Trace metal sediment loading in the Mill Creek: A spatial and temporal analysis of vehicular pollutants in suburban waterways

Trace metal sediment loading in the Mill Creek: A spatial and temporal analysis of vehicular pollutants in suburban waterways

Applied Geochemistry xxx (2017) 1e12 Contents lists available at ScienceDirect Applied Geochemistry journal homepage: www.elsevier.com/locate/apgeoc...

3MB Sizes 0 Downloads 12 Views

Applied Geochemistry xxx (2017) 1e12

Contents lists available at ScienceDirect

Applied Geochemistry journal homepage: www.elsevier.com/locate/apgeochem

Trace metal sediment loading in the Mill Creek: A spatial and temporal analysis of vehicular pollutants in suburban waterways Alexander G. Sebastiao, Eric J. Wagner, Steven T. Goldsmith* Department of Geography and the Environment, Villanova University, Villanova, PA 19085, USA

a r t i c l e i n f o

a b s t r a c t

Article history: Received 30 June 2016 Received in revised form 2 April 2017 Accepted 4 April 2017 Available online xxx

Vehicular wash-off introduces toxic heavy metals into riverine environments, but the delivery patterns of such contaminants are normally difficult to observe. This study examines the spatial and temporal heterogeneity in metal loading of riverbed sediments at two river fords in the Mill Creek Watershed located in suburban Philadelphia, Pennsylvania. Riverbed sediment samples were collected over the course of the 2015 driving season (April to December) upstream and downstream of the two fords, which greatly differ in traffic density, and analyzed for their trace metal concentrations (As, Cd, Cu, Ni, Pb, Zn) and percent particulate organic carbon (%POC). Overall, metal concentrations were in the upper range of literature values for road-derived sediments. Comparing the sites, the more heavily trafficked ford had significantly higher concentrations, confirming the role of traffic density on metal accumulation. Highly enriched concentrations of Cu and Zn coupled with diagnostic metal:Cu ratios points to the strong influence of brake lining dust. While an increase in metal concentration was observed over the driving season, the highly variable spatial and temporal signal suggests other factors such as local geomorphological and hydrological conditions also play a role in metal accumulation and dispersal. Metal concentrations in excess of levels known to impact aquatic ecosystems, coupled with strong correlations between metal concentrations and the %POC content of the sediments, suggest vehicular-derived metals may play a role in the historically observed decrease in macroinvertebrate diversity in the watershed. These findings add to a growing body of literature highlighting the potential for vehicular and roadsourced metals to persist and potentially impact riverine environments. © 2017 Elsevier Ltd. All rights reserved.

1. Introduction The rapidly growing field of urban geochemistry has centered on improving our understanding of the link between urbanization and elemental cycling, particularly with respect to waterways (Lyons and Harmon, 2012; Chambers et al., 2016). Road derived trace metals are one type of contaminant, which can be sourced either from the physical components of automobiles, their accompanying oils and lubricants, the associated urban infrastructure, and/or atmospheric deposition. Once deposited on roadways, these metals are readily transported in either the dissolved or particulate form during hydrological events to adjacent river systems via stormwater infrastructure. The marked persistence of these trace metals in sediments can lead to many deleterious impacts to aquatic ecosystems (Paul and Meyer, 2008).

* Corresponding author. E-mail address: [email protected] (S.T. Goldsmith).

Several recent studies have evaluated the delivery of automobile related trace metals to the environment (Although there are many additional sources of trace metals in urban environments this study solely focuses on those of vehicular origin). For example, motor oils and lubricants have been shown to be a predominant source of dissolved arsenic, cadmium, copper, lead, nickel, and zinc (Gourgouillon et al., 2000; Davis et al., 2001; Jonsson et al., 2002). Frictional wear associated with the routine use of brake pad linings and tires has been shown to be a major source of particulate copper and zinc, respectively (Callender and Rice, 2000; Davis et al., 2001). Other key car components (i.e., bearings, batteries, electronics etc.) have also been linked to a suite of trace metals (Sansalone and Buchberger, 1997; Davis et al., 2001; Jonsson et al., 2002). Additionally, road-related infrastructure such as concrete can be a source of zinc (Davis et al., 2001) while lead chromate has been utilized in road paint (Kim et al., 1998). Finally, historical use of lead as an anti-knock additive in gasoline has substantially contributed to legacy contamination in urban soils and sediment (Mielke et al.,

http://dx.doi.org/10.1016/j.apgeochem.2017.04.001 0883-2927/© 2017 Elsevier Ltd. All rights reserved.

Please cite this article in press as: Sebastiao, A.G., et al., Trace metal sediment loading in the Mill Creek: A spatial and temporal analysis of vehicular pollutants in suburban waterways, Applied Geochemistry (2017), http://dx.doi.org/10.1016/j.apgeochem.2017.04.001

2

A.G. Sebastiao et al. / Applied Geochemistry xxx (2017) 1e12

1999; Callender and Rice, 2000; Laidlaw and Filippelli, 2008). From a solute perspective, several studies have evaluated the delivery of trace metals from roadways during “first flush” runoff events, where the majority of pollutants are washed into waterways during the rising limb of the hydrograph (Sansalone and Buchberger, 1997; Barrett et al., 1998; Wu et al., 1998; Legret and Pagotto, 1999; Gardener and Carey, 2004; Charters et al., 2016). In general, dissolved metals have been shown to follow a general trend in abundance (Zn > Cu z Pb > Cd) with relative differences largely attributed to traffic density (Davis et al., 2001; Gardener and Carey, 2004). For example, Sansalone and Buchberger (1997) identified mean concentrations of 62 mM for Zn and 1.5 mM for Cu in a stormwater outfall draining a highway with an average traffic density of 150,000 vehicles d1, which are far higher than the mean concentrations (1.27 mM for Zn and 0.33 mM for Cu) identified for a roadway with an average traffic density of 114,000 vehicles d1 by Gardener and Carey (2004). Though some differences may also be attributed to the sample collection techniques (Gardener and Carey, 2004), a recent study found concentrations of similar magnitude in a low density residential area (Stucker and Lyons, 2016), suggesting areas even with low traffic density can be important sources of trace metals to the environment. In addition to relative traffic density, historical land use practices have also been shown as important regulating factor for both the transience and accumulation of trace metals in particulate form (Sutherland, 1999; Brown and Peake, 2005; Sekabira et al., 2010; Zhao et al., 2010; Neira et al., 2015). For example, Brown and Peake (2005) identified a two-to six-fold increase in the metal concentrations (Cu, Pb, and Zn) in road derived solids sourced from rural and urban roadways in Dunedin, New Zealand. Other studies have used grab samples of riverbed sediment to document systematic increases in metal concentrations along a development gradient within the same river system (Sutherland, 1999; Andrews and Sutherland, 2004; Sekabira et al., 2010). The persistence of trace metals in these stream sediments has been attributed to a combination of factors, including streamwater chemistry (Zhu and Schwartz, 2011), sediment grain size (Zhao et al., 2010), and the particulate organic content (Murakami et al., 2008) of the respective system. Thus, delineating the process by which these metals are preferentially sequestered is paramount to understanding their ecotoxicity. Although these studies have greatly advanced our understanding on metal delivery and persistence in riverine systems, their limited spatial focus (i.e., downstream of storm drain outlets or within urban areas of varying density) makes it difficult to tease out the effects of non-automobile related land use practices. Furthermore, the limited temporal collection of samples either during a storm event or as a singular grab sample of riverbed sediment, provides only a one-time snapshot of delivery and accumulation, respectively. This study addresses existing knowledge gaps by examining the temporal and spatial variance of trace metal sediment pollution directly sourced from two concrete paved river fords located in suburban Philadelphia, Pennsylvania. Undercarriages of vehicles are partially submerged during each crossing, resulting in a river system that is routinely exposed to “micro” first flushes over the course of a typical day. The fact that traffic volume for the two fords differs by an order of magnitude will provide loading-response data verifying its key role in contaminant delivery. Furthermore, the fact that the fords are closed to vehicular traffic for the winter season (December through March), will allow for the determination of trace metal loading to streambed sediments over the course of the 2015 driving season (April to December). Finally, the preferential adhering of trace metals to organic carbon will be explored in effort to elucidate on the potential for bioaccumulation.

