Science of the Total Environment 902 (2023) 166071

Contents lists available at ScienceDirect

Science of the Total Environment

journal homepage: www.elsevier.com/locate/scitotenv

Measuring sediment loads and particle size distribution in road runoff:

Implications for sediment removal by stormwater control measures
Ryan J. Winston a, b, *, Jon D. Witter c, R. Andrew Tirpak a
a
Department of Food, Agricultural, and Biological Engineering, The Ohio State University, 590 Woody Hayes Dr., Columbus, OH 43210, USA
b
Department of Civil, Environmental, and Geodetic Engineering, The Ohio State University, 2070 Neil Ave., Columbus, OH 43210, USA
c
Agronomy and Soils, Agricultural Technical Institute, The Ohio State University, 1328 Dover Road, Wooster, OH 44691, USA

H I G H L I G H T S G R A P H I C A L A B S T R A C T

• Particle size distribution (PSD) and TSS

characterized in Ohio road runoff
• Factors beyond AADT & catchment area
impacted TSS (climate, wind deposition,
etc.)
• PSD varied across sites; interstate PSDs
were finer than other functional classes.
• Coarser PSDs observed near low density
residential areas and during summer
storms.
• SCM selection can be guided by road
and adjacent land use characteristics.

A R T I C L E I N F O A B S T R A C T

Editor: Ashantha Goonetilleke Road runoff contributes an array of pollutants which degrade the quality of receiving waters. Sediment conveyed
in runoff results in loss of habitat and loss of reservoir capacity, among other undesirable impacts. To select and
Keywords: design stormwater control measures (SCMs), the sediment particle size distribution (PSD) is needed to quantify
Urban runoff the required hydraulic retention time for particle settling and to understand what other treatment processes (e.g.,
Stormwater control measures
filtration) are needed to meet sediment removal targets. A two-year field monitoring study was undertaken
Best management practices
across the state of Ohio, USA, to evaluate the PSD of sediment in runoff at twelve roads. The highest TSS con­
Total suspended solids
Transportation centrations were observed on interstate highways (highest annual average daily traffic [AADT]) and minor ar­
Stormwater terials (low AADT), suggesting factors beyond AADT, such as antecedent dry period, rainfall intensity, and
windborne dust and particulates, contribute to the varied sediment characteristics in runoff. The median TSS
load across all samples collected was 2.7 kg/ha per storm event, while annual TSS loads for the monitoring sites
varied from 98 kg/(ha⋅yr) to 519 kg/(ha⋅yr), with a mean value of 271 kg/(ha⋅yr). Particle size distributions
varied across the monitoring sites, with mean and median d50 of 48.6 μm and 52.5 μm, respectively. Interstate
highways (highest AADT) had significantly finer PSDs than other functional classes, while roads in low density
residential areas had coarser PSDs than other land uses. Observed differences in PSD across road characteristics
may guide SCM selection; dry detention basins and wet ponds/wetlands were predicted to provide effective

* Corresponding author at: Department of Food, Agricultural, and Biological Engineering, The Ohio State University, 590 Woody Hayes Dr., Columbus, OH 43210,
USA.
E-mail address: Winston.201@osu.edu (R.J. Winston).

https://doi.org/10.1016/j.scitotenv.2023.166071
Received 19 April 2023; Received in revised form 12 July 2023; Accepted 3 August 2023
Available online 8 August 2023
0048-9697/© 2023 Elsevier B.V. All rights reserved.
R.J. Winston et al.
removal across a variety of PSDs, while TSS reductions provided by hydrodynamic separators and high-flow
media filters (which effectively remove larger particles) may be maximized in areas with coarser PSDs (e.g.,
roads surrounded by low density residential areas studied herein).
1. Introduction
Road runoff is a contributor to the total wet-weather pollutant load
in a watershed. Highways are a source of sediment, heavy metals, nu­
trients, hydrocarbons, and bacteria due to anthropogenic and atmo­
spheric deposition processes (Kayhanian et al., 2007; Davis and Birch,
2011). Anthropogenic sources of pollutants include pavement wear, tire
and brake pad wear, and vehicular fluids (Kobriger and Geinopolis
1984; Legret and Pagotto, 1999; LeFevre et al., 2015). Successful at­
tempts have been made to tie these pollutants to factors such as annual
average daily traffic (AADT), roadway classification, pavement wearing
course, antecedent dry period (ADP), contributing drainage area, and
rainfall depth and intensity (Kim et al., 2005; Opher et al., 2009; Win­
ston and Hunt, 2017; Bakr et al., 2020). The size of particles entrained in
runoff varies substantially, with diameters from several nanometers to
several millimeters (Jayarathne et al., 2017; Zhao et al., 2022). Thus, the
particle size distribution (PSD) is an important parameter to consider in
determining pollutant removal rates of stormwater control measures
(SCMs) for sediment and sediment-bound pollutants, especially when
relying on settling-based treatment methods. In some cases, particle
shape has been shown to affect particle settling (Connolly et al., 2020),
while in Rommel et al. (2020) particle shape was not a factor which
improved model prediction of settled particulate mass.
Particulate matter is classified by its diameter into four principal
categories: dissolved solids (<2 μm), fine solids (2–75 μm), coarse solids
(75 μm – 5 mm), and gross solids (>5 mm) (WERF 2008). Others
separate coarse from fine solids at 63 μm (e.g., Semadeni-Davis 2013;
Baum et al., 2021), as this is one of the accepted thresholds between silt
and clay particle sizes. Dissolved solids consist of fine clays, colloids, and
bacteria, and are generally not able to be treated by settling-based SCMs
(Clary et al., 2020). Coarse clays, silts, and fine organic matter make up
fine solids, which require extended hydraulic retention times to settle
out of suspension (Andral et al., 1999). Coarse solids are made up of
sand, fine gravel, and organic matter, and SCMs with shorter hydraulic
retention times may be utilized to sequester these particles. Finally,
gross solids include gravels, trash, and large organic detritus (Treilles
et al., 2021; Winston et al., 2023). These largest diameter particles can
be removed from stormwater using catch basin inserts, screens, trash
racks, deep sumps, or structural SCMs.
Total suspended solids (TSS) analysis is a commonly used gravi­
metric index to quantify solids in runoff. The method typically used for
TSS analysis recommends separating out gross solids using the No. 4
sieve (mesh opening of 4.76 mm) and dissolved solids using a 2 μm filter
(WERF 2008). While TSS is a general indicator of sediment in runoff, it
provides no indication of the fraction of sediment that might be settle­
able within an SCM used to treat road runoff, such as a hydrodynamic
separator, swale, filter strip, or detention basin. However, determining
the sediment trapping efficiency of SCMs is difficult unless the PSD of
the runoff is known. For example, particles larger than 100 μm in
diameter are easily settled, while those <100 μm require lengthy (hours
to days) hydraulic retention times to be settled in an SCM (Andral et al.,
1999). Especially with regards to SCMs with limited hydraulic retention
time, a PSD dominated by clays and silts will result in a high risk of low
TSS removal. Settling velocities can substantially vary based on the
diameter, density, and geometry of a particle, with measured values
from 0.08 to 472.44 cm/min in one study (Lucas-Aiguier et al., 1998).
These vast differences in settling rates illustrate the need to understand
the PSD in road runoff. Further, water temperature is related to settling
velocity, with reductions from 20 ◦ C to 5 ◦ C diminishing modeled TSS
removal efficiencies of a sedimentation tank by 8 % (Rommel and
2
Science of the Total Environment 902 (2023) 166071
Helmreich, 2018).
Runoff PSD substantially impacts the performance of settling-based
SCMs, which are commonly employed by departments of trans­
portation (DOTs) nationwide (Ferreira and Stenstrom, 2013). PSD has
been characterized in several urban runoff studies (e.g., Characklis and
Wiesner, 1997; Sansalone et al., 1998; Li et al., 2005; Charters et al.,
2015; Winston and Hunt, 2017). Runoff PSDs exhibit a high level of
variability due to both site characteristics and climatic/runoff condi­
tions, which vary both temporally and spatially (Selbig and Bannerman,
2007). Some researchers suggest PSD varies throughout a storm and as a
function of land use in the contributing watershed (Li et al., 2006).
However, research by Charters et al. (2015) found no significant dif­
ferences between first flush and whole-event PSDs; thus, designing for
different PSDs temporally within a storm may not be justified. Distinct
PSDs by urban land use have often been suggested within a single
research study (e.g., highways appeared different from urban roads in
Charters et al., 2015). Selbig and Bannerman (2011) evaluated runoff
PSD from several land uses in Wisconsin and found considerable vari­
ability in median particle size and distribution.
In order to design effective stormwater infrastructure and mitigate
the downstream impacts of runoff, a detailed understanding of the
sediment PSD is needed to quantify the required hydraulic retention
time for particle settling and determine what other treatment processes,
such as pretreatment or filtration, might be needed to meet a target
sediment removal goal. To address this gap in knowledge, a field
monitoring study was undertaken across the state of Ohio, USA, to
determine the TSS loads and PSD of sediment conveyed by runoff.
Monitoring was undertaken at twelve geographically diverse roads with
a variety of characteristics (e.g., pavement type, traffic load, functional
class, etc.) to test the hypothesis that PSD in road runoff is affected by
site specific characteristics and that PSD influences the performance of
SCMs used for treatment of road runoff.
2. Materials and methods
2.1. Description of monitored catchments
Twelve road sites located near Columbus, Cleveland, Cincinnati,
Dayton, Lima, and Kent, Ohio, were selected for TSS and PSD monitoring
(Table 1, Fig. S1). Monitored catchments consisted solely of road con­
strained by curb and gutter draining to a single catch basin. Site selec­
tion criteria developed in the early stages of the research screened out
sites where grassed shoulders were graded toward the drainage network.
Other potential sources of sediment such as construction sites, nearby
mines or gravel pits, or openly eroding pervious areas were also
screened out such that the selected sites contributed only road runoff.
Sites were selected considering an array of potential factors which might
contribute to sediment build-up and wash-off, including rainfall pat­
terns, pavement type, AADT, functional class, adjacent land use, and
development density. Six sites were monitored in 2016 and an addi­
tional six in 2017.
Monitoring sites were identified by their corresponding road iden­
tifier: I-70, I-71, and I-90 were interstate highways, US-22, SR-43, SR-
48, SR-49, SR-59, and SR-81 were principal arterials, and SR-117, SR-
257, and US-20 were minor arterials. Interstate highways, principal
arterials, and minor arterials had AADT above 30,000, between 10,000-
25,000, and <10,000 vehicles per day, respectively. Similarly, posted
speed limits were 89–113 kph, 56–89 kph, and 56–89 kph, respectively.
The wearing course at all sites was hot mix asphalt except for US-20 and
I-90, which were paved with concrete.
 R.J. Winston et al.
Table 1
Characteristics of monitoring sites.
Site Latitude, County Catchment Pavement Functional No. of No. of No. of AADT Posted Catchment Adjacent land Development Sample size
name longitude description type class travel turn shoulder (vpd) speed limit area (ha) use density
TSS PSD
lanes lanes lanes (kph)

