Water Research 145 (2018) 509e517

Contents lists available at ScienceDirect

Water Research
journal homepage: www.elsevier.com/locate/watres

Hydrological benefits of filtering swales for metal removal

C. Monrabal-Martinez a, *, J. Aberle a, b, T.M. Muthanna a, M. Orts-Zamorano a
a
Department of Civil and Environmental Engineering, Norwegian University of Science and Technology (NTNU), NO-7491, Trondheim, Norway
b €t Braunschweig, DE-38106, Braunschweig, Germany
Leichtweiß Institute for Hydraulic Engineering and Water Resources, Technische Universita

a r t i c l e i n f o a b s t r a c t

Article history: This paper studies the hydraulic performance of two swales composed of filters for stormwater man-
Received 5 April 2018 agement (filtering swales) in a large-scale experimental study and compares them to the performance of
Received in revised form a swale composed of traditional bioretention soil (bioswale). Using experimental data, dimensionless
17 July 2018
formulations are derived to reflect the influence of swale design parameters on hydraulic performance.
Accepted 24 August 2018
Available online 24 August 2018
The developed formulas can be used to design swales accounting for practical factors for decision makers
such as local rainfall patterns, volume capture requirements, and drainage area. The experimental data
show that while the bioswale is characterized by large overland flows, the tested filtering swales manage,
Keywords:
Stormwater runoff
in the majority of cases, the complete inflow volume without overland flow. The longitudinal slope of the
Roadside filters swales does not affect the infiltration capacity of the filtering swales for the tested experimental
Hydrological balance boundary conditions, only the inflow rate and media water content are found to be statistically signif-
Swale design icant. As an example, filtering swales tested in this study captured 90% of the runoff generated by a
Infiltration performance 12.2 mm/h storm (approximately a 5-year return period 1-h duration storm event in the city of Trond-
heim) on a road 40 times larger than the swale. This highlights the capacity of such swales for handling
infrequent events.
© 2018 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND
license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction maintain the pre-development hydrologic balance of a site (Cizek

Urban runoff is well known for carrying a wide range of pol-
lutants (Hoffman et al., 1985; Huber et al., 2016c; LeFevre et al.,
2014) and the sources of pollution differ across locations and sea-
sons (Gunawardana et al., 2014; Moghadas et al., 2015). Infiltration
based solutions for the management of urban runoff have shown a
good performance with regard to pollution mitigation and volume
reduction (i.e. retention). Generally, treatment swales represent
such an infiltration-based solution, adopted as a low-tech practice
to deal with roadway runoff (Burack et al., 2008). They collect
polluted runoff generated on impervious surfaces such as roads and
convey it to a downstream wastewater network or a receiving
water after a minimum hydraulic residence time within the swale.
Sedimentation, infiltration, filtration, adsorption, and vegetation
uptake are the main processes occurring in these systems. There-
fore, they provide multiple environmental benefits such as volume
retention, pollutant removal, and percolation, which helps to
* Corresponding author.
E-mail addresses: carlos.monrabal-martinez@multiconsult.no (C. Monrabal-
Martinez), jochen.aberle@ntnu.no, jochen.aberle@tu-braunschweig.de (J. Aberle),
tone.muthanna@ntnu.no (T.M. Muthanna).
and Hunt, 2013).
Factors such as the ratio of the impervious road area to the swale
area, the size of the storm event, and the longitudinal slope
determine the overland discharge and flow depth, and hence the
flow characteristics (Winston et al., 2016). The overland flow in
swales can be either super- or subcritical. Supercritical flows in-
crease erosion, reduce volume and pollution retention, and should
therefore be avoided. Subcritical flows can be obtained by con-
structing specific design elements increasing the hydraulic resis-
tance. Such design elements include check dams (e.g., Davis et al.,
2012; Winston et al., 2018), stones (e.g., Narsimlu et al., 2004)
and grass coverage (e.g., Mishra et al., 2006). The embankment
slopes of swales can provide a bonus for volume retention, which in
turn is positively impacted by the infiltration capacity of the media,
soil moisture deficit, and side-slope length (García-Serrana et al.,
2017). For vegetated swales, the type of vegetation such as deep-
rooted grass species and proliferation of root and earthworm
channels improve infiltration (Abu-Zreig et al., 2004; Blanco-
Canqui et al., 2004). Regarding soil composition, bioswales are,
generally, composed of soils such as loamy sands, loams, or sandy
loams, providing infiltration rates higher than recommended by bio
filter design guidelines (Le Coustumer et al., 2009).

