See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/334383554

A Quantity–Quality Model to Assess the Effects of Source Control Stormwater

Management on Hydrology and Water Quality at the Catchment Scale

Article in Water · July 2019

DOI: 10.3390/w11071415

CITATIONS READS

45 1,521

6 authors, including:

Abdulrazaq Rezaei Zubaidah Ismail

University of Malaya University of Malaya
5 PUBLICATIONS 85 CITATIONS 73 PUBLICATIONS 1,360 CITATIONS

SEE PROFILE SEE PROFILE

Mohammad Hossein Niksokhan Sharif Moniruzzaman Shirazi

University of Tehran Eastern University
107 PUBLICATIONS 992 CITATIONS 65 PUBLICATIONS 1,767 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Health Monitoring of Bridges and Structures View project

Development of Environmental Flow Framework for Flora and Fauna at Turag River Basin in Dhaka View project

All content following this page was uploaded by Abdulrazaq Rezaei on 04 April 2021.

The user has requested enhancement of the downloaded file.

 water
Article
A Quantity–Quality Model to Assess the Effects of
Source Control Stormwater Management on
Hydrology and Water Quality at the Catchment Scale
Abdul Razaq Rezaei 1, *, Zubaidah Ismail 2, *, Mohammad Hossein Niksokhan 3 ,
Muhammad Amin Dayarian 3 , Abu Hanipah Ramli 4 and Sharif Moniruzzaman Shirazi 5
1 Water Resources Engineering, Civil Engineering, Institute for Advanced Studies, University of Malaya,
Kuala Lumpur 50603, Malaysia
2 Civil Engineering Department, Faculty of Engineering, University of Malaya, Kuala Lumpur 50603, Malaysia
3 School of Environment, College of Engineering, University of Tehran, Tehran 141785311, Iran
4 Department of Irrigation and Drainage (DID), Ministry of Water, Land and Natural Resources Faculty (University),
Kuala Lumpur 50626, Malaysia
5 Civil Engineering Department, World University of Bangladesh, Dhaka 1205, Bangladesh
* Correspondence: [email protected] (A.R.R.); [email protected] (Z.I.); Tel.: +60-379-675-284
(A.R.R.); +60-379-675-284 (Z.I.); Fax: +60-379-675-318 (Z.I.)




Received: 18 April 2019; Accepted: 30 May 2019; Published: 10 July 2019


Abstract: The vast development of urban areas has resulted in the increase of stormwater peak runoff
and volume. Water quality has also been adversely affected. The best management practices (BMPs)
and low impact development (LID) techniques could be applied to urban areas to mitigate these
effects. A quantity–quality model was developed to simulate LID practices at the catchment scale
using the US Environmental Protection Agency Storm Water Management Model (US EPA SWMM).
The purpose of the study was to investigate the impacts of LID techniques on hydrology and water
quality. The study was performed in BUNUS catchment in Kuala Lumpur, Malaysia. This study
applied vegetated swale and rain garden to assess the model performance at a catchment scale using
real field data. The selected LIDs occupied 7% of each subcatchment (of which 40% was swale and 30%
was rain garden). The LID removal efficiency was up to 40% and 62% for TN and TSS, respectively.
The peak runoff reduction was up to 27% for the rainfall of up to 70 mm, and up to 19% for the rainfall
of between 70 and 90 mm, respectively. For the longer storm events of higher than 90 mm the results
were not as satisfactory as expected. The model was more effective in peak runoff reduction during
the shorter rainfall events. As for the water quality, it was satisfactory in all selected rainfall scenarios.

Keywords: urbanization; runoff; water quality; best management practices; low impact development

1. Introduction
Urbanization is increasing vastly all over the world [1,2]. This massive increase in urbanization
coupled with the increasing rate of climate change are the two main factors contributing to
urban stormwater runoff that could not be handled properly by applying conventional stormwater
management [3,4]. Many urban areas are undergoing rapid development around the world and the
urbanization process is gaining more interest. This has profoundly modified the natural environment
in urban areas [2].
Urbanization results in the modification of natural landscapes, with pervious vegetated surfaces
being replaced with impervious surfaces [5]. Human activities are the main reasons for this modification
of natural surfaces and the growth of impervious surfaces in urban areas. Construction of structures
such as roofs, parking lots, and roads are some examples of impervious surfaces in urban areas.