2. Study area The study area is located in the Mill Creek watershed (20.8 km2), a suburban portion of Montgomery County in Pennsylvania (Fig. 1). A review of 2011 USGS land cover data for the region confirms land use practices in the watershed range from high density developed (i.e., multi-family residential and retail commercial buildings) in its headwaters region to low density development land use practices (i.e., suburban estates with large lawns and park land) in its mid to lower reaches. The township is serviced by separate sanitary and stormwater sewer systems, the latter of which directly discharges into the Mill Creek. The study sites consist of two two-lane concrete paved fords that traverse Mill Creek in the lower portion of the watershed e the Old Gulph (40 01038.4600 N, 75170 13.0000 W) and Righters Mill (40 01047.9200 N, 75160 17.9400 W) fords (Figs. 2e4; Picture 1). The fords are typically open annually from the beginning of April to the beginning of December and closed during the during the winter months due to icy conditions and early spring runoff events. Each ford is lined with non-continuous concrete curbing designed to both decrease relative velocity of streamwater and mark the spatial extent of the roadway. Although the curbing sections vary in length they all extend approximately 0.25 m upwards from the roadway. In upstream areas of lower flow, the curbing also appeared to assist in localized sedimentation. The Old Gulph Road Ford (OGRF) is the upstream, more heavily trafficked, crossing of the two with an average of approximately 1250 vehicle crossings during a typical workday (Lower Merion Township Police Department, personal communication). By contrast, the Righters Mill Road Ford (RMRF) experiences much lower vehicular traffic (approximately 220 vehicle crossings per workday) due to its location on a less traveled, more residential roadway. The RMRF is also located in a portion of the watershed characterized by a lower mean elevation gradient (OGRF ¼ 0.04; RMRF ¼ 0.01) and has a higher relative elevation compared to streambed, thus acting as a dam; both factors result in a lower relative water velocity at the site. The watershed is predominantly underlain by the Wissahickon Schist with relatively minor outcroppings (~3%) of serpentinite. The Wissahickon Schist is comprised primarily of oligoclase-mica schist and a hornblende gneiss (both of which contain the common metamorphic mineral biotite). The serpentinite in the region is primarily composed of olivine and pyroxene (Weiss, 1949). No trace metal data readily exists for either the Wissahickon Schist or the associated serpentinite within the literature. 3. Methods 3.1. Sampling collection and processing Samples were collected from the OGRF and RMRF multiple times over the course of the 2015 driving season (Fig. 1). Samples from OGRF were collected on four occasions (April 15th, July 24th, October 28th, and December 18th) while samples from RMRF were only collected twice (April 15th and December 18th). The RMRF was sampled less frequently due the relative lack of sediment accumulation and/or localized depositional environments. The April sampling occurred shortly after the annual opening of the fords for vehicular use while the December sampling occurred shortly before their closure. The additional sample events at the moretrafficked OGRF site were spread out over approximate 90 day intervals in an effort to capture effects of seasonality on metal introduction and storage within the system. During each sampling interval, anywhere from 20 to 22 sites were sampled at the OGRF versus a total of 11 sites for the RMRF

Please cite this article in press as: Sebastiao, A.G., et al., Trace metal sediment loading in the Mill Creek: A spatial and temporal analysis of vehicular pollutants in suburban waterways, Applied Geochemistry (2017), http://dx.doi.org/10.1016/j.apgeochem.2017.04.001

A.G. Sebastiao et al. / Applied Geochemistry xxx (2017) 1e12

3

Fig. 1. View of the Mill Creek watershed located in southeastern Pennsylvania. The OGRF and RMRF sampling locations are provided.

(Figs. 2e4). Samples were predominately collected from riffles or pools with the number of samples collected at each site varied slightly as new viable sampling locations became accessible and others became unavailable due to water depth or lack of physical accumulation. While the majority of samples were collected immediately adjacent to the upstream and downstream portions of the ford, additional downstream samples were collected at 25, 50 and 100 ft intervals to determine the effects of distance on sediment loading and dispersal (Figs. 2e4). In addition, upstream samples were collected at 25 and 50 ft intervals at each of the fords during each sampling interval. Finally, a 100 ft interval upstream sample was collected from the OGRF location during the April sampling period in order to determine background concentrations. Immediately upon return to Villanova University, the sediment samples were dried at 60  C for a total of 72 h. The samples were then processed through a 63 mm soil sampling sieve to isolate the fines (silt and clay fraction). Careful sieving practices were employed to mitigate the presence of pieces of metal or plastic from vehicles and other waste, several of which were identified in the waterways when sampling for this study. The nitric acid soluble fraction of each sediment sample was then isolated using a modi€rstner (1986) and Robertson and fied version of Kersten and Fo Taylor, 2007, and the clean procedures of Gardener and Carey (2004). Approximately 0.5 g of dried sediment was digested in Teflon beakers and watch glasses using 10 ml of Fisherbrand™ trace

metal grade nitric acid. Samples were allowed to digest overnight at 60  C and then refluxed for approximately 2 h at 120  C. Up to 3 ml of J. T. Baker™ trace metal grade hydrogen peroxide was then added to the solution to dissolve any organics. Each sample was then evaporated to less than 1 ml and then diluted to 50 ml total volume using Milli-Q™ deionized water. Sample were subsequently passed through a 0.45 mm microfiber filter and syringe prior to analysis. Acid digestion blanks were also processed using the same methodology to ensure sample integrity. Finally, a total of five digestions using the same procedures were performed on the National Institute of Standards and Technology reference material RM-8704 (Buffalo River Sediment) in order to evaluate the efficiency of the extraction methodology. Multiple digestions of RM-8704 (n ¼ 3) revealed the following element specific RSDs: As (1.4%), Cd (2.2%), Cu (3.7%), Ni (1.7%), Pb (0.6%), and Zn (1.4%). The majority of the metals revealed sample recoveries above 70% (58% As, 75% Cd, 80% Cu, 65% Ni, 83% Pb, and 75% Zn), with shortfalls attributed to incomplete digestion of silicate minerals in the residual phase. 3.2. Sample analysis Concentrations of six trace metals (As, Cd, Cu, Ni, Pb, and Zn) in the acid digested samples were measured using an Agilent Technologies 7500 Series inductively coupled mass spectrometer (ICPMS) at Bryn Mawr College. External standards were used and check

Please cite this article in press as: Sebastiao, A.G., et al., Trace metal sediment loading in the Mill Creek: A spatial and temporal analysis of vehicular pollutants in suburban waterways, Applied Geochemistry (2017), http://dx.doi.org/10.1016/j.apgeochem.2017.04.001

4

A.G. Sebastiao et al. / Applied Geochemistry xxx (2017) 1e12

Fig. 2. View of Cu concentrations at the OGRF during the A) April, B) July, C) October, and D) December sampling intervals.