https://doi.org/10.1016/j.watres.2018.08.051
0043-1354/© 2018 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
510 C. Monrabal-Martinez et al. / Water Research 145 (2018) 509e517
In some cases, larger infiltration capacities are required to; (1)
handle larger runoff events; (2) consider cold climate constraints,
such as reduced winter infiltration capacity (Paus and Braskerud,
2014), or (3) runoff from large impervious areas to swales by
keeping a certain removal capacity to meet discharge regulations
(Huber et al., 2016b). In addition, a good drainage capacity gua-
rantees low initial moisture in the media, which has been shown to
positively impact volume retention (Lucke et al., 2014). These facts
have motivated the investigation of materials for the construction
of swales offering an added performance in terms of filtration ca-
pacity, pollutant removal, or both (Kalmykova et al., 2008; Wium-
Andersen et al., 2012; Ilyas and Muthanna, 2017; Huber et al.,
2016a).
The end user (urban drainage engineer/practitioner) is, gener-
ally, interested in mathematical models that reconcile laboratory
findings to field implementation. For example, TRAVA is a 1-D
deterministic model (based on Green-Ampt and Kinematic Wave
equations) for non-submerged flow conditions and simulates the
generation of runoff and sediment transport (Deletic, 2001). The
comparison of simulation results with data from laboratory and
field experiments showed that the hydraulic conductivity of the soil
is the most sensitive parameter in the simulations (Deletic, 2005;
Deletic and Fletcher, 2006b). Moreover, the application of VFSMOD
(Mun ~ oz-Carpena et al., 1999), a model for similar purposes, showed
that the type of soil and its initial water content determines the
runoff retention. The application of VFSMOD further showed grass
spacing to be an important parameter regarding sediment removal
(Han et al., 2005). Both these models require extensive input pa-
rameters, which are typically not available for the engineers
designing swales for stormwater management purposes. This
makes empirical models generally more attractive for the user
because of their simplicity in input parameters, while providing
good predictions of, for example, trapping efficiency for specific
swale designs and residence times (Ba €ckstro€ m, 2002; Winston
et al., 2016). This motivation based on the end user applicability
resulted in testing a series of empirical equations with the aim to
find the dominant design factors. More specifically the research
questions in this study are as follows:
(1) What are the hydrological benefits of adsorbent amended
filters for roadside filters (referred to as filtering swales from
this point on) in comparison to bioswales?
(2) Which are the dominant design factors influencing the hy-
drological response of filtering swales?
(3) How can hydrological benefits be compared using a practical
approach such as the capture of storm events?
In order to answer these questions, a full-scale swale setup was
designed to evaluate factors determining the hydrological balance
of swales composed of filters studied by Monrabal-Martinez et al.
(2017). These filters are constructed with a pure sand, which was
amended with either olivine, or pine bark. At the first phase of
column scale tests the filters showed high removal rates for the
fours metals of concern (Cu, Pb, Ni, and Zn). The column test also
included several different hydraulic loadings, and periods of
infrequent storm events (rapid inflows), followed by longer dry
periods. These results, in turn, were the motivation for the present
study to understand the hydrological response of full-scale filters,
in order to confirm their potential for use in stormwater manage-
ment. The experimental data was used to propose empirical
equations for the evaluation of hydrological processes at the full-
scale for the filters, also including a comparison with a swale
composed of commonly used soil for bioretention. The results from
this study provides new and much needed improved understand-
ing of the dominant processes and factors for infiltration based
treatment filters and bioswales.
2. Material and methods
Three tiltable 500 cm long, 45 cm deep, and 80 cm wide swales
(see Fig. 1) were built at the hydraulic laboratory of the Norwegian
University of Science and Technology (NTNU). Each swale had an
underdrain layer composed of two perforated pipes with a diam-
eter of 5 cm embedded in a 10 cm thick gravel layer. Filtering swale
1 was composed of a clean homogeneous sand layer with a thick-
ness of d ¼ 40 cm (D50 ¼ 0.33 mm, which is the intercept for 50% of
the cumulative mass of the particle size distribution; uniformity
coefficient D60/D10 ¼ 2.22) overlying a 5-cm thick layer of pine bark.
Filtering swale 2 was composed of the same 40 cm quartz sand
layer but was overlying a 5 cm thick layer of granulated olivine
(D50 ¼ 1.4 mm; uniformity coefficient ¼ 1.66). Swale 3, built as
bioswale, had its entire 45 cm thickness composed of a traditional
soil for bioretention, classified as a sandy loam (USDA, 2017).
Swales 1 and 2 were superficially protected with boulders
(D50 ¼ 70 mm; approximately 256 boulders/m2; placed to reduce
filter exposure). Swale 3 was seeded (10 gr/m2) with Festuca Rubra
Rubra (Leik), a cold tolerant species with a high salinity tolerance,
which is favourable for roadside vegetation. High-pressure sodium
lamps were arranged in order to provide a photosynthetic active
radiation of approximately 300 W/m2 during 18 h/day. The height
of grass was intended at 7 cm during the experiments. Unlike most
roadside swales, the swale media in this study were not mechan-
ically compacted during filling as they are meant for infiltrating
runoff and no traffic is expected over them.
Each swale was equipped with 6 profile probes PR2/4-SDI-12
from Delta-T Devices Ltd, and each profile had three moisture
dielectric sensors distributed along the vertical axis (Fig. 1),
resulting in a total of 18 moisture sensors per swale. The moisture
sensors measured the permittivity of the surrounding soils with a
sampling frequency of 0.5 Hz and were specifically calibrated for
each media. Errors of ±4%, including installation and sampling er-
rors, can be expected according to the manufacturer of the sensors.
Each swale was divided in 18 volumetric sectors (with a volume
Vi (m3)) according to the distribution of the media moisture sen-
sors. The total moisture present in the swale at time t, Vw(t) (m3),
was computed as
X
18
VwðtÞ ¼ qiðtÞVi (1)
i¼1
where qi(t) is the volumetric water content in sector i measured by
the associated moisture sensor at time t. For the determination of
the saturated moisture content, Vsat (m3), by equation (1), the
swales were flooded by closing the valve on the underdrain. During
the experiments, the degree of saturation in the swale at time t, Sr
(t), was defined as the ratio of the total moisture at time t, Vw(t), and
the saturated water content of the swale, Vsat.
A 1-m3 tank filled with tap water served as upstream reservoir
for the experiments. From the tank, the water was pumped into a
small reservoir at the upstream end of the swale that was equipped
with a rectangular weir allowing for an even distribution of the
inflow (run-in) over the swale width. The inflow rates were
measured by an electromagnetic flowmeter (Siemens Sitrans FM
MAG 5000; accuracy ± 0.4% of the flow rate) and controlled with a
valve. The overland flow at the end of the swale (run-out) and the
infiltrated water were collected separately in two downstream
tanks with volumes of 0.72 m3 (1.1 m wide, 1.1 m long, and 0.6 m
high) and 0.21 m3 (0.5 m diameter and 1.1 m high), respectively.
The water surface level in these tanks was measured with two
 C. Monrabal-Martinez et al. / Water Research 145 (2018) 509e517 511

Fig. 1. (a) Sketch of the setup of swales 1 to 3, (b) lateral view of a filtering swale composed of an adsorbent amended filter.