I-70 39.9397, Franklin Eastbound travel Asphalt Interstate 1.5 0 1 93,940 105 0.179 Commercial Urban 14 14
− 82.9387 lanes and shoulder
I-71 40.019, Franklin Northbound travel Asphalt Interstate 1.5 1 1 131,990 89 0.127 Medium Urban 14 14
− 82.995 lanes, merging lane, density
and shoulder residential
SR- 40.1557, Delaware Northbound travel Asphalt Minor 1 0 1 7060 80 0.016 Commercial Suburban 17 15
257 − 83.121 lane and shoulder arterial
US-22 39.2803, Hamilton Eastbound travel Asphalt Principal 2 0 0.5 24,730 72 0.111 High density Suburban 17 18
− 84.3185 lanes and shoulder arterial residential
SR-48 39.8227, Montgomery Southbound travel Asphalt Principal 2 0.5 0 17,054 56 0.141 Low density Urban 13 14
− 84.24 lanes and half of arterial residential
center turn lane
SR-49 39.8292, Montgomery Northbound travel Asphalt Principal 1 0 1 15,630 89 0.059 Low density Suburban 15 16
3

− 84.294 lanes and shoulder arterial residential

SR- 40.7252, Allen Eastbound lanes Asphalt Minor 2 0 0 7960 89 0.026 Agricultural Rural 11 12
117 − 84.0604 arterial
SR-81 40.7511, Allen Westbound lanes and Asphalt Principal 2 0 0.5 12,070 72 0.048 Agricultural Rural 15 14
− 84.1570 half of center turn arterial
lane
SR-43 41.1230, Portage Northbound lanes, Asphalt Principal 2 0.5 0.25 16,840 72 0.051 Low density Urban 14 14
− 81.3478 half of turn lane, and arterial residential
3 ft. shoulder
SR-59 41.1574, Portage Eastbound lane, Asphalt Principal 1 1 1 14,048 56 0.134 Commercial Suburban 14 14
− 81.3031 shoulder, and turn arterial
lane
US-20 41.6887, Lake Northbound lanes and Concrete Minor 2 0.5 0 11,950 56 0.091 Commercial Urban 16 16
− 81.2890 half of center turn arterial

