 μm 0.014 0.006 0.035 0.0066 0.13 0.019 
63–0.45 μm 0.070 0.032 0.23 0.078 0.76 0.085 
Overallc 
1,000–400 μm 17 0.18 0.35 0.24 0.24 0.13 0.092 
400–100 μm 17 0.30 0.30 0.59 0.53 0.29 0.21 
100–63 μm 17 0.096 0.089 0.19 0.14 0.14 0.10 
63–0.45 μm 17 0.27 0.22 0.68 0.49 0.60 0.063 

aNo.: number of observations.

bSD: standard deviation.

cOverall: indicates the contents regardless of land use (from all the sampling sites).

Heavy metal chemical fractionation in bulk RDS

The chemical fractionations of metals are shown in Figure 1. It shows that Cu presented considerable proportions in residual and oxidisable fractions. The proportion order of the four fractions decreased as residual (44.0 ± 4.22%) > oxidisable (41.1 ± 3.00%) > exchangeable (11.7 ± 1.36%) > reducible (3.3 ± 0.32%). The partitioning of Cu was comparable to a similar study performed at Murcia, Spain (Acosta et al. 2014). In terms of potential ecological risk assessment, the dominant Cu contents in residual fractions indicate that Cu contents in RDS were relatively stable and would not show significant transformation under various conditions when discharged into receiving waters by stormwater runoff. Meanwhile, Cu contents in the oxidisable fraction may be associated with various forms of organic material. The high affinity of Cu to organic matters should be due to the fact that Cu can easily complex with organic matter to form the high stability complexes of organic-Cu compounds (Sundaray et al. 2011; Li et al. 2013). These compounds could pose a comparably low ecological risk to the receiving waters, but would exist for longer periods. In terms of the build-up phenomenon, stable Cu components of residual and oxidisable fractions increased with an increasing duration of ADP at sites W and B.

Figure 1

Build-up patterns and chemical fractionation of Cu, Zn and Cd in bulk RDS.

Figure 1

Build-up patterns and chemical fractionation of Cu, Zn and Cd in bulk RDS.

The exchangeable fraction (49.1 ± 2.34%) was the dominant phase of Zn, and followed the fraction order of residual (32.5 ± 3.01%) > oxidisable (9.22 ± 1.01%) > reducible (9.21 ± 0.51%). The results were consistent with the chemical fractionation pattern documented in the literature (Sutherland et al. 2012). As for the potential ecological risk assessment, metal content determined by the exchangeable extraction is considered to provide a reasonable approximation of the easily mobilised and biologically available metal content in RDS. The higher the proportion of metals in this fraction, the more mobile they are, and the higher the risk to the receiving aquatic environment (Sutherland et al. 2012). Therefore, the dominant Zn contents in the exchangeable fraction indicates that Zn contents in RDS could pose a considerably high ecological risk to aquatic biota when discharged into receiving waters. The results were consistent with previous findings in literature that Zn would be one of the most mobile and bioavailable of the anthropogenically impacted elements (Sutherland et al. 2012; Loganathan et al. 2013). In addition, Zn contents in the exchangeable fraction contributed significantly to the overall increasing tendency.

The Cd content was characteristically enriched in the exchangeable fraction. The mean proportion order of the four fractions decreased as exchangeable (59.0 ± 1.09%) > oxidisable (17.2 ± 1.35%) > reducible (19.4 ± 1.18%) > residual (4.4 ± 0.50%). The dominant content in the exchangeable fraction suggests that Cd is one of the most susceptible metals to be mobilised during stormwater runoff. Additionally, a comparably clear build-up pattern of Cd was found at site B, and was considerably attributed to the exchangeable fraction.

Heavy metal chemical fractionation in size-dependent RDS

To determine the mass percentage of chemical fractionation (P) in size-dependent RDS, according to Sutherland et al. (2012), the modified form of size fraction loading was applied in this study: 
formula
1
where Ci is the surface load of metal in size fraction i in one of the four sequentially extracted chemical fractionations; Mi is the dry mass percentage of size fraction i; and m is the number of given metal. Therefore, for a given metal, the summation of the four chemical fractionations in each size fraction = 100%.

As given in Figure 2, Cu surface loads in oxidisable and residual fractions were dominant due to the considerable contributions from the coarse size fractions (1,000–400 and 400–100 μm). All the size fractions contributed insignificantly to the unstable components of exchangeable and reducible fractions (ranging from 6.5–0.2%). Therefore, a considerable proportion of Cu in stormwater runoff could be more easily controlled and mitigated by removing coarse RDS from impervious surfaces.

Figure 2

Size-dependent percentage contributions to heavy metal chemical fractionations.