ultrasonic sensors (Microsonic micþ130/IU/TC; average resolution
0.4 mm and ±1% accuracy) with a sampling frequency of 1 Hz
allowing for the determination of the accumulated water volume in
the tanks as a function of time. Due to volume limitations of the
downstream tanks, they had to be emptied during some experi-
ments. As a consequence, linear extrapolation based on previous
filling data was used in order to fill the data gap during emptying.
The experiments with swales 1e3 were carried out for a range of
boundary conditions (Table 1) by varying the initial degree of
saturation (Sr0), the longitudinal slope (J), and the inflow rate (Qrun-
in). Each filter media was characterized by its saturated hydraulic
conductivity, Ksat (m/s), which was determined based on three
measurements along its length using the MPD infiltrometer
method (Ahmed et al., 2015). Swale slopes of 2, 3 and 4% were used
in the experiments; such slopes are common in road design in
Norway (Vegvesen, 2008).
The ratio I/P between impervious (I) and pervious areas (P) is
often used to size systems for stormwater runoff (Khan et al., 2012).
In order to mimic practical implications, Qrun-in was chosen as the
equivalent runoff generated by an event with a given storm in-
tensity, SI (mm/h), occurring over an impervious area I/P times as
large as the swale area, As (m2), by using the following relationship
which is based on the rational method (Kuichling, 1889):
Q limit the duration of the experiment, and because, visually, this rate
SI ¼ I runin 36,105
C P þ 1 As
=
The empirical runoff coefficient, C, can be assumed to be close to
1 because of the imperviousness of road surfaces. Equation (2)
considers also the precipitation that falls on the swale itself,
which might be of relevance for small I/P ratios. The inflow rates
ranged from 0.2  103 m3/s to 1  103 m3/s, which corresponds,
for an I/P ¼ 40 and C ¼ 0.9, with storm intensities from 5 to 25 mm/
h, respectively.
Each test-run was subdivided into a pre-loading, quasi-equi-
librium, and drainage phase (Fig. 2). The pre-loading phase was
defined as the time required to reach quasi-equilibrium saturation
of the soil from the start of water inflow. During this phase, the
water mainly filled the non-saturated pore space in the media,
which explains the quicker increase in the degree of saturation of
the media (DSr) with regard to the subsequent phase. The quasi-
equilibrium condition was assumed to be reached when DSr was
less than 4% in four consecutive minutes. This definition was
adopted in order to avoid an early change of phase because DSr
followed in some cases a distribution with several steps during the
preloading phase (Fig. 2b). During the quasi-equilibrium phase, the
media moisture was nearly constant, and the run-in water could be
computed as the sum of the infiltration and run-out water collected
downstream. The quasi-equilibrium phase was hence defined by
the time period in which the media moisture was nearly constant.
The drainage phase was defined as starting when a decrease of the
media moisture was monitored and delimited to 20 min after
stopping the inflow. The latter definition was chosen in order to
(2)
(drainage) typically decreased 20 min after the inflow ceased.
Dimensionless analysis is a powerful tool and its application is

Table 1
Range of the experimental boundary conditions. Srmax is the maximum degree of saturation reached during a simulation, which typically occurred slightly after stopping the
inflow.

Swale Volumetric field capacity (%)a Volumetric field saturated water content (%)b Ksat (m/s) Qrun-in 103 (m3/s) Slope (%) Sr0 Srmax