Water 2019, 11, 1415; doi:10.3390/w11071415 www.mdpi.com/journal/water

Water 2019, 11, 1415 2 of 23

The growth of impervious surfaces decreases infiltration capacity, increases runoff generation and
direct runoff, improves the connectivity of flow and leads to the reduction of groundwater recharge [5].
These changes in natural surfaces will ultimately result in the modification of magnitude and duration
of urban catchment floods [6]. Some other factors, such as climate change and population growth will
also intensify the consequences of urbanization [7]. The increased imperviousness and the resultant
contaminated stormwater will also deteriorate the water quality in urban areas [8–10]. Different types
of pollutants could be found in stormwater, such as sediments, nutrients, heavy metals, trace elements,
and pathogens [11–13]. These pollutants will be built up on impervious surfaces during dry period
and will be washed off by surface runoff during rainfall events to the waterbodies [14].
The impacts of urbanization on runoff process also depend on the extent of urban catchment
development. The small-sized and highly urbanized river basins suffer more from urban runoff rather
than large river watersheds. In the large-sized river catchments, the runoff peaks constitute only
a small portion of flow [15].
Urbanization can also have adverse impact on sewage systems, as urban development increases
the volume of runoff discharged to sewerage networks. This might lead to hydraulic overloading or
sewage backflows [16].
With regard to this, urban stormwater management should be an integral part of any urban
development. It has significant effects on both the economy and the ecology of the society [4].
The conventional approach for stormwater management use gutters and a system of sewers and
canals to convey the stormwater out of the city as fast as possible. This traditional approach could not
contribute to the sustainable urban development [3,17].
There are different types of Best Management Practices (BMPs) used to manage urban stormwater,
namely constructed wetlands, ponds, buffer strips, and bioretention systems, which are the most
popular types [18–20].
Some newer types of BMPs have been developed to control the urban stormwater at the source
such as vegetated swales, rain gardens (or bioretention systems), permeable pavements, and green
roofs. These newer techniques are generally known as Low Impact Development (LID) [21]. LID
refers to distributed small-scale treatment systems located at the sources of runoff generation, such as
roofs and streets [22,23]. The LID techniques are capable of mitigating the impact of imperviousness
on both hydrology and water quality of urban stormwater runoff. Particularly, the LID techniques
have been developed to mimic the predevelopment hydrologic conditions and promote the storage,
infiltration, and evapotranspiration processes [24]. In other words, the key principle of LID measures
is to ensure that new urban developments do not make the existing hydrologic regime flashier and
increase flooding in the catchment. The capabilities of BMPs and LID in urban runoff reduction and
water quality improvement have been well documented in numerous studies [25–31]. In comparison
with the traditional stormwater management approaches, LID techniques have the advantage of
returning the natural hydrological cycle in urban areas if the development is on a greenfield site,
such as runoff volume reduction [25] infiltration improvement [24], peak flow reduction [32], extending
the lag time, pollutant loads reduction [33], and baseflow increase [4].
Studies conducted to assess the performance of LID on runoff control, reported that LID perform
significantly differently in storms with different intensities. Hood et al. [34] studied the runoff volume,
peak discharge, and runoff coefficients of low impact residential development and compared them with
traditional approaches. They found that the effects of LID on runoff reduction were more significant
for smaller storms with shorter durations. Damodaram, Giacomoni, Prakash Khedun, Holmes, Ryan,
Saour, and Zechman [27] used a hydrologic model to evaluate the impacts of LID on the stream flow.
Their study showed that LID is able to control stormwater for smaller storms, while it is not as effective
as conventional detention ponds in case of flooding events. Thus, although LID practices are quite
essential to control urban stormwater runoff, the conventional urban drainage systems could not be
completely ignored.
Water 2019, 11, 1415 3 of 23