Please cite this article in press as: Sebastiao, A.G., et al., Trace metal sediment loading in the Mill Creek: A spatial and temporal analysis of vehicular pollutants in suburban waterways, Applied Geochemistry (2017), http://dx.doi.org/10.1016/j.apgeochem.2017.04.001

A.G. Sebastiao et al. / Applied Geochemistry xxx (2017) 1e12

Fig. 3. View of Zn concentrations at the OGRF during the A) April, B) July, C) October, and D) December sampling intervals.

Please cite this article in press as: Sebastiao, A.G., et al., Trace metal sediment loading in the Mill Creek: A spatial and temporal analysis of vehicular pollutants in suburban waterways, Applied Geochemistry (2017), http://dx.doi.org/10.1016/j.apgeochem.2017.04.001

5

6

A.G. Sebastiao et al. / Applied Geochemistry xxx (2017) 1e12

Fig. 4. View of Cu concentrations at the RMRF during the A) April and B) July sampling intervals and Zn Concentrations during the C) October and D) December sampling intervals.

Please cite this article in press as: Sebastiao, A.G., et al., Trace metal sediment loading in the Mill Creek: A spatial and temporal analysis of vehicular pollutants in suburban waterways, Applied Geochemistry (2017), http://dx.doi.org/10.1016/j.apgeochem.2017.04.001

A.G. Sebastiao et al. / Applied Geochemistry xxx (2017) 1e12

standards were run every 5 samples to account for instrument drift. Triplicate analyses did not reveal RSDs greater than 6%. Replicates performed on select samples revealed metal concentrations within 1%. Acid digestion and filter blanks analyzed did not reveal any detectable concentrations of trace metals. A separate aliquot of approximately 0.18 g of re-dried fine sediment sample was weighed out in individual foil cups and folded into pellets. They were then combusted in a TruSpec CN analyzer to determine the organic carbon percentages associated with each sample. Sample replicates did not reveal RSDs greater than 6%. 3.3. Statistical methods A variety of statistical tools were used to evaluate similarities and differences in trace metal concentrations in the dataset. In order to determine whether metal concentrations differed between the two fords with contrasting traffic density, a series of two-tailed t-tests (a ¼ 0.05) were performed to test the null hypothesis of no difference between the means. Secondly, a series of one-way analysis of variance (ANOVA) tests followed by Tukey honestly significant difference (HSD) post hoc tests (a ¼ 0.05) were performed on the average concentrations for each metal for the four

7

sampling intervals at the OGRF sampling location to determine whether significant temporal differences exist (Fig. 5). Finally, Pearson correlation coefficients (r) were calculated to determine relationships between the metals within each ford site. Calculations were performed on the logarithmic-transformed average concentrations at each of the fords, as the transformed values revealed a normal distribution. 3.4. Hydrological data analysis Metal accumulation in sediments over the course of a driving season will not only be a product of their rate of introduction but also hydrological events that can effectively mobilize and or redistribute sediments itself. Although the Mill Creek is not gauged by the USGS, hydraulic variability at each site was previously determined through the collection of monthly instantaneous discharge measurements during baseflow conditions, as well as a spring runoff event. All measurements were collected using a Teledyne RDI Stream Pro Acoustic Doppler Current Profiler. Baseflow discharge measurements for the OGRF site ranged from 0.09e0.25 m3/s with a peak storm flow value of 1.79 m3/s, compared to base flow values of 0.11e0.46 m3/s and a peak storm

Fig. 5. Mean concentrations (mg/g) of arsenic, cadmium, copper, lead, and Zinc monthly for the OGRF. Means that do not share the same letter are statistically significantly different from one another according to an ANOVA and Tukey HSD Post-Hoc Test (a¼0.05).

Please cite this article in press as: Sebastiao, A.G., et al., Trace metal sediment loading in the Mill Creek: A spatial and temporal analysis of vehicular pollutants in suburban waterways, Applied Geochemistry (2017), http://dx.doi.org/10.1016/j.apgeochem.2017.04.001

8

A.G. Sebastiao et al. / Applied Geochemistry xxx (2017) 1e12

value of 6.07 m3/s for the RMRF. A much higher proportional increase for the storm discharge value at the RMRF site was attributed to the input from Trout Run, a tributary which intersects the Mill Creek between the two fords. Discharge data over the course of the study was also obtained from the nearby USGS Cobbs Creek gauging station (01475530) to determine the number of high intensity runoff events experienced during the course of the study. The Cobbs Creek has a comparable watershed area with that above the RMRF (12.8 and 16.1 km2, respectively) and is characterized by similar relative land use practices; thus, it is expected to experience hydrological events of roughly equivalent magnitude. The dataset was evaluated for the frequency of discharge thresholds similar to those observed on the Mill Creek, which have been observed to mobilize visually noticeable amounts of sediment within the creek. 4. Results and discussion 4.1. Range of values The mean and maximum concentrations of As (6.0, 14.4 mg kg1), Cd (0.4, 4.8 mg kg1), Ni (68.0, 530 mg kg1), and Pb (153, 675 mg kg1) in the riverbed sediments are in-line with what has been reported in the literature for similar settings (Andrews and Sutherland 2004; Sekabira et al., 2010; Sutherland, 1999; Zhao et al., 2010; (Table 1). Mean and maximum Cu (623, 2703 mg kg1) and Zn concentrations (555, 1250 mg kg1) are at the high end of what has been reported for streambed sediments but in-line with mean concentrations for street dust (129 and 528 mg kg1, respectively), baseflow suspended sediments (632 and 4915 mg kg1, respectively) and stormwater suspended sediments (560 mg kg1 and 5901 mg kg1, respectively) previously obtained from an urban catchment in New Zealand (Brown and Peake, 2005). High standard deviation values for each of the metals reflect the high spatial heterogeneity in the dataset and suggest factors, such as anthropogenic inputs, and local localized geomorphological conditions (i.e., erosion versus deposition) are influencing sediment concentrations. Copper, Zn and Pb (in varying order) were the three most abundant metals in 76 of the 85 collected from the OGRF, with the remaining samples exhibiting relatively elevated concentrations of nickel. All 22 samples collected from the RMRF revealed Cu, Zn, and Pb (in varying order) as the three most abundant metals. These findings are similar to other studies focusing on road-derived runoff, which have typically identified Zn, Cu, and Pb amongst the three most abundant metals (Sutherland, 1999; Brown and Peak, 2005; Callender and Rice, 2000; Zhao et al., 2010). Spatial distribution of metal concentrations revealed preferential accumulation on the upstream side of the OGRF. For example, sample sites U3, U4, U-4A, U-5 and U-8, consistently ranked amongst the top seven locations for concentrations of Cu, Ni and Zn during the July, October and December sampling periods (The remaining metals exhibited more heterogeneous spatial variation over time). These concentration “hot spots” are likely due both to a ripple effect experienced on each side of the ford during every automobile transit as well as the downhill transport and preferential accumulation of runoff generated from the undercarriage after its passage. The latter factor may explain the relatively elevated concentrations observed in several sample locations adjacent to the northwestern (upstream) portion of the OGRF. Interestingly, a sampling location immediately downstream of stormwater outlet draining an adjacent roadway (D-1) did not reveal Cu, Pb or Zn concentrations in the top 60th percentile for any of the sampling intervals; thus, suggesting direct input from the ford plays a greater role in the cumulative loading of metals in