Swale 1 16 30 (6.45 ± 1.35) 104 0.2e1 2e4 0.51e0.67 0.67e0.91

Swale 2 16 30 (5.59 ± 1.14) 104 0.2e1 2e4 0.56e0.72 0.70e0.97
Swale 3 37.5 52 (6.53 ± 1.04) 105 0.2e0.7 2e4 0.71e0.91 0.95e1
a
Volumetric field capacity ¼ water content after draining for more than 48 h ¼ Volume of water/Volume of swale.
b
Volumetric field saturated water content ¼ water content when all pores are filled with water.
512 C. Monrabal-Martinez et al. / Water Research 145 (2018) 509e517
Fig. 2. Degree of saturation as a function of time in swale 3 (a) and swale 1 (b) during a test with 0.3  103 m3/s run-in and 2% longitudinal slope for both swales.
common in hydraulic engineering applications. Derived relation-
ships can be upscaled since results are dimensionally homogenous.
Therefore, the data were analyzed in a dimensionless framework
using P-parameters (Buckingham, 1914) constructed using the
depth, d (m), and width, w (m), of the swales as well as the ac-
celeration of gravity, g (m2/s) as returning variables (see below).
The operational parameters considered in the analysis were the
swale slope J, the inflow rate Qrun-in and the saturated hydraulic
conductivity Ksat of the different media, as these parameters have
been shown to impact the infiltration performance of swales
(Deletic and Fletcher, 2006a; García-Serrana et al., 2017). During
the quasi-equilibrium phase, the moisture content was considered
constant and the hydrological performance of the swale was eval-
uated by the flow rate Qinf. This flow rate was determined from the
water collected at the infiltration tank, which corresponded
approximately to the difference between run-in and run-out flow
rate. Based on the derived P-parameters, the infiltration rate was
parameterized according to:
!a
Qinf Ksat Qrunin
¼A Jb
g 1=2 ðwdÞ5=4 g1=2 ðwdÞ1=4 g 1=2 ðwdÞ5=4
where A, a, b, and g are dimensionless coefficients. Experimental
data were used to determine these coefficients by regression ana-
lyses. To characterize swale performance in the final phase, the
decrease in media moisture in the first 20 min after stopping the
inflow, DSr20 (%), was adopted as parameter to assess the drainage
capacity of the swale. The type of media, the swale slope and the
media moisture at the time the inflow stopped were used to model
the drainage capacity of the swale. The acquired data were hence
used to parameterize the coefficients r, B, s, and t in the empirical
relationship:
!r
Ksat
DSr20 ¼ B J s Strmax
g1=2 ðwdÞ1=4
which follows from the dimensional analysis.
The relationships given by equations (3) and (4) were trans-
formed into log-log-space to facilitate multiple linear regression
analysis. Data from 37 experiments, performed under different
boundary conditions, were used for ANOVA analysis to obtain sta-
tistically significant hydrological variables as well as associated
coefficients composing the empirical models (p-value < 5%). A total
of 16 independent experiments were used to validate the derived
relationships. The R2 values were used to assess the goodness of fit
of the models.
3. Results and discussions
The hydrological response of swale 1 and 2 differed to a great
extent from the response of swale 3 (traditional soil for bio-
retention). This can be observed in Figs. 2 and 3 showing the degree
of media saturation over time and the degree of media saturation
together with the measured volumes of run-in, infiltration, and
run-out, respectively. As indicated by Fig. 2a and b, the initial de-
gree of saturation was different between swale 3 and 1, which
makes the pre-loading phase not comparable despite the fact the
tests had similar boundary conditions, i.e., run-in rate and slope.
However, compared to swale 3, swale 1 showed a considerably
quicker degree of saturation response during the 20-min drainage
phase (DSr/min was 0.64 in swale 1 and 0.17 in swale 3). This can be
associated with the high drainage capacity of the filter, which
quickly empties the water stored in the media. In addition, unlike
!g swale 3 that reached full saturation during the quasi-equilibrium
(3) phase, swale 1 only reached 73% of its storage capacity for the
same experimental conditions. The occurrence of saturation im-
plies the appearance of overland flow in the swale, i.e. some of the
run-in water will not be captured by the swale or treated by the
media. However, full saturation was not reached in the filtering
swales (Srmax < 1, Table 1) even though run-out was collected in
tests where Qrun-in ¼ 1  103 m3/s. In these tests, the run-in
discharge was so large that there was not enough hydraulic
retention time to saturate the filter media before run-out was
observed.
Fig. 3 presents the time-development of the degree of saturation
and cumulative run-in, infiltration, and run-out volumes for the
three swales for an experiment with 4% slope and Qrun-
3
in ¼ 0.3  10 m3/s. Fig. 3a and b shows that the slope of the cu-
mulative inflow volume and the infiltrated volume are equal for
(4) swales 1 and 2 with quasi-equilibrium conditions being satisfied.
This can be explained by the fact that after a lag time, corre-
sponding to approximately the time to reach quasi-equilibrium,
practically all run-in was infiltrated. As no overland-outflow was
observed, the differences between the inflow and infiltrated vol-
ume are associated with the change in the media moisture, which,
in average, is equivalent to 5.6% of the run-in volume (Table 2). In
contrast, in the experiments with swale 3, overland flow was
registered at the downstream end (Fig. 3c); thus, only swale 1 and 2
could completely infiltrate the event represented in Fig. 3. Addi-
tionally, Table 2 shows that the amount of run-in used in swale 1
 C. Monrabal-Martinez et al. / Water Research 145 (2018) 509e517 513