Science of the Total Environment 902 (2023) 166071

lane
I-90 41.7343, Lake Westbound lane and Concrete Interstate 1 0 1 30,575 113 0.013 Agricultural Rural 15 15
− 81.1052 shoulder - bridge deck
runoff
R.J. Winston et al.
2.2. Data collection
A rain gage cluster, consisting of a tipping bucket and a manual rain
gage attached to a 1.8 m tall wooden post, was installed at each of the 12
monitoring sites in locations free from overhead obstructions. Rainfall
data were collected using 0.254 mm resolution Davis Rain Collector
tipping bucket rain gages (Davis Instruments) and stored on Hobo
Pendant data loggers (Onset Computer Corporation). Rainfall data were
stored on a 1-min interval and downloaded to a field laptop approxi­
mately once per month. Rainfall events analyzed herein had a minimum
rainfall depth of 2.54 mm and were separated by a minimum antecedent
dry period (ADP) of 6 h.
Runoff volume-proportional, composite runoff samples were
collected by ISCO 6712 automated samplers (Teledyne ISCO). ISCO 730
bubbler modules (Teledyne ISCO) were used to monitor flow depths
over weirs installed in a catch basin at each monitoring site; these were
used to determine runoff hydrographs using standard weir equations.
Flow rates were integrated with time to determine stormwater volume
and trigger sample aliquots. Once triggered, the automated sampler
pumped a 200 mL aliquot out of the catch basin through 9.5 mm
diameter plastic tubing and deposited it into 1 L plastic containers which
were later composited prior to laboratory analysis. Sample intake
strainers were utilized to remove gross solids (i.e., solids with >5 mm
diameter) and were located upstream of the weir where flow was well-
mixed. Each sampler employed a 24, 1-l bottle configuration with a
distributor arm to disperse samples into bottles. For each sampling
event, a minimum of five and a maximum of 96, 200 mL aliquots
describing >80 % of the pollutograph (U.S. EPA, 2002) were collected.
During sample collection, each set of sample bottles from a single storm
were composited in a 25 L plastic container. Thus, laboratory reported
concentrations were representative of an event mean concentration
(EMC) for TSS or an event mean PSD.
Rainfall, hydrologic, and water quality data were collected during
the 8-month period from May 2016 to December 2016 for the sites in
Franklin, Delaware, Montgomery, and Hamilton counties (Table 1).
Sites in Allen, Portage, and Lake counties were monitored from April
2017 through December 2017.
2.3. Laboratory methods
Samples were composited in a 25 L polypropylene container before
being vigorously shaken for 30 s and subsampled into laboratory sample
bottles. Composite samples were divided among a 1 L plastic bottle for
PSD analysis and a 200 mL plastic bottle for TSS analysis. Water quality
samples were placed immediately on ice and chilled to <4 ◦ C for transit
to the laboratory. Total suspended solids samples were analyzed at Pace
Analytical, Inc. (Englewood, Ohio) using American Public Health As­
sociation (APHA et al. 2017) methods. Particle size distribution samples
were shipped on ice to the biogeochemistry laboratory at North Carolina
State University (Raleigh, NC) and analyzed using a Beckman-Coulter
13–320 Laser Diffraction Particle Size Analyzer equipped with a Uni­
versal Liquid Module. A refractive index of 1.59 + 0.0001i was used for
PSD analysis based on standards for nondispersed samples. One-hundred
seventeen divisions of particle size were evaluated across the range of
0.04–2000 μm; for each division, the volumetric percentage of the total
particle volume in the sample was determined. The percentage of sand
(50–2000 μm), silt (2–50 μm), and clay (<2 μm) was thus enumerated
for each sample. All TSS and PSD samples were analyzed within a hold
time of seven days.
2.4. Data analysis
2.4.1. Rainfall and runoff
Summary statistics for each precipitation event were developed,
including rainfall depth (mm), rainfall duration (hrs), average rainfall
intensity (mm/h), peak 5-min rainfall intensity (mm/h), ADP (days),
4
Science of the Total Environment 902 (2023) 166071
and season. In rare instances, rainfall data were lost due to data logger
battery failure or because of debris clogging the rain gage funnel. In
these cases, rainfall data from a paired site located with 16 km (e.g.,
rainfall data from SR-48 would be substituted for data missing at the SR-
49 gauge) were utilized to fill gaps in data. Of the 490 observed rainfall
events across the 12 sites, this occurred for just 20 events, or 4 % of the
rainfall data set.
Hydrologic measurements obtained using the bubbler flow meters
were used to quantify runoff timing, volume, and rate. Runoff volume
was determined by integrating under the hydrograph, while peak flow
rate was determined as the instantaneous 2-min maximum flow rate
over the flow duration. Hydrologic data collected at the I-70 site were
deemed unreliable; thus, TSS loads could not be determined at this site.
For about half of observed storms at I-70, far more runoff depth was
measured than rainfall depth. It was determined that two factors
resulted in changes to the contributing catchment area for this catch
basin, namely: (1) upslope catch basins were overwhelmed during
intense periods of rainfall, and (2) clogging of upslope catch basin inlets
with gross solids; both of these factors resulted in additional water
entering the monitored catch basin.
2.4.2. Water quality
Area-normalized TSS loads (kg/ha) were calculated as the product of
TSS EMC and measured runoff volume for each rainfall event and
dividing by the catchment area:
EMCTSS x Vevent
L= xC (1)
A
where L is the area-normalized TSS load (kg/ha), EMCTSS is the TSS
event mean concentration (mg/L), Vevent is the runoff volume during a
given rain event (L), A is the drainage area (ha) to the monitored catch
basin, and C is a constant to convert units to kg/ha. Annual area-
normalized TSS loading rates (kg/(ha⋅yr)) were calculated to scale
measured event TSS loads to levels that could be expected from similar
road catchments during an average year of rainfall:
n (
∑ ) Pmean ann
EMCTSS,i x Vevent,i x ∑ P
(2)
samp,loc
i=1
Lann = xC
Aloc
where Lann is the annual area normalized TSS load (kg/(ha⋅yr)) at a
given monitoring location, n is the number of sampled storm events,
EMCTSS,i and Vevent, i are the event mean TSS concentration (mg/L) and
runoff volumes (L) observed for rain event i, Pmean ann is the mean annual
rainfall depth (mm) measured at the nearest airport rain gauge,
∑
Psamp,loc is the total rainfall depth measured at a given monitoring site
during the study (mm), Aloc is the contributing drainage area to the
monitored catch basin (ha), and C is a constant to convert units to kg/
(ha⋅yr). Pmean ann values varied from 932 to 1057 mm due to geographic
variability in rainfall across the state and were calculated from rainfall
data sets measured from 1948 to 2013.
2.4.3. Statistical analysis
The Kruskal-Wallis test was used to assess significant differences in
rainfall characteristics between sites and seasons (Kruskal and Wallis,
1952). Pairwise differences in rainfall between sites and seasons were
identified using Dunn's test with a Bonferroni adjustment (Dunn, 1961).
These statistical tests were also used to evaluate between-site, seasonal,
and pairwise differences in TSS loads. Spearman's rank correlation was
used to identify significant relationships between site characteristics,
climatic variables, as well as TSS and PSD parameters.
Kruskal-Wallis tests were performed to evaluate differences in PSD
parameters between sites and seasons. Dunn's tests with Bonferroni
adjustments were again used to identify significant pairwise differences
between sites and seasons. Bootstrapping was performed to determine
the 95 % confidence intervals around particle sizes of interest for various
R.J. Winston et al.
PSDs.
All statistical analyses were performed in the statistical software R
version 4.2.2 (R Core Team, 2023). Results were considered statistically
significant at p < 0.05 unless otherwise noted.
2.4.4. Sediment removal by SCMs
Sediment removal rates as a function of particle size were gleaned
from previous research for various SCMs. These data were used with
PSDs measured herein to calculate the potential sediment removal rates
of various SCMs implemented to treat road runoff. Inherent in this
calculation is the assumption that all particles in stormwater evaluated
in this study had similar density (i.e., to convert from volumetric to mass
percentage). Percent removal rates reported for a particle size were
considered applicable for larger particles up to the next reported particle
removal rate. For example, the removal rate reported for the d20 particle
size was assumed for all particles up to the next reported value (e.g., the
d50 particle diameter). Removal rates across a variety of SCMs were
applied to mean overall PSD observed in this study, as well as mean PSDs
for different road and climatic conditions which resulted in significantly
different PSDs herein.
3. Results and discussion
3.1. Precipitation
At each monitoring site, between 26 and 52 rainfall events were
observed over the monitoring periods. A total of 490 rainfall events were
observed; of these, 176 and 190 were sampled and analyzed for PSD and
TSS, respectively. A minimum of 12 and a maximum of 18 events were
sampled for PSD at each monitoring site. Sampled storms represented
46.2 % of the 6924 mm of rainfall observed at the 12 monitoring sites.
Median event depth for sampled storms was 16.0 mm, while that for
all observed storms (i.e., including both sampled and unsampled events)
was 10.4 mm. The maximum observed rainfall depth was 91.7 mm at
SR-257; this event was sampled for PSD and TSS. The median peak
rainfall intensity for sampled storms (22.9 mm/h) was similar to that for
all observed rainfall events (24.4 mm/h). Median rainfall duration for
sampled and all observed events was also nearly equal (i.e., 5.6 h and
5.8 h, respectively); however, median ADP for sampled events (3.6 d)
was greater than that all observed rainfall events (2.8 d) at eight of the
12 monitoring sites. These data suggest that sampled storms generally
had greater rainfall depth and longer ADP but similar rainfall intensity
and duration to the population of rainfall events.
Substantial variability in rainfall characteristics was observed across
the monitoring sites (Fig. S2); however, statistical testing showed no
significant differences in rainfall depth, peak intensity, and ADP across
the 12 monitored sites. Between-site significant differences were
observed for rainfall duration and average intensity; follow-on paired
comparisons with Dunn's test with a Bonferroni correction showed that
the I-71 site exhibited significantly lower rainfall duration than SR-43 (p
= 0.036), resulting in significantly higher average intensity at I-71 than
at SR-43 (p = 0.019). Significantly greater average intensity was also
observed at I-71 than at I-90. Rainfall duration and average intensity
were not significantly different for all other combinations of sites;
generally, these analyses suggested that rainfall characteristics were not