Table 1 (A).Descriptive statistics for select metal concentrations (in mg kg1) and %POC in samples collected from the OGRF for each sample interval compared to their respective consensus-based threshold effect concentrations (TEC) and probable effect concentrations (PEC). (B) Descriptive statistics for select metal concentrations (in mg kg1)and %POC in samples collected from the RMRF for each sample interval compared to their respective consensus-based TEC. A Ni

Cu

Zn

As

Cd

Pb

%POC

April Mean Standard Deviation Median Minimum Maximum

73.0 10.4 70.7 53.7 92.9

732 565 729 55.7 2047

407 201 390 155 935

5.88 0.80 5.93 3.78 7.50

0.57 0.19 0.55 0.27 1.01

199 156 138 54.8 675

7.36 1.94 8.42 3.49 9.80

July Mean Standard Deviation Median Minimum Maximum

104 103 77.9 7.62 530

1156 998 778 58.8 2704

798 581 626 154 1640

6.36 1.05 6.53 4.46 8.59

0.69 0.33 0.62 0.22 1.46

171 45.0 161 112 322

6.54 1.91 6.72 3.90 10.8

October Mean Standard Deviation Median Minimum Maximum

72.0 94.6 58.1 15.4 487

619 589 514 61.5 2147

650 425 579 190 1534

6.29 1.60 6.26 3.39 9.23

0.53 0.36 0.43 0.18 1.63

97.8 29.0 90.7 50.9 180

6.72 1.72 7.13 2.60 10.0

December Mean Standard Deviation Median Minimum Maximum

60.7 23.5 53.2 33.4 146

549 374 505 49.9 1404

649 385 610 220 1765

6.00 1.12 5.77 4.47 9.40

0.97 1.16 0.38 0.27 4.80

168 124 109 64.6 597

6.47 1.60 6.75 3.78 9.52

TEC valuesa PEC values

22.7 48.6

31.6 149

121 459

9.79 33.0

0.99 4.98

35.8 128

Ni

Cu

Zn

As

Cd

Pb

%POC

April Mean Standard Deviation Median Minimum Maximum

48.4 14.6 47.6 30.4 76.5

110 63.6 98.4 41.3 275

182 72.1 176 94.2 320

5.04 1.76 4.74 3.27 8.57

0.36 0.19 0.36 0.12 0.64

144 88.2 108 76.4 318

4.88 2.55 4.26 2.65 11.1

December Mean Standard Deviation Median Minimum Maximum

17.1 26.5 4.14 1.78 85.5

76.7 46.5 62.4 42.8 206

255 237 182 79.6 908

6.06 3.12 5.67 2.88 14.4

0.31 0.18 0.26 0.15 0.80

129 76.9 105 53.9 326

3.49 0.93 3.67 1.75 4.62

TEC valuesb PEC values

22.7 48.6

31.6 149

121 459

9.79 33.0

0.99 4.98

35.8 128

B

a Consesus-basedTEC and PEC values for select heavy metals in sediments for freshwater systems are taken from Macdonald (2000). b Consesus-based threshold effect concentration (TEC) and probable effect concentrations (PEC) values for select heavy metals in sediments for freshwater systems are taken from Macdonald (2000).

sediments in this portion of the stream than runoff events draining adjacent roadways. Finally, samples collected from areas further upstream (U-10, U-11, and U12) and downstream (D-7 and D-8) of the ford always exhibited concentrations of Cu and Zn in the bottom 25th percentile of all samples analyzed during each sampling interval (note: location U-12 was only sampled during April), thus confirming the role of the ford as a source. The RMRF also exhibited spatial “hot spots” of elevated concentrations with sites U1, D1, D5 and D6 consistently exhibiting metal concentrations amongst the top five for As, Cd, Cu, Ni, and Zn during each sampling interval. Average values and ranges for all constituents were greater for

Please cite this article in press as: Sebastiao, A.G., et al., Trace metal sediment loading in the Mill Creek: A spatial and temporal analysis of vehicular pollutants in suburban waterways, Applied Geochemistry (2017), http://dx.doi.org/10.1016/j.apgeochem.2017.04.001

A.G. Sebastiao et al. / Applied Geochemistry xxx (2017) 1e12

the OGRF during both of the overlapping sample intervals (April & December). A subsequent series of two-tailed t-tests (ɑ ¼ 0.05) performed on the average values for the two sites confirmed statistically significant differences for both the April and December sampling periods Cd (p ¼ 0.008, p ¼ 0.02), Cu (p ¼ 9.47E-05, p ¼ 7.22E-06), Ni (p ¼ 0.0001, p ¼ 0.0002), and Zn (p ¼ 0.0001; p ¼ 0.001). These findings add to a growing body of literature which have identified positive correlations between elevated sediment metal concentrations and traffic density. Brown and Peake (2005) identified approximately ~5 higher concentrations of Zn and ~3 higher concentrations of Cu and Pb in stormwater sediments draining streets in an urban catchment compared to its rural counterpart. In addition, a study in suburban Beijing found roughly equivalent concentrations of several heavy metals in an urban village compared to college and residential area, which can be attributed to high traffic volume in the latter (Zhao et al., 2010). Furthermore, studies documenting an increase in sediment metal concentrations along an urbanizing catchment have at least partially attributed these findings to a corresponding increase traffic density (Callender and Rice, 2000; Sutherland, 1999). 4.2. Seasonal differences in sediment concentrations A unique aspect of the study was to determine the variability of sediment metal concentration in each ford over time. At the OGRF, an increase in average mean concentrations (outside of the margin of error) with respect to the initial April values was observed for five of the six metals during at least one of the subsequent sample intervals. However, the sample interval exhibiting the highest overall mean varied for each metal. A series of one way ANOVA and Tukey HSD post-hoc tests (a ¼ 0.05) for elemental means revealed statistically significant higher concentrations for Cu (pvalue < 0.05) and Zn (p-value < 0.05) in July with similarly high levels in October and December (Fig. 5). The tests also revealed statistically higher concentrations of Pb (p-value < 0.05) in April with similarly high levels in July and December. These findings are in agreement with those from a previous study identifying significant temporal variability in the Cu, Pb and Zn content of road sweepings collected monthly from a roadway in Manchester England (Robertson and Taylor, 2007). Lee et al. (1997) also identified highly variable concentrations of Cd, Pb, and Zn in particles delivered over a two year period to a roadside stormwater settling pond along a major roadway in Sologne, France. Both studies also revealed varying relative abundances of individual metals over each sampling interval. Frequency of hydrological events, local geomorphological conditions, as well as the mode of delivery (dissolved versus particulate) likely all play a collective role in the relative accumulation or dissipation of metals in the streambed sediments. A review of the hydrograph for the nearby Cobbs Creek USGS gauging station for the sample period identified 20 storm events capable of producing a discharge value of 1.87 m3 s1 at the OGRF (our previously measured storm discharge value). A total of eight of these events occurred between the April and July and the July and August sampling intervals while four events occurred between the October and December sampling intervals. No statistical correlations were observed between the number of storm events between the sampling intervals and metal concentrations. At the OGRF, localized deposition between the April and July sampling intervals resulted in three new sampling locations (U-2A, U-3A, U-4A) and a fourth sampling location (U-1A) in October. These locations were visually dominated by fine sediment and typically exhibited elevated metal concentrations. Conversely, localized erosion prevented the collection of sediment from one sample location in July and October (U-9) and location in December (D-7). In what can be described as a