Fig. 3. Time-development of the degree of saturation (dash dot dot line), and cumulative volumes of run-in, infiltration, and run-out in swale 1 (a), swale 2 (b), and swale 3 (c)
during experiments with 4% slope and 0.3  103 m3/s run-in discharge.
514 C. Monrabal-Martinez et al. / Water Research 145 (2018) 509e517
and 2 was around 20% larger than in swale 3. In these experiments,
it was decided to extend the duration of inflow to observe run-out.
However, the media moisture analysis confirmed that after reach-
ing quasi-equilibrium, the water balance in the swale remained
practically stable and, therefore, no run-out would have happened
regardless of the duration of the inflow. In addition, the infiltration
volume presented in Table 2 highlights a similar response of swale
1 and 2 towards run-in flows regardless of the adsorbent used in
the media. Note that full saturation in swale 3 can lead to larger
uncertainties in the moisture sensors because in such cases the
sensors are operating outside their factory given range (0e40%,
volumetrically). This uncertainty could lead to an under estimation
of the moisture increase in the media, which explains the larger
imbalance that was observed in swale 3.
The porosity of the soil composing swale 3 (volumetric field
saturated water content ¼ 52%, Table 1) was larger than the
porosity of the sand used in swale 1 and 2 (volumetric field satu-
rated water content ¼ 30%), which apparently should allow for
retaining larger water volumes. However, the overall capacity of
swale 3 for handling run-in discharges was considerably less
(evaluated by the fact that the presence of run-out was observed
with Qrun-in ¼ 0.2  103 m3/s in swale 3 and Qrun-in ¼ 1  103 m3/s
in swales 1 and 2). Several points can be used to explain this
observation; 1) the hydraulic gradient responsible of the driving
forces for water infiltration was lower in swale 3 from the begin-
ning of the experiments due to high initial soil water contents
(0.71 < Sr0 < 0.91; in contrast to the filtering swales where
0.51 < Sr0 < 0.72); 2) for swale 3, Srmax was close to 1 (Table 1) in all
experiments, implying that Vw at quasi-equilibrium conditions
equals Vsat. At saturation, the infiltration rate reaches a constant
value (Ksat), and overflow occurs when the run-in rate exceeds this
value. It was observed that run-in discharges used during the ex-
periments exceeded the infiltration capacity of swale 3
(Ksat ¼ 6.53  10-5 m/s), which was approximately 10 times lower
than the infiltration capacity of the alternative filters (in average
Ksat ¼ 6  10-4 m/s, see Table 3); 3) The soil nature of swale 3 might
be susceptible of developing surface crusting (sealing), which in
turn reduces the infiltration.
Despite the fact that the swales were drained in a similar
manner, the lowest initial degree of saturation in swale 3 was 0.71
(Table 1), which practically corresponds to the highest initial de-
gree of saturation observed in swale 2 (0.72) and which is still
higher than the highest initial degree of saturation in swale 1 (0.67)
(note that volume retention is positively impacted by low initial
moisture in the media (Lucke et al., 2014)). This may be explained
Table 2
Volumetric balance of the experiments presented in Fig. 3. Percentage over the corresponding run-in volume is shown in parentheses.
Experiment Run-in (L) Run-out (L)
Swale 1 1218.9 (100) 0 (0)
Swale 2 1279 (100) 0 (0)
Swale 3 1025.8 (100) 579.13 (49.8)
Table 3
Ksat values measured with MPD test and coefficients for equations (3) and (4). Zero implies that the associated parameter was non-statistically significant according to the
ANOVA test. R2 represents the goodness of the model in the validation.
Ksat (m/s) Equation (3)
A a b g
Swale 1 (6.45 ± 1.35)$104 e0:831 e 0
Swale 2 (5.59 ± 1.14)$104 e1:437 e 0
Swale 3 (6.53 ± 1.04)$105 e8:798 e 0.298
Joint Swale 1 þ 2 e e1:222 0 0
by the presence of fines (silt and clay particles) composing the soil
(Ke and Takahashi, 2012), which were absent in swales 1 and 2. The
presence of fines and organic matter will favor the retention of
gravimetric, capillary, and adsorbed water and will subsequently be
detrimental to the effective porosity and impede a low initial de-
gree of saturation.
Overall, swale 3 had lower handling capacity as well as initial
storage, which accelerated the appearance of run-out. This has
direct consequences for handling the first and, generally, most
polluted fraction of runoff. An early run-out appearance might pose
an important risk to receiving water bodies due to large untreated
pollutant loads associated with the first flush (Sansalone and
Buchberger, 1997).
Festuca Rubra shows a mean root diameter similar to species
that have been shown to improve soil hydraulic conductivity by the
creation of macro-pores (Fort et al., 2012; Le Coustumer et al.,
2012). In addition, the observed flow depth was always lower
than the grass height, which favors its hydraulic resistance and,
subsequently, the infiltration potential of the swale (Winston et al.,
2016). However, these enhancing effects were observed insufficient
to counteract the presence of fines.
The statistically significant factors (p-value < 5%) in the empir-
ical models defined by equations (3) and (4) were used to repro-
duce the behaviors of the swales (Table 3). The model defined by
equation (3) describing the quasi-equilibrium phase performed
well for the sandy filters (non vegetated swales 1 and 2) but poor
for the traditional media (swale 3) due to the insufficient infiltra-
tion capacity of the soil for the experimental boundary conditions.
Two different responses were observed with regard to the influence
of slope on infiltration. Fig. 4 shows that Qinf, for the filtering
swales, is only affected by Qrun-in, and the slope has no effect on
infiltration. However, for the bioswale, Qrun-in is independent of
Qinf, and the slope has an impact on Qinf, i.e., gentler slopes allow for
higher infiltration, as Mun ~ oz-Carpena et al. (1993) showed. This
confirms the hypotheses that; 1) the critical slope over which the
peak runoff becomes constant was not achieved in swale 3 (Nassif
and Wilson, 1975); 2) the filtering swales 1 and 2 have not devel-
oped any surface sealing (in the absence of surface sealing the slope
does not influence the infiltration rate; Poesen, 1984). It is worth
mentioning that conflicting conclusions have been reported in the
literature regarding the role of the slope for infiltration due to many