substantially different across the geographically varied sites monitored
in this project.
Rainfall depth and antecedent dry period did not statistically differ
across seasons. Kruskal-Wallis tests showed significant seasonality in
rainfall duration, peak intensity, and average intensity. Rainfall dura­
tion in the fall and spring were significantly greater than summer, and
duration in the fall was significantly greater than spring. Peak and
average rainfall intensities were significantly greater in summer than in
fall and spring, which is expected given the convective thunderstorms
that occur in Ohio during the summer months (Fritsch et al., 1986).
5
Science of the Total Environment 902 (2023) 166071
3.2. Water quality
3.2.1. TSS concentrations
Observed TSS EMCs at the 12 monitoring sites were relatively low,
with an overall mean and median concentrations of 35 and 26 mg/L,
respectively (Fig. 1). Mean concentrations for each site ranged from 13
to 70 mg/L. Maximum TSS concentrations ranged from 40 mg/L to 312
mg/L (mean of the maximum concentrations observed at each of the 12
sites was 106 mg/L). Measures of central tendency for TSS concentra­
tions from roads in the literature (National Stormwater Quality Data­
base, California, Texas, Oregon, and North Carolina) suggest a range of
TSS concentrations from 19 to 163 mg/L (Barrett et al., 1998; Barrett
et al., 2004; Nason et al., 2012; Winston and Hunt, 2017; Pitt et al.,
2018; Simpson et al., 2022). These data suggest that TSS concentrations
observed in this study were on the lower end of those in the literature,
but not entirely dissimilar from those in past studies.
Kruskal-Wallis tests indicated no significant seasonal changes in TSS
concentrations, suggesting that road runoff TSS concentrations are
consistent throughout the year. TSS concentrations differed significantly
between monitoring sites (p < 0.001). TSS concentrations at I-70, I-71,
SR-257, US-20, and SR-81 were significantly greater than those
measured at SR-49, while concentrations at US-20 and SR-81 were
significantly greater than SR-48. Thus, a mixture of interstate highways
(i.e., I-70 and I-71) with the highest AADT and largest catchment areas
as well as minor arterials (i.e., SR-257) with low AADT and a small
catchment area produced comparatively higher TSS concentrations.
These data suggest that other factors beyond AADT, such as ADP, rain­
fall intensity, and windborne dust and particulates, may contribute to
TSS concentrations in runoff.
Spearman's rank correlation was used to explore relationships be­
tween TSS concentrations and potential influential variables, including
rainfall characteristics and quantitative site characteristics such as
AADT, speed limit, and catchment area (Fig. 2). TSS concentrations were
not significantly correlated to site or rainfall characteristics. Previous
studies, including Winston and Hunt (2017), Kayhanian et al. (2007),
Drapper et al. (2000), and Yu et al. (1994) observed significant corre­
lations between various rainfall, runoff, and site characteristics to TSS
concentrations in road runoff. However, the causal relationships be­
tween such variables and TSS concentrations observed in these studies
were not found herein.
3.2.2. TSS loads
Total suspended solid loads were determined for each sampled
rainfall event and were subsequently normalized by watershed area
(Fig. 3). TSS loads ranged from 0.1 to 64.9 kg/ac across the 177 events
sampled for TSS. The mean TSS load across all samples collected at the
11 sites (recall the I-70 flow data were unreliable) was 6.1 kg/ha per
storm event, with mean loads for each site ranging from 1.0 kg/ha to
13.3 kg/ha. These values were relatively low compared to previous
research on TSS loads in road runoff; for example, Irish et al. (1998)
observed TSS loads of 1.1–109.4 kg/ha for runoff events from the MoPac
Expressway in Austin, TX.
Storm event loads by site were summed and scaled by the ratio of
mean annual rainfall to the total sampled rainfall event depth to
determine an annual mass loading rate (Fig. 3). Annual TSS loads varied
from 107.0 kg/(ha⋅yr) to 569.2 kg/(ha⋅yr), with a mean value of 297.5
kg/(ha⋅yr) across the 11 sites. Previous studies on roads in North Car­
olina, Texas, Taiwan, Virginia, and 10 sites across the U.S. observed TSS
loads between 403 and 2156.5 kg/(ha⋅yr) (Chui et al., 1982; Stotz, 1987;
Wu et al., 1998; Barrett et al., 1998; Järveläinen et al., 2017; Yazdi et al.,
2021). Generally, these data suggest that annual sediment loading rates
from roads in Ohio are on the low end of those observed in the literature.
This may be explained in part by the variety in road characteristics for
the site studied herein, whereas many previously published studies
focused solely on highway runoff.
R.J. Winston et al.
Fig. 1. Boxplots of TSS concentrations by site. Dashed line indicates overall mean TSS concentration across all sites (35 mg/L). TSS concentrations are plotted on
logarithmic scale.
3.2.3. Particle size distribution
A total of 176 PSD samples were collected across the twelve moni­
toring sites. Between 12 and 18 samples were collected at each site, with
SR-117 and US-22 having the smallest and largest number of samples
collected during the study, respectively. Median and mean PSD char­
acteristics were determined for all samples obtained at each site
(Table S1). Median d50 ranged from 24 μm at SR-81 to 72 μm at SR-49.
Mean d50 particle size ranged from 29 μm at SR-81 to 102 μm at SR-48.
The median d10 ranged from 3 μm to 17 μm, while the median d90
ranged from 89 μm to 200 μm. Coefficients of uniformity (Cu) and cur­
vature (Cc), calculated as the ratio of d60 to d10 and the ratio of the
square of d30 to the product of d60 and d10, respectively, ranged from 5 to
10 and 1.2–2.6, respectively.
Clays, silts, and sands can be defined as particles with diameters <2
μm, between 2 and 50 μm as silt, and those >50 μm, respectively (USDA
1975). Sand, silt, and clay fractions are reliable estimators of the unit
processes necessary for sediment removal from runoff; these percentages
were calculated for the 12 sites monitored herein (Table S2). Across the
sites, the mean PSD was 48.7 % sand, 47.4 % silt, and 3.9 % clay. PSDs
varied between sites, ranging from 27.1 to 66.3 % sand, 31.5–64.5 %
silt, and 2.2–8.4 % clay, suggesting that the PSD of Ohio road runoff is
dominated by the silt and sand fractions, similar to past studies in other
states and countries (Selbig and Feinen 2012; Charters et al., 2015;
Winston and Hunt, 2017). The median particle size at five of the road
runoff monitoring sites (i.e., I-90, US-22, SR-257, SR-48, and SR-49) was
characterized as sand, while the d50 for the remaining seven sites was
silt. Based on these PSDs, SCMs with short hydraulic retention time, such
as roadside vegetated filter strips and swales, may provide substantial
TSS removal if implemented in these areas (Lucke et al., 2014; Winston
6
Science of the Total Environment 902 (2023) 166071
et al., 2017).
Particle size distributions from the 176 sampled storm events are
presented in Fig. 4, where each of the 117 particle size divisions eval­
uated during laser diffraction analysis is represented by a separate
boxplot. The overall median PSD is approximated by connecting the
median horizontal line in each boxplot (represented by the black curve
shown in Fig. 4). Little inter-sample variability was observed for clay-
sized and coarse sand (>500 μm) particles; the interquartile range was
largest for the silt fraction. Very few particles larger than 1000 μm were
observed, representing just 0.7 % of the mean PSD; this result was
similar to findings reported by Selbig and Bannerman (2007), Charters
et al. (2015) and Selbig et al. (2016).
The median d50 for all PSD samples was 52.5 μm, equivalent to a very
fine sand particle size. This was modestly smaller than the PSD of road
runoff in New Zealand (71.6 μm; Charters et al., 2015) and Wisconsin
(70–200 μm; Selbig and Bannerman, 2007) and on the lower end of
values reported for eight road sites in North Carolina (31 to 144 μm;
Winston and Hunt, 2017). The d50 observed herein was most similar to
values reported by Brodie and Dunn (2009), Wu et al. (2015), and
Westerlund and Viklander (2006), who found d50 values <50 μm for
sediment in road runoff. The mean d10 and d90 measured herein (8.4 and
175.5 μm, respectively) were similar to those from an asphalt road in
New Zealand (23.2 and 177.2 μm, respectively; Charters et al., 2015).
Selbig and Bannerman (2007) found d10 from 2 to 5 μm and d90 from
300 to 400 μm for roads in Madison, Wisconsin.
To meet a target treatment threshold of 80 % TSS load reduction (i.e.,
the current requirement for Ohio's National Pollutant Discharge Elimi­
nation System permit; Ohio Environmental Protection Agency, 2018),
nearly all particles larger than the d20 would need to be removed from
R.J. Winston et al.
Fig. 3. Boxplots of event TSS load (kg/ha) by site. Median sampled storm event TSS load (kg/ha) and TSS annual loading rate (kg/(ha⋅yr)) are noted above
each boxplot.
runoff (assuming poor removal below this threshold). The median d20 in
this study was 22.1 μm, suggesting that SCMs would need to be effective
in treating particulate matter larger than a mid-sized silt particle to
attain permit compliance (Fig. 4). Mean Cu and CC values across all
measured PSDs were 6.0 and 1.4, respectively. Based on the coefficients
of uniformity and curvature, the mean PSD measured herein would be
considered well graded. Similar results were observed for PSDs in North
Carolina (Winston and Hunt, 2017). Understanding the target particle
size required to achieve permit compliance can guide SCM selection for
implementation on various roads (see Section 4).
7
Science of the Total Environment 902 (2023) 166071
Fig. 2. Correlogram of TSS concentrations,
particle size statistics (d10, d20, d50, d90,
coefficient of uniformity (Cu), and coeffi­
cient of curvature (Cc)) and potential pre­
dictor variables, including climatic factors
(ADP, peak and average rainfall intensity
(Peak Int and Avg Int, respectively), rain­
fall depth, and duration) and site charac­
teristics (speed limit, catchment area
(CatchArea), AADT). Blank cells in the
correlogram were not statistically signifi­
cant. Larger circles with darker colors
suggest higher correlation coefficients.
Statistical testing was undertaken to determine significant differ­
ences between sites for measured PSDs (Table S3). This testing sup­
ported the hypothesis that SR-48, SR-49, and I-90 tended to have the
coarsest PSDs compared to the other monitoring sites. SR-48 and SR-49
were in the same region, located just 4.8 km apart near Dayton, while
the I-90 site was unique because it was paved with concrete (Table 1). It
is possible that local materials used to pave the Dayton sites and dif­
ferences in concrete mix design vis-à-vis hot mix asphalt resulted in
coarser PSDs as the pavements wear and contribute sediment to runoff.
Conversely, the I-70, I-71, and SR-81 sites had the finest PSDs of the
R.J. Winston et al. Science of the Total Environment 902 (2023) 166071