9

study design most similar to our own, Andrews and Sutherland (2004) did not identify a statistical differences in Cu, Pb, and Zn concentrations in streambed sediments collected above and below storm drains throughout a suburban watershed. The authors attributed this finding to the frequent storm events observed throughout their study period, which likely diluted sediments with those from undeveloped headwaters region. These findings in conjunction with our own confirm the dynamic nature of these environments and suggest that substantial amounts of road derived metals are continuously flushed through river systems. The limited seasonal sampling at the RMRF (April and December, only) prevented a similar systematic comparison of metal concentrations over the course of a driving season. However, a series of t-tailed t-tests (a ¼ 0.05) performed between the April and December metal concentrations did not indicate any significant differences between the means with the exceptions of Ni (p ¼ 0.003). 4.3. Enrichment ratios and elemental correlation Although the analyses revealed distinct differences in metal concentrations with respect to traffic density and variable responses to cumulative delivery over time, it is important to define the anthropogenic signature of inputs to the system. Furthermore, understanding the interrelatedness of elements found in sediments can help distinguish specific contaminant sources. One commonly used quantitative tool in road derived sediment studies to distinguish relative differences in sample concentrations with respect to background values is the Index of Geoaccumulation (Igeo; eq. (1)). The index was developed by Müeller (1969) and is defined as:

 Igeo ¼ log2

Cn 1:5* BEn

 (1)

where Cn represents the concentration of the element in question; BEn is the best estimate of its concentration in the background material (For the purpose of this study BEn was obtained from the 100 m upstream sample at the OGRF site (U-12); and the factor 1.5 is used to due to possible variations of the background due to lithogenic effects. Since it is possible that this location experiences dilute trace elemental loading from dissolved metal fluxes entering the watershed upstream of the OGRF, all calculations evaluating elemental enrichment to “background” values should be considered conservative and toxicity indices may be underestimated (An existing geochemical dataset for the Wissahickon Schist was unavailable, and while we could have potentially used soil as a background reference material, it would not have adequately accounted for contamination potentially delivered to our site from upstream sources). Müeller (1969) subsequently compared obtained index values for sediments in the Rhine River with existing biochemical water quality classifications for the same locations to develop the following ranking system: <0, practically unpolluted; 0e1, unpolluted to moderately polluted; 1e2, moderately polluted; 2e3, moderately to strongly polluted; 3e4, strongly polluted; 4e5, strong to very strongly polluted; and >5, very strongly polluted. Overall, Igeo values are in-line with what has been reported for streambed sediments impact by road runoff (Table 2) (Sutherland, 1999; Sekabira et al., 2010; Adamiec et al., 2016). For both sites, a series of two-tailed t-tests (ɑ ¼ 0.05) performed on the lumped Igeo values between sites immediately adjacent to the ford (OGRF, U-1 through U-9 versus D-1 through D-5; RMRF, U1 through U5 versus D-3 through D-6) did not indicate any significant differences between the means, suggesting no preferential accumulation on either side of the ford. However, a two tailed-test (ɑ ¼ 0.05) performed between the lumped Igeo values for sites adjacent to the

Please cite this article in press as: Sebastiao, A.G., et al., Trace metal sediment loading in the Mill Creek: A spatial and temporal analysis of vehicular pollutants in suburban waterways, Applied Geochemistry (2017), http://dx.doi.org/10.1016/j.apgeochem.2017.04.001

10

A.G. Sebastiao et al. / Applied Geochemistry xxx (2017) 1e12

Table 2 Igeo percentages in each pollution category and median, minimum, and maximum enrichment ratio values for trace elements in the OGRF streambed sediments.

April Ni Cu Zn As Cd Pb

<0a

0e1

1b (19)c 0.11 (2) 0.21 (4) 0.26 (5) 0.89 (17) 0.74 (14)

0.11 0.37 0.74 0.11 0.21

July, October, & December Ni 0.89 (58) Cu 0.10 (7) Zn 0.10 (7) As 0.37 (24) Cd 0.77 (50) Pb 0.94 (61)

(2) (7) (14) (2) (4)

(combined 0.08 (5) 0.14 (9) 0.32 (21) 0.63 (41) 0.15 (10) 0.05 (3)

>5

1e2

2e3

3e4

4e5

0.11 (2) 0.37 (7)

0.05 (1) 0.05 (1)

0.47 (9)

0.16 (3)

0.22 (14)

0.18 (12)

0.05 (1)

0.14 (9) 0.28 (18) 0.06 (4) 0.02 (1)

0.03 (2) 0.20 (13) 0.30 (19)

0.02 (1)

0.02 (1)

Mediand

Minimum

Maximum

0.56 3.2 0.87 0.07 0.62 1.16

0.96 0.23 0.52 0.2 1.64 2.44

0.17 4.61 2.01 0.4 0.27 1.18

0.78 2.82 1.41 0.11 0.85 1.22

3.78 0.74 0.59 0.74 2.18 2.55

2.34 5.02 2.93 0.73 2.52 1.01

a Igeo values adhere to the ranking system of Müeller (1969): <0, practically unpolluted; 0e1, unpolluted to moderately polluted; 1e2, moderately polluted; 2e3, moderately to strongly polluted; 3e4, strongly polluted; 4e5, strong to very strongly polluted; and >5, very strongly polluted. b Percentage of samples falling within each pollutuion category. c Number of samples falling within each pollutuion category. d Median, minimum, and maximum represent values and not percentages.