confounding factors such as incoming runoff, micro topography,
vegetation type, soil texture, and vertical and horizontal hetero-
geneity in soil properties (Morbidelli et al., 2018).
The media moisture content is an important factor for
Infiltration (L) Media moisture increase (L) Imbalance (L)
1116.5 (91.6) 52.1 (4.3) 49.8 (4.1)
1160.6 (90.7) 88 (6.9) 30.4 (2.4)
200.9 (33.4) 145.4 (14.2) 96.7 (9)
Equation (4)
R2 B r s t R2
0.902 0.99 e1:3 e 0 3.1 0.97
0.828 0.95 e1:468 e 0 3.025 0.99
0 0.03 e2:739 e 0 0 0.04
0.85 0.96 e7:656 1.089 0 3.054 0.84
 C. Monrabal-Martinez et al. / Water Research 145 (2018) 509e517
Fig. 4. Infiltration rates vs run-in rates in swale 1 (a), swale 2 (b), and swale 3 (c)
during the quasi-equilibrium phase.
515
considerations of the transition phases, i.e., pre-loading and
drainage. The time to reach stationary conditions depends on the
soil type as well as the initial water content (Fig. 2). The water
content of the media at the end of the storm event strongly
impacted the drainage capacity of swales 1 and 2. In these cases, the
slope was not statistically significant when parameterizing the
models (Table 3). As mentioned above, swale 3 reached practically
in all tests full saturation, i.e., Srmax ¼ 1. This fact might explain that
swale 3 did not depend on the degree of saturation, and its behavior
could not be explained by any of the proposed variables in the
model defined by equation (4); however, the drainage capacity of
the alternative filters can be well predicted with equation (4).
The drainage capacity of swales is an important design factor
because it influences the capacity for handling the next storm
event. In cold climates, drainage gains more relevance because of
the importance of avoiding high water content under negative
temperatures, which can lead to the formation of concrete frost
with practically absent infiltration (Muthanna et al., 2007). In this
context, Figs. 2 and 3 show a steeper drainage slope in the alter-
native filters (DSr/min ¼ 0.64, 0.77, and 0.17 in swale 1, 2 and 3,
respectively) and, therefore, a more desirable drainage response for
stormwater management in cold climates.
Guidelines for sizing sustainable urban drainage systems (SuDS)
vary widely (Guo et al., 2013). Generally, a water quality volume
capture can be adopted as a design parameter in order to guarantee
the removal of the majority of stormwater pollution on an average
annual basis. This runoff volume capture generally refers to the
volume of runoff generated from a given percentile (normally, be-
tween 80th and 90th) of individual 24-h rain events (Schueler and
Claytor, 1996; Urbonas et al., 1989). Swales are flow-based treat-
ment practices and, instead, water quality flow rates should be used
for sizing (Burack et al., 2008). Therefore, the run-in rates (Qrun-in)
used in this study were extrapolated using equation (2) to estimate
the road runoff generated by a rainfall event. Subsequently, the
coefficients reported in Table 3 (swale 1 þ 2) were used in equation
(3) to compute the percentage of run-in that will be infiltrated,
captured and treated by the media. It is noteworthy to mention that
Monrabal-Martinez et al. (2017) showed an effective treatment of
polluted water (spiked with 1 g/l of Pb, Zn, Cu, and Ni) through the
rapidly flowing filters used in swale 1 and 2.
Table 4 can be used as a guideline for dimensioning swales
composed of alternative filters similar to swales 1 and 2. It presents
the rain intensity events that can be captured at a given percentage
according to the size of the road (defined by the ratio I/P). As an
example, for a 100% volume capture and a design storm event of
5 mm/h, the filtering swale will handle the runoff from a catchment
40 times its size. If the capture requirement is set at 90%, a 12.2 mm/
h storm event could be handled by the system for the same I/P ratio.
The latter storm intensity is slightly less intense than a 5-year 1-h
rainfall for the city of Trondheim. Therefore, Table 4 can help to
design filtering swales according to local rainfall patterns and
volume capture requirements. Note that infiltration systems with
large ratio I/P make better use and advantage of urban space,
although such I/P ratios might accelerate clogging due to high
sediment loading rates (Le Coustumer et al., 2012).
The infiltration capacity of the media is an important design
factor for filtering swales. Additionally, preventative measures
against clogging are crucial for assuring long-term service life and
functioning according to design requirements. Established vegeta-
tion can counteract the clogging phenomena because of the
development of a root network as well as the activity of earth-
worms (Paus et al., 2013). Therefore, systems that lack vegetation
will require a closer study and control with regard to sediment
pretreatment, clogging prevention and maintenance; for example,
pretreatment swales have been found to be efficient for
516 C. Monrabal-Martinez et al. / Water Research 145 (2018) 509e517
Table 4
Combination of storm intensity-catchment size-volume capture for swales composed of filter 1 or 2.
Storm intensity, SI (mm/h) Impervious area to swale area, I/P
5 20
Volume capture (%) 100 34.2 9.8
90 83.3 23.8
80 166.7 47.6
downstream treatment systems with regard to suspended particles
(Lucke et al., 2014), and periodic replacement of the 5e10-cm layer
of top media has been shown enough for hydraulic conductivity
restoration (Li and Davis, 2008).
Several considerations must be accounted for when interpreting
results from this study, which could differ from field conditions:
 The maximum slope tested in this study was 4%. However, as a
treatment system is often placed parallel to road, slopes are not
expected to exceed this value. Filtering practices should avoid
steep slopes for optimal performance and, therefore, should be
implemented in nearby flatter areas (Yousef et al., 1987; Samani
and Kouwen, 2002).
 By using the rational method to relate storm intensities (SI) to
runoff generated (which is the run-in flow in the experiments),
it is assumed that the rainfall intensity is constant over time and
over the impervious area. In real conditions, rainfall intensity
oscillations and the existence of lag-time until full catchment
contribution will provide a runoff discharge pattern that favours
the infiltration response.
 By design, the runoff was assumed to flow directly into the
upstream end of the swale. Therefore, no initial abstraction of