Figure 2

Size-dependent percentage contributions to heavy metal chemical fractionations.

In terms of Zn and Cd, surface load in 63–0.45 μm dominated all chemical fractionations and especially the exchangeable fraction. This is notable because this is the size fraction most widely digested and assimilated by aquatic biota, especially benthic feeders. Therefore, the management of fine size fractions is a key issue in the potential mitigation of Zn and Cd in stormwater runoff.

Heavy metal primary source identification

The source identification of RDS-associated heavy metals is a complex research area. Heavy metals have a broad range of potential sources. Therefore, it is necessary to identify the primary contributors in order to facilitate source-oriented mitigation strategies for stormwater pollution. PCA was used to qualitatively identify the major sources of heavy metals. The purpose of PCA is to reduce the dimensionality of the original heavy metal data to a minimum number of principal components. Each principal component is a linear combination of the original variables. All the components are orthogonal to each other, which results in the smallest possible covariance. The principal components were computed using the varimax rotation method with Kaiser normalisation by SPSS Statistics 19. More details regarding analysis setting and other technical usage of PCA can be found elsewhere (Zhang et al. 2013; Banerjee et al. 2015).

According to the component matrix in Table 5, two principal components (potential sources) were extracted. The sum of the variances of the first two components exceeded 98.73% of the total variance of the original data. The first component was responsible for 78.22% of the total variance. This component is heavily weighted by Cu and Zn. According to the source fingerprints (tracers) reviewed from literature, the dominant factor loadings of Cu and Zn have been suggested for vehicle-related sources (Sutherland et al. 2012; Loganathan et al. 2013). More explicitly, the elevated levels of Cu and Zn were usually regarded as source fingerprints of auto brake pad erosion and tyre debris, respectively (Banerjee et al. 2015).

Table 5

Component matrix of heavy metals

  Component
 
Item 
Cu 0.941 −0.311 
Zn 0.967 −0.214 
Cd 0.726 0.688 
% of variancea 78.22 20.51 
Cumulative %b 78.22 98.73 
Potential source vehicle associated atmospheric deposition 
  Component
 
Item 
Cu 0.941 −0.311 
Zn 0.967 −0.214 
Cd 0.726 0.688 
% of variancea 78.22 20.51 
Cumulative %b 78.22 98.73 
Potential source vehicle associated atmospheric deposition 

aThe percentage of the total variability (initial eigenvalues) explained by each principal component.

bThe amount of variance accounted for by each component.

In addition, the data of chemical fractionation further assist the potential source identification. As shown in Figures 1 and 2, Cu presented considerable proportions in residual and oxidisable fractions. In fact, Cu (Cu0) is a component of brake pads with content up to 16% (www.copper.org, accessed July 2013). The considerable proportions of Cu in oxidisable and residual fractions are consistent with the Cu (Cu0) speciation in brake pads, and further confirms the Cu contributor of brake pad erosion.

The exchangeable fraction was the dominant phase of Zn. Zn is added as Zn oxide (ZnO) to the tyre rubber as an activator to accelerate vulcanisation when manufacturing tyres. The tyre rubber contains 1.2% of ZnO for car tyres (min 0.4%, max 2.9%) and 2.1% (min 1.2%, max 4.3%) for truck tyres (Smolders & Degryse 2002). The high proportion of Zn in the exchangeable fraction and ZnO fractionation in tyre debris are highly linked, and further confirms the Zn contributor of tyre debris. Consequently, this component was tentatively attributed to the vehicle-associated sources.

The second component was responsible for 20.51% of the total variance. This component was characterised by high loading of Cd. The source apportionment of Cd is a much debated subject. Muhammad et al. (2012) found that Zn and Cd share a similar geochemical behaviour and often occur together as environmental pollutants in soils, sediments, plants, and other endpoints. However, Cu and Zn had negative loadings in this component, suggesting that Cd might have alternative sources compared with Cu and Zn.

Saeedi et al. (2012) reported that Cd in RDS does not have a local anthropogenic source. Fujiwara et al. (2011) stated that Cd is a component of petroleum and its concentration depends on the origin of petroleum, and generally it can be found at sub-ppm levels. Davis et al. (2001) have stated that the largest contributor (60%) to Cd pollution was atmospheric deposition. Banerjee et al. (2015) reported that Cd was usually considered as a source fingerprint of vehicular and industrial emissions which are eventually released into the atmosphere. Accordingly, this component appeared to be highly indicative of atmospheric deposition attributed to the emissions of vehicles and/or industries.

In addition, the Cd content was characteristically enriched in the exchangeable fraction as shown in Figures 1 and 2, which was consistent with the chemical fractionation found in urban atmospheric aerosols whereby dominant components of Cd were in the exchangeable forms (Sun et al. 2014; Li et al. 2015). It would further lead to the suggestion that atmosphere deposition might be an alternative source for this component.