The main purpose of this study was to assess the effects of LID techniques on hydrology and
water quality at a catchment scale using EPA SWMM and collected field data. In this study, vegetated
swale and rain garden, among all types of LIDs, were selected to be applied to impervious areas to
reduce the effect of imperviousness on hydrology and water quality at the catchment scale. These two
LIDs have the advantage of controlling the stormwater quantity and quality at the source, and are
able to reduce the flow volume, and thus delay the hydrologic response and reduce the pollutant load
washed-off from urban surfaces [4].
Rain garden (bioretention) is relatively highly efficient for both runoff and pollutant reduction [35].
For example, bioretention cells reduced the average peak flows by at least 45% in Maryland and
North Carolina during a series of rainfall events [36]. It has also been proved that bioretention are
capable of reducing sediment and nutrient from 0% to 99% [35]. Swales have also been shown to have
an average retention of 14% to 98% for nutrients and TSS, and up to 93 % for metals [24].
Previous studies also investigated the impact of other types of LID controls on urban runoff,
flooding, and water quality. Jackisch and Weiler [37] carried out a study on the hydrological performance
of LIDs. They used some LIDs, such as permeable pavement, vegetated roofs, and bioretention.
They found out that LIDs were able to capture 73% of rainfalls and the runoff volume was also reduced
by 66–87%. The results confirmed that LID could replace the conventional stormwater management
systems. Wilson, et al. [38] compared the water quality performance of commercial LID with the
conventional development. The results showed that LID control water quality performance was
much better than the conventional systems. There are many more similar researches investigating the
hydrological or water quality performance of LIDs [39–44].
Based on this, the main aim of this study was to investigate the impacts of LID practices to
manage stormwater runoff and water quality in urban areas by developing a quantity–quality model
at the catchment scale. The first specific objective was to assess the impact of the two types of
LIDs: vegetated swale and rain garden on the hydrologic response and water quality of an urban
catchment. This objective aims to assess the performance of the selected LIDs to reduce peak runoff
and pollutants in the case of stormwater. The purpose was to investigate how much the peak runoff
and selected pollutants would be reduced by applying vegetated swale and rain garden to the
catchment. The second objective of the study was to evaluate the hydrologic response and water quality
improvement of the selected LIDs in different rainfall scenarios with different intensities. The purpose
was to investigate the hydrological and water quality performance of the selected LIDs in higher
intensities and longer duration rainfall events. In other words, the specific aim of the study was to
investigate the efficiency of the selected LIDs in reducing the stormwater runoff and its pollutants
in various rainfall intensities and durations. For this purpose, nine different rainfall scenarios were
achieved using the intensity–duration–frequency (IDF) curves, which were already developed for
the study area. Three different return periods, T, namely 5, 10, and 20 years, with three different
durations, D, namely 1, 1.5, and 2 hours, were selected accordingly, nine rainfalls with different
intensities were achieved from IDF curves to be simulated in the developed model. The developed
model was tested with these nine rainfalls to evaluate the model performance with LID in peak runoff
and pollutant reduction. Total suspended sediments (TSS) and total nitrogen (TN) were selected as
available pollutants to evaluate the model performance in pollutants reduction. To achieve the best
results, samples were collected from a real rainfall event of the catchment on the 20 September 2018.
The samples were analyzed in the environmental laboratory to achieve the selected pollutants, namely
TSS and TN. The rainfall and amount and the respective flow were also collected from the automated
hydrological station at the site. After that the model was carefully calibrated and validated for both
quantity and quality using the 20 September 2018 data. Historical rainfall-flow data were also collected
from the on-site automated rain gauge for model validation. Other required data were collected via
site inspection and GIS maps. A sensitivity analysis and goodness-of-fit test were also carried out
to find out the most sensitive parameters of the model and to check the reliability of the developed
model. The significance of the study was to support the idea that LID application in urban areas
rainfall and amount and the respective flow were also collected from the automated hydrological station
at the site. After that the model was carefully calibrated and validated for both quantity and quality using
the 20 September 2018 data. Historical rainfall-flow data were also collected from the on-site automated
rain gauge for model validation. Other required data were collected via site inspection and GIS maps. A
Water 2019, 11, 1415 4 of 23
sensitivity analysis and goodness-of-fit test were also carried out to find out the most sensitive parameters
of the model and to check the reliability of the developed model. The significance of the study was to
support the idea that
can remarkably LID peak
reduce application in enhance
run off, urban areas canquality,
water remarkably reduce
and can alsopeak run the
replace off, conventional
enhance water
quality, and can
stormwater also replacesystems.
management the conventional stormwater management systems.