Fig. 4. Aggregated particle size distributions for the 176 PSD samples across the 12 monitored road sites. Each boxplot represents the variability in particle size for
that particle size division. The median PSD is approximated by connecting the median value in each boxplot with the black curve shown.

monitored sites. Interestingly, the two sites with the highest AADT (i.e.,
I-70 and I-71) had significantly smaller particles entrained in runoff.
Higher AADT, and particularly presence of a greater number of trucks
(on interstate highways, for instance), results in greater pavement wear
(Ramaswamy and Ben-Akiva, 1990), perhaps pulverizing pieces of
asphalt and other larger particles into smaller particles during dry
periods.
Summary statistics for d10, d20, d50, d90, CU, and CC were calculated
across all measured PSDs for different seasons, wearing courses (asphalt,
concrete), functional classes (i.e., interstate, principal arterial, and
minor arterial), development densities (i.e., urban, suburban, rural), and
surrounding land uses (i.e., commercial, low density residential, high
density residential, and agricultural). Characterizing differences across
these categorical variables may aid in SCM planning and implementa­
tion. Based on bootstrapped confidence intervals for d10, d20, d50, and
d90, no significant differences in PSD were observed across different
development densities or between concrete and asphalt wearing
courses.
Conversely, surrounding land use was a significant predictive factor
in road runoff PSD (Fig. 5). Sites in low density residential areas had
significantly larger d10 than high density residential and commercial
sites. Low density residential sites had significantly larger d20, d50, and
d90 than commercial, high density residential, and agricultural land
uses. All other land use particle size comparisons were not significantly
different. Previous studies have observed substantial differences in PSD
between different land uses (Selbig and Bannerman, 2011). There is the
potential for surrounding land uses to contribute to runoff PSD through
vehicular activities and transport of local soils onto roads by vehicles
and wind (Charters et al., 2015). Kayhanian et al. (2007) found signif­
icant evidence to show that surrounding land use substantially impacts
stormwater quality, which concurs with the results herein.
Road functional class has been correlated to runoff quality, with
differences in AADT across functional classes potentially causing these
effects (Opher and Friedler, 2010). Bootstrapped 95 % confidence in­
tervals suggested that the largest particles in runoff (represented by the
d90) were significantly smaller for interstate highways than for principal
arterial roads (Fig. S4). Significant differences for the d10, d20, or d50
particle size were not observed, albeit the overlap in d50 confidence

Fig. 5. Mean particle size distributions for roads with commercial (“Com”), low density residential (“Low Den Res”), high density residential (“High Den Res”), and
agricultural (“Agric”) surrounding land uses. Bootstrapped confidence intervals (95 %) for d10, d20, d50, and d90 are tabulated.

8
R.J. Winston et al.
intervals for interstate and principal arterial roads was minor. Winston
and Hunt (2017) found no significant difference between road runoff
PSD across functional classes; similarly, Selbig (2015) found no rela­
tionship between d50 and functional class. Taken together, results sug­
gest that SCM design should not be modified across road functional
classes.
Significant differences in PSD by season were observed for d10, d20,
d50, d90, CU, and CC. Follow-up paired comparisons showed that d10, d20,
and d50 were greater in the summer than in the fall and spring (α = 0.10
for d10). The sand fraction (represented by the d90) was greater in both
summer and fall than spring (α = 0.10). Concurrently, CU was signifi­
cantly greater in fall than in spring and summer and CC was greater in
spring than in summer. Thus, generally PSD was coarser in the summer
across the 12 monitored sites than in the fall or spring (Fig. S3). As
monitoring site characteristics did not change seasonally, it is hypoth­
esized that these results are related to seasonal differences in rainfall
characteristics. Because rainfall duration was significantly shorter in
summer than in fall or spring and both peak and mean rainfall intensities
were greater in summer than in fall or spring, resulting hydrographs
during the summer were flashier and resulted in significantly higher
peak flow rates. Elevated flow rates entrain larger particles in storm­
water, similar to sediment transport processes in-stream (Shields, 1936),
resulting in coarser PSDs during the summer than in spring or fall.
Correlations were explored between PSD parameters (i.e., d10, d20,
d50, d90, CU, and CC) and potential causal variables such as rainfall
characteristics and quantitative site characteristics such as AADT, speed
limit, and catchment area. These analyses were performed considering
all PSD data across all monitoring sites. Site and rainfall characteristics
were not significantly correlated to any PSD parameter. TSS concen­
tration was the only explanatory variable that was moderately nega­
tively correlated to d10, d20, and d50 and weakly negatively correlated to
d90. In general, smaller TSS concentrations were observed for coarser
PSDs.
Fig. 6. Comparison of mean particle size distribution measured in this study and the NJDEP particle size distribution utilized for approval of proprietary devices.
Numbers adjacent to symbols are percent passing values at the corresponding particle size.
9
Science of the Total Environment 902 (2023) 166071
3.2.4. Comparison to NDJEP distribution
Particle size distribution in untreated runoff is particularly important
for the design and function of SCMs, since it determines what unit
processes are needed in the system to achieve a desired removal effi­
ciency or effluent concentration. Shorter hydraulic retention times and
processes such as settling can be relied upon to remove particles from
suspension for coarser PSDs. Processes such as filtration may be required
as PSDs trend toward silt and clay particles. Standard PSDs, such as the
OK110 and New Jersey Department of Environmental Protection
(NJDEP) distributions (Guo et al., 2008), are utilized in laboratory
testing to certify the TSS removal capability of various SCMs, including
proprietary stormwater treatment systems. The NJDEP distribution is of
particular interest because other state jurisdictions often defer to NJDEP
for benchmarking TSS removal rates of proprietary devices.
The median PSD measured in runoff from the 12 roads herein was
compared against the NJDEP distribution (Fig. 6). The two PSDs align
closely, with at most a 16 % difference in percent passing across the
particle size distributions. The PSD from this study tends to be finer than
the NJDEP distribution for particle diameters larger than the d50.
However, most of these larger particles are well sequestered by SCMs
with very short hydraulic retention time, such as hydrodynamic sepa­
rators (Ferreira and Stenstrom, 2013). Median particle diameters for the
two distributions were somewhat similar: 75 μm for the NJDEP distri­
bution and 52.5 μm for the distribution measured herein. Below the
median particle size, the median PSD measured herein was coarser than
that for NJDEP. This suggests that if designing for 80 % TSS removal,
SCMs designed to treat the NJDEP PSD would conservatively provide 80
% TSS removal for road runoff in Ohio. Designing to trap the d20 particle
size conservatively meets 80 % TSS removal, since some particles
smaller than the d20 will be trapped in the SCM. The d20 for the NJDEP
distribution was 8 μm, while that for the distribution herein was 18 μm.
This suggests that technologies certified for 80 % TSS removal for the
NJDEP distribution would provide slightly >80 % TSS removal for the
R.J. Winston et al. Science of the Total Environment 902 (2023) 166071