OGRF with the three downstream sites (D-6 through D-7) revealed significant differences for As (p ¼ 7.69E-07, Cd p ¼ 0.02, Cu (p ¼ 7.69E-07), and Zn (6.36E-09), thus confirming the ford as a source. A similar comparison for the RMRF did not identify any statistical differences between the means. For the OGRF site, Igeo values were calculated for April and the remaining periods (July, October, and December) collectively in an effort to track elemental accumulation over time. During the April sample period, a minimum of 10% of the samples exhibited values indicative of “unpolluted to moderately polluted” for five elements (As, Cd, Cu, Pb, and Zn) and a minimum of 5% of the samples exceeded the “moderately to strongly polluted” threshold Cu and Zn. Given the ford was open to traffic only one week before collection, the existence of these elevated Igeo values confirms the perennial persistence of automobile related contaminants in the adjacent sediments. A pronounced shift towards the percentage of samples exhibiting higher average Igeo values was observed for all metals between the April and remaining lumped intervals, with the exception of Pb, further confirms the relative accumulation of contaminants over the driving season. In addition, a minimum of 2% of the samples exhibited “moderately to strongly polluted” Igeo values or above for four metals (Cd, Cu, Ni, and Zn) and a minimum of 20% of the samples exhibited “strong to very strongly polluted” Igeo values or above for Cu. A Pearson correlation coefficient matrix was subsequently calculated for each of the two fords to determine the interrelatedness of the metals (Table 3a and b). For the OGRF, a particularly strong correlation was observed between Cu and Zn (r ¼ 0.91, p ¼ <0.0001). The robust association of these two metals, coupled with their highly elevated concentrations in the range of street dust, may suggest a particulate source and/or similar contaminant pathway. This assumption is further supported by a previous analysis of road runoff which showed preferential delivery of Cu and Zn in particulate form (Camponelli et al., 2010). Prior studies analyzing composite brake wear debris have identified Cu concentrations as high as 10.8% (g/g) (Hur et al., 2003) coupled with much lower concentrations of Zn (8300e12,000 mg/kg), Cd (1.20e2.34 mg/kg), and Pb (120e2897 mg/kg) (Hjortenkrans et al., 2007; Adamiec et al., 2016). Similar studies focusing on tire debris have identified elevated concentrations of Zn (3400e12,000 mg/kg) coupled with lower concentrations of Cu (5e7.4 mg/kg), Cd (0.86e1.7 mg/kg) and Pb (9.4e17 mg/kg) (Davis et al., 2001; Hjortenkrans et al., 2007). A

Table 3 A). Pearson correlation coefficent matrix for all trace metal concentrations at the OGRF. (B). Pearson correlation coefficent matrix for all trace metal concentrations at the RMRF. A

As Cd Cu Ni Pb Zn POC

As

Cd

Cu

Ni

Pb

Zn

0.61a 0.59 0.48 NS 0.69 0.25c

0.59 0.43 NS 0.66 0.35

0.36 NS 0.91 0.67

0.34b 0.35 NS

NS 0.23

0.57

As

Cd

Cu

Ni

Pb

Zn

0.76d 0.70 NS 0.66 NS 0.53

0.78 0.45f 0.84 0.60 0.70

0.59e 0.53 0.76 0.82

NS NS 0.53

0.60 0.48

0.57

POC

B

As Cd Cu Ni Pb Zn POC

POC

NS ¼ pearson coffecients were not significant at p ¼ 0.05. a Samples where r < 0.35 or r > 0.35 are statistically significant at p ¼ 0.001. b Samples where 0.35 < r < 0.25 or 0.25 < r < 0.35 are statistically significant at p ¼ 0.01. c Samples where 0.25 < r < 0.23 or 0.23 < r < 0.25 are statistically significant at p ¼ 0.05. d Samples where r > 0.66 are statistically significant at p ¼ 0.001. e Samples where 0.53 < r < 0.66 are statistically significant at p ¼ 0.01. f Samples where 0.45 < r < 0.53 are statistically significant at p ¼ 0.05.

subsequent comparison of median Zn:Cu (0.99), Pb:Cu (0.26), and Cd:Cu (0.001) ratios from this study with those previously determined for brake lining and tire treads identified a strong agreement with bulk brake lining dust (Zn:Cu ¼ 0.59, Pb:Cu ¼ 0.21, and Cd:Cu ¼ 0.0002) (Adamiec et al., 2016). Furthermore, a simple two component mixing model using the metal:Cu ratios for break lining dust of Adamiec et al. (2016) and bulk tire scrubbings of Davis et al. (2001) showed our median values contained greater than 98e99% brake lining dust. These findings are also supported by visual observations, which confirmed the thorough soaking of tires and associated components with each automobile passage through the ford. Overall variability in the Cu ratios for the remainder of the

Please cite this article in press as: Sebastiao, A.G., et al., Trace metal sediment loading in the Mill Creek: A spatial and temporal analysis of vehicular pollutants in suburban waterways, Applied Geochemistry (2017), http://dx.doi.org/10.1016/j.apgeochem.2017.04.001

A.G. Sebastiao et al. / Applied Geochemistry xxx (2017) 1e12

OGRF samples is likely a reflection of a several factors, including relative differences in metal concentrations of the brake pads, inputs from other sources and sediment mixing (see Table 4). Significant correlations identified among the remaining metals at the OGRF, with the exception of lead, may also suggest relative inputs from oils/lubricants, gasoline, and tires. Previous analyses of used motor oil has shown it can be a major source of Zn (1.25  105 mg/l) and lesser quantities Pb (1100 mg/l) Cu (2100 mg/l) and Cd (100 mg/l) Davis et al. (2001). Much lower concentrations of As (13.2e32.5 mg/l), Cd (2.9e8.1 mg/l), Cu (89.4e486mg/l), and Zn (50.5e175mg/l) have been identified in gasoline and diesel fuel (Becker et al., 2012; Nomngongo and Ngila, 2014). In addition, an equal number of statistically significant correlations (p < 0.05) were identified for the RMRF. However, observed increases in average Zn:Cu (2.02), Pb:Cu (1.56), and Cd:Cu (0.004) suggests higher relative contributions from these additional sources.

4.4. Potential impacts to aquatic ecosystems The dataset was evaluated using the potential ecological risk index (RI) of Hakanson (1980) to determine the possible impacts to the Mill Creek ecosystem. The RI is a diagnostic tool that takes into account both the relative abundance of a metal (assuming a response between toxicity and rarity) as well as its affinity for solid substrates. The RI is defined as follows:

RI ¼

n X

Er i ; Er i ¼ Tr i *cif ; cif ¼

i¼1

C01 Cn

where C0-1 and Cn are the concentrations of the metal in question in each sample and its corresponding background value, respectively; Tri is a toxic response factor which takes into account the toxicity of each heavy metal (Cd ¼ 30, As ¼ 10, Cu ¼ 5, Cr ¼ 2, Zn ¼ 1) (Hakanson (1980); and Eri is the ecological risk potential of each metal. Hakanson (1980) proposed the following categorization of RI values: 1) RI < 150, low ecological risk, 2) 150  RI < 300, moderate ecological risk; 3) 300  RI < 600, considerable ecological risk, 4) RI > 600, very high ecological risk. A total of 22 samples from the OGRF revealed RI values in the “moderate” category and six samples in the “considerable” category. Five of the six samples in the latter category were associated with the July sampling interval while the remaining sample was

Table 4 Metal:Cu ratios in brake linings, tires, and riverbed sediments (this study).