2.2.Materials
Materialsand Methods
and Methods

2.1.
2.1. Study
Study Area
Area
Thestudy
The studyhashasbeen
beencarried
carriedout outinin the
the Bunus River Subcatchment
Subcatchment located
located in
in Kuala
Kuala Lumpur
LumpurCity,
City,
Malaysia.The
Malaysia. TheBunus
BunusSubcatchment
Subcatchmentisispart part of
of the bigger
bigger catchment:
catchment: Klang
KlangRiver
RiverCatchment.
Catchment.TheTheBunus
Bunus
catchment
catchmentarea
areaisis~18
~18kmkm2 2ininterms
termsofoftotal
total size,
size, with
with the main river
river stretching
stretching9.5
9.5km
kmoriginating
originatingfrom
from
Wangsa
WangsaMaju
Maju(3.212° ◦
(3.212 NNand 101.735◦E)
and101.735° E) and
and joining Klang river
river next
nexttotoJalan
JalanMunshi
MunshiAbdullah
Abdullah(3.153 ◦ NN
(3.153°
and
and101.698° ◦
101.698E).E).The
Thecatchment
catchment hashas
twotwohydrological
hydrological monitoring
monitoring points (S1 (S1
points andand
S2 in
S2 Figure 1) and
in Figure one
1) and
automated streamflow gauge (S2) run by the SMART (Stormwater Management
one automated streamflow gauge (S2) run by the SMART (Stormwater Management and Road Tunnel) and Road Tunnel) control
center which
control is under
center whichthe Malaysian
is under Department
the Malaysian of Irrigation
Department ofand Drainage
Irrigation and(DID).
Drainage (DID).

Figure 1.
Figure 1. Study area location
location within
within Kuala
KualaLumpur,
Lumpur,Malaysia.
Malaysia.

The city of Kuala Lumpur has undergone a remarkable development in the past few decades
which has widely resulted in the vast urbanization and urban runoff. As the Bunus River subcatchment
is one of the most densely populated areas within Kuala Lumpur, the urban runoff quantity and quality
is a major issue in the area. Figure 1 shows the Bunus River subcatchment within Kuala Lumpur city
in Malaysia. The arrows depict the direction of drainage in the catchment which is from the upper
right to the lower left.

2.2. SWMM Model

The US EPA Storm Water Management Model (SWMM) [45] (has been selected to evaluate the
effects of LID on runoff reduction and water quality improvement in the study area. A dynamic
Water 2019, 11, 1415 5 of 23

rainfall–runoff module and a hydraulic module are included in SWMM for piped systems. SWMM
is used to simulate runoff quantity and quality mostly from urban areas. An LID control module
has been provided to the model from version (5.1.010), which can precisely model the hydrologic
performance of LID controls in urban areas. In SWMM, LID controls are depicted by a combination of
vertical layers and their properties (such as thickness, void volume, hydraulic conductivity, underdrain
characteristics, etc.) are defined on a per unit area basis. The defined LID controls can be assigned
within the selected subcatchments with different designed sizes (or areal coverage). The SWMM model
has been successfully used in numerous studies to investigate the impacts of stormwater management
based on traditional drainage systems [46] or LID designs [22]. The SWMM model has been used
in many studies related to urban stormwater, and the capabilities of this model have been well proved
in all these studies [17,19,21,43,47–50].
SWMM is a deterministic and spatially distributed hydrological model, which is able to simulate
the hydrological cycle mainly within urban areas [45,51]. SWMM uses the continuity equation along
with the Manning’s equation to simulate hydrological outflows [51].
To apply SWMM in urban areas, hourly, or subhourly, rainfall data, topographic slope and
elevation, soil, land use data, sewer system network map, and storm sewer discharge data are required
for calibration and validation of the model [52]. Input parameters required by the SWMM model to
simulate stormwater runoff include rainfall and climatology data (for continuous modeling), parameters
for hydrologic components (subcatchments, pipes, storage units, etc.), and run time controls (time step,
starting and ending time, etc.).