mean PSD measured herein.
3.3. Implications for SCM selection
Using published literature, the sediment removal performance of
several SCMs was explored as a function of particle size to demonstrate
their potential water quality benefits if implemented on roads similar to
those in this study (Table 2). The focus of this was not to analyze dif­
ferences in treatment efficacy between the SCM types (including
infiltration-based versus non-infiltrating SCMs), but rather to assess the
sediment treatment function of different SCMs for factors resulting in
significantly different road runoff PSD herein (e.g., AADT, season, and
surrounding land use).
As expected, potential sediment removal performance varied greatly
between SCMs, primarily due to differences in hydraulic retention time
and fundamental sediment removal processes for each technology. Only
dry detention and wet ponds/wetlands provided sufficient treatment of
the mean PSD observed herein to achieve the target TSS reduction (80
%) required in Ohio (OEPA 2018). The removal performance provided
by these SCMs across a range of particle diameters led to minimal dif­
ferences in TSS reduction between road use intensity (i.e., AADT), sea­
son, and land use characteristics. These results suggests that dry
detention basins and wet ponds/wetlands could be considered for all
roads monitored in this study and, provided sufficient conditions for
construction (e.g., topography, available land), implemented to achieve
80 % TSS removal.
While filter strips and swales have the potential to substantially
reduce TSS concentrations through sedimentation and filtration for the
mean PSD measured herein (by 59.9 % and 75.8 %, respectively), this
analysis suggests these SCMs alone do not provide sufficient removal to
achieve TSS reduction targets. Like dry detention and wet ponds/wet­
lands, little performance differences were observed across PSDs for high-
and low-AADT roadways, seasons, or land uses, likely due to the limited
removal of smaller particles provided by these technologies. Seasonal
differences in performance associated with vegetation growth may
bolster these removal rates. Due to the consistent performance across
statistically different PSDs observed in this study, filter strips and swales
may be appropriate SCMs for all the monitored roads; however, these
systems must be paired with other SCMs or management practices (e.g.,
street sweeping) which mitigate smaller particles to achieve 80 % TSS
removal.
Hydrodynamic separators (Lee et al., 2014) and high-flow media
filters investigated in Smolek et al. (2018) provided the lowest TSS
removal rates (12.8 % and 30.2 %, respectively) for the mean PSD
observed in this study, suggesting these technologies may best serve as
pretreatment devices for other SCMs to achieve water quality targets.
Unlike the other SCMs in question, road attributes appear to impact the
TSS removal performance of these devices; potential TSS removal rates
for hydrodynamic separators and high-flow media filters were 55.3 %
and 22.7 % higher, respectively, when treating the mean PSD for roads
sounded by low density residential areas compared to all other land
uses. This is best explained by the significantly coarser PSDs found on
roads in low density residential areas compared to other land uses and
the efficacy of these technologies to remove larger particles from runoff
(and, conversely, their inability to remove smaller particles). Increased
removal rates for summer PSDs, which were significantly coarser than
other seasons, also support this hypothesis. Thus, while hydrodynamic
separators and high-flow media filters will require additional SCMs to
achieve 80 % TSS removal, their contributions to sediment removal may
be maximized when implemented on roads in low density residential
land use or other areas with coarser PSDs.
4. Conclusions
Particle size distribution in runoff impacts the design of treatment
SCMs, effectively determining what treatment processes are needed to
control TSS. To understand the variability in PSD from roads and factors
which affect it, PSD and TSS were monitored 12 geographically diverse
road sites across Ohio over two years. Results indicate that interstate
highways with the highest AADT and largest catchment areas as well as

Table 2
Potential TSS removal performance (%) of various SCMs receiving Ohio road runoff. Removal rates by particle size were applied to the mean PSD measured in field
monitoring in this study and to significantly different PSDs of various site attributes (i.e., AADT, season, land use).
SCM Removal performance by Mean AADTa Season Land Use Source
particle size (%) PSD
High Low Spring Summer Fall Low density All other
residential land uses

Hydrodynamic <70 μm: 0 % 12.8 10.0 13.3 8.1 15.0 12.7 17.7 11.4 Lee et al. (2014)
separator 70–150 μm: 20.0 %
150–250 μm: 41.0 %
250–425 μm: 87.0 %
>425 μm: 95.0 %
Dry detention basin <8 μm: 66.0 % 89.5 88.1 89.7 88.8 90.2 88.6 91.3 88.9 Ferreira and Stenstrom (2013)
8–20 μm: 87.5 %
20–100 μm: 89.5 %
>100 μm: 100 %
Wet pond/wetland <2 μm: 84 % 97.1 96.6 97.2 96.9 97.3 96.9 97.7 96.9 Toronto and Region
2–63 μm: 96.0 % Conservation Authority
>63 μm: 100 % (2002)
Filter strip <10.6 μm: 0 % 59.9 59.5 60.0 61.1 60.4 58.0 61.2 59.5 Knight et al. (2013)
10.6–88.2 μm: 72.0 %
88.2–468.4 μm: 60.0 %
>468.4 μm: 80 %
Swale <9.7 μm: 0 % 75.8 74.9 75.9 76.6 76.6 73.6 77.7 75.1 Knight et al. (2013)
9.7–80.2 μm: 88.0 %
80.2–420.4 μm: 79.0 %
>420.4 μm: 80.0 %
High flow media <27.4 μm: 0 % 30.2 26.6 30.7 27.9 32.3 27.9 35.1 28.6 Smolek et al. (2018)
filter 27.4–78.3 μm: 38.0 %
73.8–175.3 μm: 40.0 %
175.3–241.9 μm: 61.0 %
241.9–872 μm: 66.0 %
>872 μm: 74.0 %
a
Note: High AADT included I-71 and I-70 sites (min. AADT 93940 vpd); remaining sites categorized as Low AADT (max. AADT 30575 vpd).