Brake linings Hjortenkrans et al. (2007) Front linings Rear linings Adamiec et al. (2016) Break lining dust (bulk) Tires Hjortenkrans et al. (2007) Retread tire tread rubber Nontreated tire tread rubber Davis et al. (2001) Tire scrubbings (bulk)

Zn:Cu

Pb:Cu

Cd:Cu

0.21 0.28

0.0009 0.02

9.23E-06 3.08E-05

0.59

0.21

1.67E-04

1622 1093

1.28 1.09

0.12 0.20

680

3.40

0.20

This study - OGRF (n ¼ 85) Median Range

0.99 0.43e4.84

0.26 0.05e9.12

0.001 0.0002e0.01

This study - RMRF (n ¼ 22) Median Range

2.02 1.16e4.40

1.56 0.28e2.70

0.004 0.001e0.005

11

identified in December. The fact that six of the samples in the “moderate” category were associated with the April sampling interval further confirms the pervasiveness of contaminants within this system. The majority of the samples in these elevated RI categories revealed the following sequence: Cu > Cd > Zn > Ni > Pb. All of the samples collected from the RMRF revealed RI values in the “low” category. The presence of several samples with elevated RI categories at the OGRF suggests automobile sourced metals are being retained at concentrations potentially harmful to aquatic ecosystems. This potential impacts to aquatic ecosystems is further supported by median monthly elemental values for Cu, Ni, Pb (with the exception of October), and Zn at the OGRF in excess of both their consensus based threshold effect concentrations (TEC) and probable effect concentrations (PEC) for benthic macroinvertebrate communities previously determined by Macdonald, (2000) (Table 1). In particular, median monthly Cu concentrations were approximately 3 to 5 greater than the PEC with individual samples up to 18 greater than the PEC. Likewise, median monthly Ni, Zn, and Pb were approximately 1 to 2 their respective PEC value with individual samples ranging from 10, 3 and 5 their PEC, respectively. However, monthly median values for As and Cd were below their respective TEC values. While a sequential extraction approach would be necessary to determine the preferential phase of the metals (i.e., soluble, exchangeable, carbonate, labile organic, reducible, and residual phases) in the system, a number of positive, statistically significant correlations between metal concentrations and %POC of the sediments at the OGRF (As, r ¼ 0.25, p ¼ 0.03; Cd, r ¼ 0.35, p ¼ 0.002; Cu, r ¼ 0.67, <0.001; Zn, r ¼ 0.57, p ¼ < 0.001) and RMRF sites (As, r ¼ 0.53, p ¼ 0.01; Cd, r ¼ 0.70, p ¼ 0.0004; Cu, r ¼ 0.82, p ¼ < 0.001; Pb, r ¼ 0.53, p ¼ 0.03; Zn, r ¼ 0.57, p ¼ 0.007) suggests the % POC pathway may play an important role in metal retention. An observed strengthening for the correlations at the RMRF suggests some site specific factors (i.e., lower traffic density, relatively lower water velocity during baseflow, etc.) may be playing an important role in the adherence of metals to POC. Alternatively, the relatively less robust correlations at the OGRF could be due to preferential deposition of metals in particulate form (i.e., brake pad dust, tire fragments, etc.). Although metal concentrations in this study decreased with movement downstream, there are still important implications regarding toxicology and the aquatic life within the stream. Bivalves and other trace-metal sensitive aquatic organisms could be greatly impacted by these vehicular inputs via bioaccumulation, which could play a role in the observed decline in both the number and richness of macroinvertebrates in the Mill Creek compared to other tributaries in the greater Schuylkill River watershed (Jackson, 2009). Furthermore, these metals have the potential to move up through the food web when such aquatic life is consumed and eventually reach human trophic levels (Abdallah, 2013). 5. Conclusions Trace metal sediment chemistry data has been presented for two river fords with varying traffic densities collected over multiple sampling periods during the course of a driving season. Concentrations of most metals were in the upper range of literature values for streams impacted by road derived sediments. Statistically significant differences in mean concentrations between the two fords for the majority of metals analyzed confirms the role of traffic density on metal accumulation in sediments. The OGRF sampling site shows the strong influence of brake lining dust while more elevated ratios for the RMRF suggest the relative importance of other sources such as tires, motor oil and gasoline. A pronounced

Please cite this article in press as: Sebastiao, A.G., et al., Trace metal sediment loading in the Mill Creek: A spatial and temporal analysis of vehicular pollutants in suburban waterways, Applied Geochemistry (2017), http://dx.doi.org/10.1016/j.apgeochem.2017.04.001

12

A.G. Sebastiao et al. / Applied Geochemistry xxx (2017) 1e12

shift towards higher enrichment ratios was observed between the April sampling period and the lumped values for the three remaining sampling intervals at the OGRF site documenting the accumulation of metals over the course of a driving season. Observed spatial and temporal heterogeneity in metal concentrations over the course of a driving season at the OGRF suggests other factors, such as local hydrological and geomorphological conditions, can also influence accumulation and dispersal of metals in river sediments and that contamination is readily transported to downstream portions of the watershed. Furthermore, metal concentrations in excess of consensus TEC and PEC values along coupled with a positive relationship with %POC content of the sediments provides a cautionary tale with regards to the bioaccumulations potential of metals introduced from vehicular related pathways. These results add to a growing body of literature which delineate the impact of vehicular traffic on aquatic ecosystems. Acknowledgements We would like to thank Kate Henderson for her assistance with sample collection. We would also like to thank Dr. Pedro Marenco of Bryn Mawr College for his assistance with the ICP-MS analysis of our sediment samples. We also express gratitude to Dr. Stephen Levas for his discussions related to an earlier version of this manuscript. References  ska, E., Wieszała, R., 2016. Heavy metals from nonAdamiec, E., Jarosz-Krzemin exhaust vehicle emissions in urban and motorway road dusts. Environ. Monit. Assess. 188, 1e11. Abdallah, M.A.M., 2013. Bioaccumulation of heavy metals in mollusca species and assessment of potential risks to human health. B. Environ. Contam. Tox. 90, 552e557. Andrews, S., Sutherland, R.A., 2004. Cu, Pb and Zn contamination in Nuuanu watershed, Oahu, Hawaii. Sci. Total Environ. 324, 173e182. Barrett, M.E., Irish, L.B., Malina, J.F., Charbeneau, R.J., 1998. Characterization of highway runoff in Austin, Texas area. J. Environ. Eng. 124, 131e137. Becker, E.M., Dessuy, M.B., Boschetti, W., Vale, M.G.R., Ferreira, S.L., Welz, B., 2012. Development of an analytical method for the determination of arsenic in gasoline samples by hydride generationegraphite furnace atomic absorption spectrometry. Spectrochim. Acta Part B At. Spectrosc. 71, 102e106. Brown, J.N., Peake, B.M., 2005. Sources of heavy metals and polycyclic aromatic hydrocarbons in urban stormwater runoff. Sci. Total Environ. 359, 145e155. Callender, E., Rice, K.C., 2000. The urban environmental gradient: anthropogenic influences on the spatial and temporal distributions of lead and zinc in sediments. Environ. Sci. Technol. 34, 232e238. Camponelli, K.M., Lev, S.M., Snodgrass, J.W., Landa, E.R., Casey, R.E., 2010. Chemical fractionation of Cu and Zn in stormwater, roadway dust and stormwater pond sediments. Environ. Pollut. 158 (6), 2143e2149. Chambers, L.G., Chin, Y.P., Filippelli, G.M., Gardner, C.B., Herndon, E.M., Long, D.T., Lyons, W.B., Macpherson, G.L., McElmurry, S.P., McLean, C.E., Moore, J., Moyer, R.P., Neumann, K., Nezat, C.A., Soderberg, K., Teutsch, N., Widom, E., 2016. Developing the scientific framework for urban geochemistry. Appl. Geochem. 67, 1e20. Charters, F., Cochrane, T., O'Sullivan, A., 2016. Untreated runoff quality from roof and runoff surfaces in a low intensity rainfall climate. Sci. Total Environ. 550, 265e272. Davis, A.P., Shokouhian, M., Ni, S., 2001. Loading estimates of lead, copper, cadmium, and zinc in urban runoff from specific sources. Chemosphere 4, 997e1009. Gardener, C.B., Carey, A.E., 2004. Trace metal and major ion inputs into the Olentangy River from an urban storm sewer. Environ. Sci. Technol. 38, 5319e5326. Gourgouillon, D., Schrive, L., Sarrade, S., Rios, G.M., 2000. An environmentally