2.3. Input Data

The historical rainfall and the relevant flow data were collected from the SMART control center
which is part of the DID. There are two hydrological monitoring points within the Bunus catchment
located at Jalan (Street) Genting Klang (S1) and Jalan Tun Razak (S2) area. The S2 is an automated
streamflow gauge as well. All rainfall and flow data used for quantity and quality modeling were
collected from S2 automated hydrological monitoring station located at Jalan Tun Razak. The DID
office also provided us with relevant GIS maps of the Klang catchment and Bunus subcatchment
consisted of the main Klang River and the tributaries.
For the purpose of calibration and validation of the model, we collected runoff samples during
one rainfall event at Jalan Tun Razaq Hydrological station (S2) on 20 September 2018. Later, the rainfall
data of the same event was collected from the SMART control center to calibrate the model. A total of
17 samples were collected from the rainfall event and were analyzed in the Environmental Laboratory at
the University of Malaya for the quality parameters. We used the water quality parameters to calibrate
the quality model. Among the analyzed pollutants from the collected samples, total suspended
sediments (TSS) and total nitrogen (TN) were selected to be used for the stormwater quality modeling.
TSS and TN are among the most widespread Pollutants in urban areas [53].

2.4. Modeling Procedure

To set up the SWMM model, different types of data are required. The data required to set-up
the model consist of rainfall and flow data, topographic data of the study area, river characteristics,
catchment drainage networks, catchment slope, land use of the study area, and soil type. The rainfall
and flow data were provided by the SMART control center, while GIS maps were provided by DID.
Topographic data, such as elevation and slopes were delineated from GIS maps. The catchment
drainage networks were also provided by DID. The land use and soil type in the study area were
achieved through GIS maps, site inspection and Google Earth. The study area is a heavily populated
residential part of Kuala Lumpur consisted of 70% residential and 30% commercial, which consist of
mostly food outlets and shops. The soil type in the area is mostly clay loam. The river characteristics
were achieved through site inspection. The most important characteristics of the river used for
modeling are river width, depth, and shape. The river length was also delineated from the GIS maps.
Water 2019, 11, 1415 6 of 23

The catchment map was imported into the SWMM model and the subcatchments were drawn
following the contour lines. The system of nodes and conduits were also added to the model, likewise.
Figure 2 illustrates the drainage system in SWMM in the base case with all subcatchments, nodes
and conduits network. Totally, we came up with 35 subcatchments and 10 junctions in the model.
Some of the principal input parameters for each subcatchment in SWMM include assigned rain gage,
outlet node for each subcatchment, assigned land uses, tributary surface area, imperviousness, slope,
characteristic width of overland flow, Manning’s n for overland flow on both pervious and impervious
areas, depression storage in both pervious and impervious areas, and percent of impervious area with
no depression storage. Some of these parameters are assigned by the user, like rain gage, and some
of them were assigned by site inspection and using ArcGIS and google earth, such as land uses and
imperviousness, and some of them were calculated mathematically, such as slope and width. SWMM
offers three main choices for modeling infiltration, namely Horton, Green-Ampt, and Curve Number
methods. In the hydrologic module of SWMM, the Green-Ampt model was employed in the infiltration
model, which calculates the amount of rainfalls infiltrated into the unsaturated upper soil zone of
a pervious surface area, whereas Manning’s equation was used to compute the surface runoff [54].
Based on Green-Ampt and on-site soil type, the pertinent parameters to the infiltration model were
adjusted in SWMM. The main parameters are hydraulic conductivity, suction head, and initial deficit.
Having set-up
Water 2019, 11, x the
FOR required parameters, the model was ready to operate by employing
PEER REVIEW 7 ofthe
25 rainfall
data series.