10
R.J. Winston et al.
minor arterials with low AADT and small catchment areas produced the
highest TSS concentrations, suggesting that other factors beyond AADT,
such as ADP, rainfall intensity, and windborne dust and particulates,
may contribute to TSS concentrations in runoff. Road runoff contributed
98 kg/(ha⋅yr) to 519 kg/(ha⋅yr) of TSS, with a mean of 271 kg/(ha⋅yr).
The mean and median d50 were 48.6 μm and 52.5 μm, respectively, with
very little variability in the percentage of clay and coarse sand-sized
particles in observed samples. TSS concentration was moderately
negatively correlated to d10, d20, and d50 and weakly negatively corre­
lated to d90, suggesting that as TSS concentration decreased, PSD was
coarser.
Road functional class, surrounding land use, and season all signifi­
cantly affected measured PSD. However, these differences were small in
magnitude and thus modifications to BMP design to treat these various
PSDs are not recommended. While dry detention basins and wet ponds/
wetlands provided sufficient TSS removal (i.e., 80 %) for all PSDs, filter
strips and swales may need to be paired with other technologies or
management practices to achieve this target. Given their effective
removal of larger particles, the TSS removal contributions of hydrody­
namic separators and high-flow media filters may be maximized on
roads with coarser PSDs, such as roads near low density residential
catchments.
CRediT authorship contribution statement
Ryan J. Winston: Conceptualization, Data curation, Formal anal­
ysis, Funding acquisition, Methodology, Supervision, Writing – original
draft, Writing – review & editing. Jon D. Witter: Conceptualization,
Funding acquisition, Project administration, Supervision, Writing – re­
view & editing. R. Andrew Tirpak: Data curation, Formal analysis,
Writing – review & editing.
Declaration of competing interest
Ryan Winston reports financial support was provided by Federal
Highway Administration. Ryan Winston reports financial support was
provided by Ohio Department of Transportation.
Data availability
Data will be made available on request.
Acknowledgements
This research was funded by the Ohio Department of Transportation
(job number 135258) and the Federal Highway Administration of the
United States of America. We appreciate the aid of Ohio Department of
Transportation (ODOT) staff in Districts 1, 2, 4, 6, 7, 8, and 12 who
contributed to identifying monitoring sites and ensuring the safety of
research personnel during sampling. We would like to thank Jon Prier
and Rebecca Humphries of ODOT central office for their coordination of
this research project. Shawn Kennedy of North Carolina State University
aided in instrumentation of the research sites and installed the weirs
used to measure flow in catch basins. Denin Miletto, Haley Jenkins,
Viktor Lillard, Lauren Sester, and David Wituszynski (formerly of Ohio
State University) performed various field sampling and/or laboratory
tasks that contributed to the success of this project. The biogeochemistry
laboratory at North Carolina State University is appreciated for their
efforts in performing particle size distribution analysis.
Appendix A. Supplementary data
Supplementary data to this article can be found online at https://doi.
org/10.1016/j.scitotenv.2023.166071.
11
Science of the Total Environment 902 (2023) 166071
References
American Public Health Association (APHA), American Water Works Association
(AWWA), Water Environment Federation (WEF), 2017. In: Rice, E.W., Baird, R.B.,
Eaton, A.D. (Eds.), Standard methods for the examination of water and wastewater,
23rd ed. Washington, DC.
Andral, M.C., Roger, S., Montréjaud-Vignoles, M., Herremans, L., 1999. Particle size
distribution and hydrodynamic characteristics of solid matter carried by runoff from
motorways. Water Environ. Res. 71 (4), 398–407.
Bakr, A.R., Fu, G.Y., Hedeen, D., 2020. Water quality impacts of bridge stormwater
runoff from scupper drains on receiving waters: a review. Sci. Total Environ. 726,
138068.
Barrett, M., Lantin, A., Austrheim-Smith, S., 2004. Storm water pollutant removal in
roadside vegetated buffer strips. Transport. Res. Record J. Transport. Res. Board
1890, 129–140.
Barrett, M.E., Walsh, P.M., Malina Jr., J.F., Charbeneau, R.J., 1998. Performance of
vegetative controls for treating highway runoff. J. Environ. Eng. 124 (11),
1121–1128.
Baum, P., Kuch, B., Dittmer, U., 2021. Adsorption of metals to particles in urban
stormwater runoff—does size really matter? Water 13 (3), 309.
Brodie, I.M., Dunn, P.K., 2009. Suspended particle characteristics in storm runoff from
urban impervious surfaces in Toowoomba, Australia. Urban Water J. 6 (2), 137–146.
Characklis, G.W., Wiesner, M.R., 1997. Particles, metals, and water quality in runoff
from large urban watershed. J. Environ. Eng. 123 (8), 753–759.
Charters, F.J., Cochrane, T.A., O’Sullivan, A.D., 2015. Particle size distribution variance
in untreated urban runoff and its implication on treatment selection. Water Res. 85,
337–345.
Chui, T.W., Mar, B.W., Horner, R.R., 1982. Pollutant loading model for highway runoff.
J. Environ. Eng. Div. 108 (6), 1193–1210.
Clary, J., Jones, J., Leisenring, M., Hobson, P., Strecker, E., 2020. International
Stormwater Best Management Practices (BMP) Database: 2020 Summary Statistics.
Available at: https://www.waterrf.org/system/files/resource/2020-11/DRPT-4
968_0.pdf.
Connolly, B.J., Loth, E., Smith, C.F., 2020. Shape and drag of irregular angular particles
and test dust. Powder Technol. 363, 275–285.
Davis, B.S., Birch, G.F., 2011. Spatial distribution of bulk atmospheric deposition of
heavy metals in metropolitan Sydney, Australia. Water Air Soil Pollut. 214 (1–4),
147–162.
Drapper, D., Tomlinson, R., Williams, P., 2000. Pollutant concentrations in road runoff:
Southeast Queensland case study. J. Environ. Eng. 126 (4), 313–320.
Dunn, O.J., 1961. Multiple comparisons among means. J. Am. Stat. Assoc. 56 (293),
52–64.
Ferreira, M., Stenstrom, M.K., 2013. The importance of particle characterization in
stormwater runoff. Water Environ. Res. 85 (9), 833–842.
Fritsch, J.M., Kane, R.J., Chelius, C.R., 1986. The contribution of mesoscale convective
weather systems to the warm-season precipitation in the United States. J. Clim. Appl.
Meteorol. 25 (10), 1333–1345.
Guo, Q., England, G., Johnston, C.E., 2008. Development of certification guidelines for
manufactured stormwater BMPs. In: World Environmental and Water Resources
Congress 2008: Ahupua'A.
Irish Jr., L.B., Barrett, M.E., Malina Jr., J.F., Charbeneau, R.J., 1998. Use of regression
models for analyzing highway storm-water loads. J. Environ. Eng. 124 (10),
987–993.
Järveläinen, J., Sillanpää, N., Koivusalo, H., 2017. Land-use based stormwater pollutant
load estimation and monitoring system design. Urban Water J. 14 (3), 223–236.
Jayarathne, A., Egodawatta, P., Ayoko, G.A., Goonetilleke, A., 2017. Geochemical phase
and particle size relationships of metals in urban road dust. Environ. Pollut. 230,
218–226.
Kayhanian, M., Suverkropp, C., Ruby, A., Tsay, K., 2007. Characterization and prediction
of highway runoff constituent event mean concentration. J. Environ. Manag. 85,
279–295.
Kim, L.H., Kayhanian, M., Lau, S.L., Stenstrom, M.K., 2005. A new modeling approach for
estimating first flush metal mass loading. Water Sci. Technol. 51 (3–4), 159–167.
Knight, E.M.P., Hunt, W.F., Winston, R.J., 2013. Side-by-side evaluation of four level
spreader-vegetated filter strips and a swale in eastern North Carolina. J. Soil Water
Conserv. 68 (1), 60–72.
Kobriger, N.P., Geinopolos, A., 1984. Sources and migration of highway runoff
pollutants. In: Rep. FHWA/RD-84/059 (PB86-227915J, V.S. Dept. of Transp., Fed.