friendly process for the regeneration of used oils. Environ. Sci. Technol. 34, 3469e3473. Hakanson, L., 1980. An ecological risk index for aquatic pollution control. A sedimentological approach. Water Res. 14, 975e1001. €ck, B.G., Ha €ggerud, A.V., 2007. Metal emissions from brake Hjortenkrans, D.S., Bergba linings and tires: case studies of Stockholm, Sweden 1995/1998 and 2005. Environ. Sci. Technol. 41, 5224e5230. Hur, J., Yim, S., Schlautman, M.A., 2003. Copper leaching from brake wear debris in standard extraction solutions. J. Environ. Monit. 5, 837e843. Jackson, J.K., 2009. Understanding Stream Conditions: Lessons from an 11-year Study of Macroinvertebrates in Eastern Pennsylvania's Schuylkill River Watershed, with a Focus on Exceptional-value and High-quality Streams. Stroud Water Research Center, Avondale, P.A, p. 10. Jonsson, A., Lindstrom, M., Bergback, B., 2002. Phasing out cadmium and lead emissions and sediment loads in an urban area. Sci. Total Environ. 292, 91e100. €rstner, U., 1986. Chemical fractionation of heavy metals in anoxic Kersten, M., Fo estuarine and coastal sediments. Wat. Sci. Tech. 18, 121e130. Kim, K.W., Myung, J.H., Ahn, J.S., Chon, H.T., 1998. Heavy metal contamination in dusts and stream sediments in the Taejon area. Korea. J. Geochem. Explor. 64, 409e419. Laidlaw, M.A., Filippelli, G.M., 2008. Resuspension of urban soils as a persistent source of lead poisoning in children: a review and new directions. Appl. Geochem. 23, 2021e2039. Lee, P.K., Touray, J.C., Baillif, P., Ildefonse, J.P., 1997. Heavy metal contamination of settling particles in a retention pond along the A-71 motorway in Sologne, France. Sci. Total Environ. 201, 1e15. Legret, M., Pagotto, C., 1999. Evaluation of pollutant loadings in the runoff waters from a major rural highway. Sci. Total Environ. 235, 143e150. Lyons, W.B., Harmon, R.S., 2012. Why urban geochemistry? Elements 8, 417e422. Macdonald, D.D., 2000. Development and evaluation of consensus-based sediment quality guidelines for freshwater ecosystems. Arch. Environ. Con. Tox. 39, 20e31. Mielke, H.W., Gonzales, C.R., Smith, M.K., Mielke, P.W., 1999. The urban environment and children's health: soils as an integrator of lead, zinc, and cadmium in New Orleans, Louisiana, USA. Environ. Res. 81, 117e129. Müeller, G., 1969. Index of geo-accumulation in sediments of the Rhine river. Geo. J. 2, 108e118. Murakami, M., Nakajima, F., Furumai, H., 2008. The sorption of heavy metal species by sediments in soakaways receiving urban road runoff. Chemosphere 70, 2099e2109. Neira, P., Cobelo-Garcia, A., Besada, V., Santos-Echeandia, J., Bellas, J., 2015. Evidence of increased anthropogenic emissions of platinum: time-series analysis of mussels (1991e2011) of an urban beach. Sci. Total Environ. 51, 366e370. Nomngongo, P.N., Ngila, J.C., 2014. Determination of trace Cd, Cu, Fe, Pb and Zn in diesel and gasoline by inductively coupled plasma mass spectrometry after sample clean up with hollow fiber solid phase microextraction system. Spectrochim. Acta Part B At. Spectrosc. 98, 54e59. Paul, M.J., Meyer, J.L., 2008. Streams in the urban landscape. In: Marzluff, J., Schulenberger, E., Endlicher, W., Alberti, M., Bradley, G., Ryan, C., ZumBrunnen, C., Simon, U. (Eds.), Urban Ecology. Springer, US, pp. 207e231. Robertson, D.J., Taylor, K.G., 2007. Temporal variability of metal contamination in urban road-deposited sediment in Manchester, UK: implications for urban pollution monitoring. Water Air Soil Pollut. 186, 209e220. Sansalone, J.J., Buchberger, S.G., 1997. Partitioning and first flush of metals in urban roadway stormwater. J. Environ. Eng. 123, 134e143. Sekabira, K., Origa, H.O., Basamba, T.A., Mutumba, G., Kakudidi, E., 2010. Assessment of heavy metal pollution in the urban stream sediments and its tributaries. Int. J. Environ. Sci. Tech. 7, 435e446. Stucker, J.D., Lyons, W.B., 2016. Dissolved trace metals in low-order, urban stream water, Columbus, Ohio. Appl. Geochem. http://dx.doi.org/10.1016/j.apgeochem.2016.12.003 (in press). Sutherland, R.A., 1999. Bed sediment-associated trace metals in an urban stream, Oahu, Hawaii. Environ. Geol. 39, 611e627. Weiss, J., 1949. Wissahickon schist at Philadelphia, Pennsylvania. Geol. Soc. Amer. Bull. 60, 1689e1726. Wu, J.S., Allan, C.J., Saunders, W.L., Evett, J.B., 1998. Characterization and pollutant loading estimation for highway runoff. J. Environ. Eng. 124, 584e592. Zhao, H., Li, Z., Wang, X., Tian, D., 2010. Grain size distribution of road-deposited sediment and its contribution to heavy metal pollution in urban runoff in Beijing. China. J. Hazard. Mater. 183, 203e210. Zhu, C., Schwartz, F.W., 2011. Hydrogeochemical processes and controls on water quality and water management. Elements 7, 169e174.

Please cite this article in press as: Sebastiao, A.G., et al., Trace metal sediment loading in the Mill Creek: A spatial and temporal analysis of vehicular pollutants in suburban waterways, Applied Geochemistry (2017), http://dx.doi.org/10.1016/j.apgeochem.2017.04.001