water by filter strips or side slopes was taken into account,
which would have contributed to a better overall performance
(Flanagan et al., 2017).
 Unlike in field scenarios, the water used in this study was
sediment-free. Sediments carried by real runoff will, to some
extent, affect the infiltration capacity due to clogging of pore
space. However, for a long-term service life it is highly recom-
mended to maximize the retention of sediments prior to
entering the adsorbent amended section of the treatment train.
 Swale 3 showed that the traditional soil used in bioretention
was unpredictable using our modelling methods within the
studied boundary conditions.
4. Conclusions
The hydrological processes occurring in three swales designed
for stormwater management were studied under controlled labo-
ratory conditions. Two of the swales were composed of promising
filters for stormwater management, while the third swale was
composed of a typical soil used for bioretention areas. Boundary
conditions that varied among the experiments were the type of
media, the initial degree of saturation, the swale slope, and the
inflow rate.
In addition to the previously demonstrated metal removal ca-
pacity, the filtering swales showed a considerably improved hy-
drological response compared to the traditional bioswale. This was
evaluated, for a common boundary setup, by the overland flow
collected at the end of the swale (volume retention) in addition to
the drainage capacity. The swale composed of traditional bio-
retention soil registered large run-out volumes for all the inflow
intensities of Qrun-in, while the filtering swales managed the run-in
volume completely up to Qrun-in ¼ 0.7  10-3 m3/s. In addition, the
40 60 80 100
5.0 3.4 2.5 2.0
12.2 8.2 6.2 5.0
24.4 16.4 12.3 9.9
initial degree of saturation was considerably higher in the bioswale,
which negatively impacted the retention capacity of the
infiltration-based system with earlier onset of run-out flow
(overflow).
The models derived in this study explained the variability of the
target variables used to assess the performance in the alternative
media satisfactory. However, the application range is limited to the
experimental boundary conditions: Slope (2e4%), run-in
(0.2  103 e 1  103 m3/s) and rock coverage. The infiltration
capacity of the media, Ksat, was found to play a significant role in
each phase. Longitudinal slope showed to have no influence on the
filtering swales in any of the studied phases, i.e. only the inflow rate
and media water content were statistically significant.
The proposed dimensionless empirical models can be used,
along with the rational method, for dimensioning and performance
predictions of filtering swales. As an example, the filtering swales
presented in this study might capture 90% of the runoff generated
by a 12.2 mm/h storm event for on a road with a surface area 40
times larger than the swale. This storm intensity corresponds
approx. with a 5-year (return period), 1-h (duration) rainfall in the
city of Trondheim, which highlights the capacity of these swales to
also handle higher return period events. This will of course be a
function of local rainfall pattern, but it does show the capacity to
handle at the minimum the first flush and significantly more for
many climatic regions.
Acknowledgements
The authors wish to acknowledge the National Public Road
Administration of Norway and the NORWAT program (Project
number 603019) for their financial support for this project. The
authors also want to show their gratitude to Vladimir Hamouz (PhD
student at NTNU) for helping in the construction of the experi-
mental setup.
References
Abu-Zreig, M., Rudra, R.P., Lalonde, M.N., Whiteley, H.R., Kaushik, N.K., 2004.
Experimental investigation of runoff reduction and sediment removal by
vegetated filter strips. Hydrol. Process. 18, 2029e2037.
Ahmed, F., Gulliver, J.S., Nieber, J.L., 2015. Field infiltration measurements in grassed
roadside drainage ditches: spatial and temporal variability. J. Hydrol. 530,
604e611.
Blanco-Canqui, H., Gantzer, C., Anderson, S., Alberts, E., Thompson, A., 2004. Grass
barrier and vegetative filter strip effectiveness in reducing runoff, sediment,
nitrogen, and phosphorus loss. Soil Sci. Soc. Am. J. 68, 1670e1678.
Buckingham, E., 1914. On physically similar systems; illustrations of the use of
dimensional equations. Phys. Rev. 4, 345.
Burack, T., Walls, M., Stewart, H., 2008. New Hampshire Stormwater Manual, vol. 2.
Post-Construction Best Management Practices. Selection and Design.
€ckstro
Ba €m, M., 2002. Sediment transport in grassed swales during simulated runoff
events. Water Sci. Technol. 45, 41e49.
Cizek, A.R., Hunt, W.F., 2013. Defining predevelopment hydrology to mimic pre-
development water quality in stormwater control measures (SCMs). Ecol. Eng.
57, 40e45.
Davis, A.P., Stagge, J.H., Jamil, E., Kim, H., 2012. Hydraulic performance of grass
swales for managing highway runoff. Water Res. 46, 6775e6786.
Deletic, A., 2001. Modelling of water and sediment transport over grassed areas.
J. Hydrol. 248, 168e182.
Deletic, A., 2005. Sediment transport in urban runoff over grassed areas. J. Hydrol.
301, 108e122.
 C. Monrabal-Martinez et al. / Water Research 145 (2018) 509e517
Deletic, A., Fletcher, T.D., 2006a. Performance of grass filters used for stormwater
treatmentda field and modelling study. J. Hydrol. 317, 261e275.
Deletic, A., Fletcher, T.D., 2006b. Performance of grass filters used for stormwater
treatmentda field and modelling study. J. Hydrol. 317, 261e275.
Flanagan, K., Branchu, P., Ramier, D., Gromaire, M.C., 2017. Evaluation of the relative
roles of a vegetative filter strip and a biofiltration swale in a treatment train for
road runoff. Water Sci. Technol. 75 (4), 987e997.
Fort, F., Jouany, C., Cruz, P., 2012. Root and leaf functional trait relations in Poaceae
species: implications of differing resource-acquisition strategies. J. Plant Ecol. 6,
211e219.
García-Serrana, M., Gulliver, J.S., Nieber, J.L., 2017. Infiltration capacity of roadside
filter strips with non-uniform overland flow. J. Hydrol. 545, 451e462.
Gunawardana, C., Egodawatta, P., Goonetilleke, A., 2014. Role of particle size and
composition in metal adsorption by solids deposited on urban road surfaces.
Environ. Pollut. 184, 44e53.
Guo, J.C., Urbonas, B., MacKenzie, K., 2013. Water quality capture volume for storm
water BMP and LID designs. J. Hydrol. Eng. 19, 682e686.
Han, J., Wu, J.S., Allan, C., 2005. Suspended sediment removal by vegetative filter