Hwy. Admin., Washington, D.C).
Kruskal, W.H., Wallis, W.A., 1952. Use of ranks in one-criterion variance analysis. J. Am.
Stat. Assoc. 47 (260), 583–621.
Lee, D.H., Min, K.S., Kang, J.-H., 2014. Performance evaluation and a sizing method for
hydrodynamic separators treating urban stormwater runoff. Water Sci. Technol. 69
(10), 2122–2131.
LeFevre, G.H., Paus, K.H., Nataragan, P., Gulliver, J.S., Novak, P.J., Hozalski, R.M., 2015.
Review of dissolved pollutants in urban storm water and their removal and fate in
bioretention cells. J. Environ. Eng. 141 (1), 04014050.
Legret, M., Pagotto, C., 1999. Evaluation of pollutant loadings in the runoff waters from a
major rural highway. Sci. Total Environ. 235, 143–150.
Li, Y., Lau, S.-L., Kayhanian, M., Stenstrom, M.K., 2005. Particle size distribution in
highway runoff. J. Environ. Eng. 131 (9), 1267–1276.
Li, Y., Lau, S.-L., Kayhanian, M., Stenstrom, M.K., 2006. Dynamic characteristics of
particle size distribution in highway runoff: implications for settling tank design.
J. Environ. Eng. 132 (8), 852–861.
R.J. Winston et al.
Lucas-Aiguier, E., Chebbo, G., Bertrand-Krajewski, J.L., Gagné, B., Hedges, P., 1998.
Analysis of the methods for determining the settling characteristics of sewage and
stormwater solids. Water Sci. Technol. 37 (1), 53–60.
Lucke, T., Mohamed, M.A.K., Tindale, N., 2014. Pollutant removal and hydraulic
reduction performance of field grassed swales during runoff simulation experiments.
Water. 6 (7), 1887–1904.
Nason, J.A., Bloomquist, D.J., Sprick, M.S., 2012. Factors influencing dissolved copper
concentrations in Oregon highway storm water runoff. J. Environ. Eng. 138 (7),
734–742.
Ohio Environmental Protection Agency (OEPA), 2018. Storm Water Discharged From
Small and Large Construction Activities – General Permit. Permit No: OHC000005.
Opher, T., Friedler, E., 2010. Factors affecting highway runoff quality. Urban Water J. 7
(3), 155–172.
Opher, T., Ostfeld, A., Friedler, E., 2009. Modeling highway runoff pollutant levels using
a data driven model. Water Sci. Technol. 60 (1), 19–28.
Pitt, R., Maestre, A., Clary, J., 2018. The National Stormwater Quality Database (NSQD,
Version 4.02). Department of Civil and Environmental Engineering.
R Core Team, 2023. R: A Language and Environment for Statistical Computing. R
Foundation for Statistical Computing, Vienna, Austria. Available at: http://www.Rpr
oject.org.
Ramaswamy, R., Ben-Akiva, M., 1990. Estimation of highway pavement deterioration
from in-service pavement data. Transp. Res. Rec. 1272, 96–106.
Rommel, S.H., Helmreich, B., 2018. Influence of temperature and de-icing salt on the
sedimentation of particulate matter in traffic area runoff. Water 10 (12), 1738.
Rommel, S.H., Gelhardt, L., Welker, A., Helmreich, B., 2020. Settling of road-deposited
sediment: influence of particle density, shape, low temperatures, and deicing salt.
Water 12 (11), 3126.
Sansalone, J.J., Koran, J.M., Smithson, J.A., Buchberger, S.G., 1998. Physical
characteristics of urban roadway solids transported during rain events. J. Environ.
Eng. 124 (5), 427–440.
Selbig, W.R., 2015. Characterizing the distribution of particles in urban stormwater:
advancements through improved sampling technology. Urban Water J. 12 (2),
111–119.
Selbig, W.R., Bannerman, R.T., 2007. Evaluation of street sweeping as a stormwater-
quality management tool in three residential basins in Madison, Wisconsin. Scientific
Investigations Report 2007–5156.
Selbig, W.R., Bannerman, R.T., 2011. Characterizing the Size Distribution of Particles in
Urban Stormwater by Use of Fixed-point Sample-collection Methods. United States
Geological Survey. Report 2011-1052.
Selbig, W.R., Fienen, M.N., 2012. Regression modeling of particle size distributions in
urban storm water: advancements through improved sample collection methods.
J. Environ. Eng. 138 (12), 1186–1193.
Selbig, W.R., Fienen, M.N., Horwatich, J.A., Bannerman, R.T., 2016. The effect of
particle size distribution on the design of urban stormwater control measures. Water
8 (1), 17.
Semadeni-Davies, A.F., 2013. Classification of Stormwater-Borne Solids: A Literature
Review (Prepared by the National Institute of Water and Atmospheric Research
(NIWA) for Auckland Council. Auckland Council technical report, TR2013/017).
12
Science of the Total Environment 902 (2023) 166071
Shields, A., 1936. Application of similarity principles and turbulence research to bed-
load movement. In: Mitteilungen der Preussischen Versuchsanstalt fur Wasserbau
und Schifbau, Berlin (In: Ott, W.P. & van Uchelen, J.C. (translators), California Inst.
Tech., W.M. Keck Lab. of Hydraulics and Water Resources, Rept. No. 167).
Simpson, I.M., Winston, R.J., Brooker, M.R., 2022. Effects of land use, climate, and
imperviousness on urban stormwater quality: a meta-analysis. Sci. Total Environ.
809, 152206.
Smolek, A.P., Anderson, A.R., Hunt, W.F., 2018. Hydrologic and water-quality evaluation
of a rapid-flow bioinfiltration device. J. Environ. Eng. 144 (2), 05017010.
Stotz, G., 1987. Investigations of the properties of the surface water run-off from federal
highways in the FRG. Sci. Total Environ. 59, 329–337.
Toronto and Region Conservation Authority, 2002. Performance Assessment of a Pond-
Wetland Stormwater Management Facility – Markham, Ontario (Prepared by:
Stormwater Assessment Monitoring and Performance (SWAMP) Program).
Treilles, R., Gasperi, J., Saad, M., Tramoy, R., Breton, J., Rabier, A., Tassin, B., 2021.
Abundance, composition and fluxes of plastic debris and other macrolitter in urban
runoff in a suburban catchment of Greater Paris. Water Res. 192, 116847.
U.S. EPA, 2002. Urban Storm Water BMP Performance Monitoring: A Guidance Manual
for Meeting the National Storm Water BMP Database Requirements. In: EPA-821-B-
02-001. U.S. Environmental Protection Agency, Washington, DC.
USDA (U.S. Department of Agriculture), 1975. Soil Taxonomy: A Basic System of Soil
Classification for Making and Interpreting Soil Surveys. Washington, DC.
Water Environment Research Foundation (WERF), 2008. Improved Protocol for
Classification and Analysis of Stormwater-Borne Solids. Report No. 04SW4.
Westerlund, C., Viklander, M., 2006. Particles and associated metals in road runoff
during snowmelt and rainfall. Sci. Total Environ. 362 (1), 143–156.
Winston, R.J., Hunt, W.F., 2017. Characterizing runoff from roadways: particle size
distributions, nutrients, and gross solids. J. Environ. Eng. 143 (1), 04016074.
Winston, R.J., Hunt, W.F., Anderson, A.R., 2017. Modeling sediment reduction in grass
swales and vegetated filter strips using particle settling theory. J. Environ. Eng. 143
(1), 04016075.
Winston, R.J., Witter, J.D., Tirpak, R.A., Sester, L., Jenkins, H., Lillard, V., 2023.
Abundance and composition of anthropogenic macrolitter and natural debris in road
runoff in Ohio, USA. Water Res. 239, 120036.
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 (7), 584–592.
Wu, J., Ren, Y., Wang, X., Wang, X., Chen, L., Liu, G., 2015. Nitrogen and phosphorus
associating with different size suspended solids in roof and road runoff in Beijing,
China. Environ. Sci. Pollut. Res. 22 (20), 15788–15795.
Yazdi, M.N., Sample, D.J., Scott, D., Wang, X., Ketabchy, M., 2021. The effects of land
use characteristics on urban stormwater quality and watershed pollutant loads. Sci.
Total Environ. 773, 145358.
Yu, S.J., Kaighn, R.J., Shih-Long, L., 1994. Testing of Best Management Practices for
Controlling Highway Runoff: Phase II. Virginia Transportation Research Council,
Charlottesville, VA.
Zhao, H., Ma, Y., Fang, J., Hu, L., Li, X., 2022. Particle size distribution and total
suspended solid concentrations in urban surface runoff. Sci. Total Environ. 815,
152533.