CONCLUSIONS

In this study, chemical fractionation determination and primary source identification of heavy metals in bulk and size-dependent RDS were conducted. Results show that the highest surface loads of heavy metals were consistently found in the finest size fraction of 63–0.45 μm. Zn and Cd were identified predominantly in exchangeable fractions, especially in 63–0.45 μm fraction, which can easily be mobilised and biologically incorporated. Cu posed a comparable low risk due to the high proportions of residual and oxidisable fractions. With the assistance of PCA and chemical fractionation, vehicle-related sources and atmospheric deposition were identified as primary source contributors to heavy metals in RDS. More explicitly, the potential source of Cu in RDS was identified as auto brake pad erosion. Zn was significantly attributed to tyre debris. The atmospheric deposition was tentatively identified as a significant source contributor for Cd.

The results reported herein could assist the potential integrated stormwater management designed to remove hazardous materials from roads, and could be used for source-oriented mitigation of rainfall-induced urban runoff pollution.

ACKNOWLEDGEMENTS

The authors would like to gratefully thank Mr Gregor Pfalz for his contribution to the sample collection. Thanks to Mrs Dr Heike Brückner, Mrs Ulrike Gebauer, Mrs Sinaida Heidt and Miss Yuan Wang for their assistance with laboratory analysis. The authors also gratefully acknowledge the state-sponsored scholarship program provided by China Scholarship Council (CSC) (No. 2010605020 and No. 2010605021). This work was also jointly supported by Completion Grant (PSP-Element: F-003661-553-52A-2330000) provided by the Graduate Academy of the Technische Universität Dresden by means of the Excellence Initiative by the German Federal and State Governments. Mention of trade names or commercial products does not constitute endorsement or recommendation for use.

REFERENCES

REFERENCES
Acosta
J. A.
Faz
A.
Kalbitz
K.
Jansen
B.
Martinez-Martinez
S.
2014
Partitioning of heavy metals over different chemical fraction in street dust of Murcia (Spain) as a basis for risk assessment
.
Journal of Geochemical Exploration
144
,
298
305
.
Banerjee
T.
Murari
V.
Kumar
M.
Raju
M. P.
2015
Source apportionment of airborne particulates through receptor modeling: Indian scenario
.
Atmospheric Research
164–165
,
167
187
.
Egodawatta
P.
Ziyath
A. M.
Goonetilleke
A.
2013
Characterising metal build-up on urban road surfaces
.
Environmental Pollution
176
,
87
91
.
Fujiwara
F.
Rebagliati
R. J.
Dawidowski
L.
Gomez
D.
Polla
G.
Pereyra
V.
Smichowski
P.
2011
Spatial and chemical patterns of size fractionated road dust collected in a megacitiy
.
Atmospheric Environment
45
(
8
),
1497
1505
.
Hillenbrand
T.
Toussaint
D.
Böhm
E.
Fuchs
S.
Scherer
U.
Rudolphi
A.
Hoffmann
M.
Kreißig
J.
Kotz
C.
2005
Discharges of Copper, Zinc and Lead to Water and Soil-Analysis of the Emission Pathways and Possible Emission Reduction Measures
,
Report 202 242 20/02
,
German Federal Ministry for the Environment, Nature Conservation, and Nuclear Safety
,
Berlin
.
Lau
S. L.
Stenstrom
M. K.
2005
Metals and PAHs adsorbed to street particles
.
Water Research
39
(
17
),
4083
4092
.
Loganathan
P.
Vigneswaran
S.
Kandasamy
J.
2013
Road-deposited sediment pollutants: a critical review of their characteristics, source apportionment, and management
.
Critical Reviews in Environmental Science and Technology
43
(
13
),
1315
1348
.
Quevauviller
P.
Rauret
G.
LopezSanchez
J. F.
Rubio
R.
Ure
A.
Muntau
H.
1997
Certification of trace metal extractable contents in a sediment reference material (CRM 601) following a three-step sequential extraction procedure
.
Science of The Total Environment
205
(
2–3
),
223
234
.
Sartor
J. D.
Boyd
G. B.
1972
Water Pollution Aspects of Street Surface Contaminants
.
US Environmental Protection Agency
,
Washington, DC
.
Smolders
E.
Degryse
F.
2002
Fate and effect of zinc from tires debris in soil
.
Environmental Science and Technology
36
(
17
),
3706
3710
.
Wicke
D.
Cochrane
T. A.
O'Sullivan
A.
2012
Build-up dynamics of heavy metals deposited on impermeable urban surfaces
.
Journal of Environmental Management
113
,
347
354
.