strip treating highway runoff. J. Environ. Sci. Health. A, Toxic/Hazardous Sub-
stances & Environ. Eng. 40, 1637e1649.
Hoffman, E.J., Latimer, J.S., Hunt, C.D., Mills, G.L., Quinn, J.G., 1985. Stormwater runoff
from highways. Water, Air, & Soil Pollution 25, 349e364.
Huber, M., Badenberg, S.C., Wulff, M., Drewes, J.E., Helmreich, B., 2016a. Evaluation
of factors influencing lab-scale studies to determine heavy metal removal by six
sorbents for stormwater treatment. Water 8 (2), 62.
Huber, M., Welker, A., Dierschke, M., Drewes, J.E., Helmreich, B., 2016b. A novel test
method to determine the filter material service life of decentralized systems
treating runoff from traffic areas. J. Environ. Manag. 179, 66e75.
Huber, M., Welker, A., Helmreich, B., 2016c. Critical review of heavy metal pollution
of traffic area runoff: occurrence, influencing factors, and partitioning. Sci. Total
Environ. 541, 895e919.
Ilyas, A., Muthanna, T.M., 2017. Assessment of upscaling potential of alternative
adsorbent materials for highway stormwater treatment in cold climates. Envi-
ron. Technol. 38 (6), 705e717.
Kalmykova, Y., Stro €mwall, A.M., Steenari, B.M., 2008. Alternative materials for
adsorption of heavy metals and petroleum hydrocarbons from contaminated
leachates. Environ. Technol. 29 (1), 111e122.
Ke, L., Takahashi, A., 2012. Strength reduction of cohesionless soil due to internal
erosion induced by one-dimensional upward seepage flow. Soils Found. 52,
698e711.
Khan, U.T., Valeo, C., Chu, A., van Duin, B., 2012. Bioretention cell efficacy in cold
climates: Part 1 d hydrologic performance. Can. J. Civ. Eng. 39, 1210e1221.
Kuichling, E., 1889. The relation between the rainfall and the discharge of sewers in
populous districts. Trans. Am. Soc. Civ. Eng. 20, 1e56.
Le Coustumer, S., Fletcher, T.D., Deletic, A., Barraud, S., Lewis, J.F., 2009. Hydraulic
performance of biofilter systems for stormwater management: influences of
design and operation. J. Hydrol. 376, 16e23.
Le Coustumer, S., Fletcher, T.D., Deletic, A., Barraud, S., Poelsma, P., 2012. The in-
fluence of design parameters on clogging of stormwater biofilters: a large-scale
column study. Water Res. 46, 6743e6752.
LeFevre, G., Paus, K., Natarajan, P., Gulliver, J., Novak, P., Hozalski, R., 2014. Review of
dissolved pollutants in urban storm water and their removal and fate in bio-
retention cells. J. Environ. Eng. 141 (23), 04014050.
Li, H., Davis, A.P., 2008. Urban particle capture in bioretention media. I: laboratory
and field studies. J. Environ. Eng. 134 (6), 409e418.
Lucke, T., Mohamed, M.A.K., Tindale, N., 2014. Pollutant removal and hydraulic
reduction performance of field grassed swales during runoff simulation
517
experiments. Water 6, 1887e1904.
Mishra, P.K., Neelkanth, J.K., Maheswara Babu, B., Kumathe, S.S., 2006. Effectiveness
of Bermuda grass as vegetative cover in grassed waterway: a simulated study.
J. Irrigat. Drain. Eng. 132, 288e292.
Moghadas, S., Paus, K., Muthanna, T.M., Herrmann, I., Marsalek, J., Viklander, M.,
2015. Accumulation of traffic-related trace metals in urban winter-long road-
side snowbanks. Water, air. & Soil Pollution 226, 404.
Monrabal-Martinez, C., Ilyas, A., Muthanna, M.T., 2017. Pilot scale testing of adsor-
bent amended filters under high hydraulic loads for highway runoff in cold
climates. Water 9, 230.
Morbidelli, R., Saltalippi, C., Flammini, A., Govindaraju, R.S., February 2018. Role of
slope on infiltration: a review. J. Hydrol. 557, 878e886.
Mun~ oz-Carpena, R., Parsons, J.E., Gilliam, J.W., 1993. Numerical approach to the
overland flow process in vegetative filter strips. Transactions of the ASAE 36,
761e770.
Mun~ oz-Carpena, R., Parsons, J.E., Gilliam, J.W., 1999. Modeling hydrology and
sediment transport in vegetative filter strips. J. Hydrol. 214, 111e129.
Muthanna, T.M., Viklander, M., Blecken, G., Thorolfsson, S.T., 2007. Snowmelt
pollutant removal in bioretention areas. Water Res. 41, 4061e4072.
Narsimlu, B., Rao, B.V., Mishra, P.K., Rao, K.V., 2004. Effect of covers for soil and
water conservation using tilting flume. J. Irrigat. Drain. Eng. 130, 154e159.
Nassif, S.H., Wilson, E.M., 1975. The influence of slope and rain intensity on runoff
and infiltration L'influence de l'inclinaison de terrain et de l'intensite  de pluie
sur l'ecoulement et l'infiltration. Hydrol. Sci. J. 20 (4), 539e553.
Paus, K., Braskerud, B., 2014. Suggestions for designing and constructing bio-
retention cells for a nordic climate. Journal of Water Management and Research
70, 139e150.
Paus, K., Morgan, J., Gulliver, J.S., Leiknes, T., Hozalski, R.M., 2013. Assessment of the
hydraulic and toxic metal removal capacities of bioretention cells after 2 to 8
Years of service. Water, Air, Soil Pollut. 225, 1803-1803.
Poesen, J., 1984. The influence of slope angle on infiltration rate and Hortonian
overland flow. Zeitschrift für Geomorpholgie, Supplement Band 49, 117e131.
Samani, J.M., Kouwen, N., 2002. Stability and erosion in grassed channels. J. Hydraul.
Eng. 128 (1), 40e45.
Sansalone, J.J., Buchberger, S.G., 1997. Partitioning and first flush of metals in urban
roadway storm water. J. Environ. Eng. 123, 134e143.
Schueler, T.R., Claytor, R.A., 1996. Design of Stormwater Filter Systems. Center for
Watershed Protection, Silver Spring, MD.
Urbonas, B., Guo, J., Tucker, L.S., 1989. Sizing capture volume for stormwater quality
enhancement. Flood Hazard News 19, 1e9.
USDA, U. S. Department of Agriculture, 2017. Natural Resources Conservation Ser-
vice Soil. Internet, url: https://www.nrcs.usda.gov/wps/portal/nrcs/detail/soils/
survey/?cid¼nrcs142p2_054167 (accessed February 14, 2017).
Vegvesen, S., 2008. Veg-og gateutforming. In: Håndbok N100. Vegdirektoratet.
Winston, R.J., Anderson, A.R., Hunt, W.F., 2016. Modeling sediment reduction in
grass swales and vegetated filter strips using particle settling theory. J. Environ.
Eng. 143 (1), 04016075.
Winston, R.J., Powell, J.T., Hunt, W.F., 2018. Retrofitting a grass swale with rock
check dams: hydrologic impacts. Urban Water J. 1e8.
Wium-Andersen, T., Nielsen, A.H., Hvitved-Jacobsen, T., Kristensen, N.K., Brix, H.,
Arias, C., Vollertsen, J., 2012. Sorption media for stormwater treatmentda
laboratory evaluation of five low-cost media for their ability to remove metals
and phosphorus from artificial stormwater. Water Environ. Res. 84 (7),
605e616.
Yousef, Y.A., Hvitved-Jacobsen, T., Wanielista, M.P., Harper, H.H., 1987. Removal of
contaminants in highway runoff flowing through swales. Sci. Total Environ. 59,
391e399.