Figure
Figure2. Drainage system
2. Drainage in the base
system casebase
in the in thecase
Storm
in Water Management
the Storm Water Model (SWMM)Model
Management with subcatchments.
(SWMM) with
subcatchments.
2.5. Sensitivity Analysis
2.5. Sensitivity
DifferentAnalysis
parameters, such as Manning’s roughness coefficient, depression storage depth, infiltration
parameters
Different (e.g., curve number),
parameters, suchsubcatchments
as Manning’s percent imperviousness,
roughness subcatchment
coefficient, slope,storage
depression percentage
depth,
of impervious surfaces with no depression storage, and channel roughness values have been used in
infiltration parameters (e.g., curve number), subcatchments percent imperviousness, subcatchment
previous studies as calibration parameters [39,48,55–57]. Before conducting the model calibration in this
slope, percentage
study, of impervious
the sensitivity of the modelsurfaces with
to different no depression
parameters storage,
was analyzed and channel
to figure out whatroughness
parameters values
are
more sensitive to the model. In order to achieve the best match between the observed and modeled flow,
these parameters were adjusted in the model. To carry out sensitivity analysis, the parameters were
checked one by one. For each parameter to be analyzed, different amounts of the parameter were used in
the model, while keeping all other parameters fixed. The model was run and the results of the simulated
outflow hydrograph was compared with the observed one each time. The process was repeated several
times to figure out the most sensitive parameters to the model.
Water 2019, 11, 1415 7 of 23

have been used in previous studies as calibration parameters [39,48,55–57]. Before conducting the
model calibration in this study, the sensitivity of the model to different parameters was analyzed
to figure out what parameters are more sensitive to the model. In order to achieve the best match
between the observed and modeled flow, these parameters were adjusted in the model. To carry out
sensitivity analysis, the parameters were checked one by one. For each parameter to be analyzed,
different amounts of the parameter were used in the model, while keeping all other parameters fixed.
The model was run and the results of the simulated outflow hydrograph was compared with the
observed one each time. The process was repeated several times to figure out the most sensitive
parameters to the model.

2.6. Calibration and Validation

In the model calibration and validation approach, the predicted (modeled) and the observed
outflow hydrographs were compared with each other. Particularly, two values, namely the discharge
volume and the peak flow rate were assessed for an event. The calibration was conducted using
the data of 20 September 2018 event. In order to calibrate the model more accurately, we collected
samples of a real rainfall event on 20 September 2018 from the monitoring station at Jalan Tun Razak.
The rainfall and the respective flow were also collected from the automated hydrological station run
by the SMART control center. The collected data was used for both quantity and quality calibration
of the model. As some of the historical data might not be accurate, collecting data from a real event
was very important to make sure the calibration is precise and accurate. The historical rainfall and
flow data were also collected from SMART control center to be used for model validation. For the
validation purpose, the data on 1 August 2017 was used. For a good validation result, it is required
that the general shape of both hydrographs follow each other, and the results of goodness-of-fit test
are satisfactory. With regard to this, to evaluate the model calibration and validation performance,
the goodness-of-fit test was performed to compare the modeled discharge volume against the observed
discharge volume for a specific analysis period [43].
Modeling of stormwater quantity usually aims for the prediction of stormwater runoff volumes,
hydrographs and peak discharge rates to mitigate the runoff peak and volume in urban areas. Modeling
of urban stormwater quality normally focuses on the pollution prevention and impact assessment
resulting from stormwater.
Several mechanisms determine stormwater quality, with build-up and wash-off being the most
important ones. It is generally assumed that a supply of contaminants is built up on the land surface
during dry weather in an impervious urban area. The buildup of pollutants may or may not be related
to time, traffic flow, dry fallout and street sweeping [14]. When the storm occurs, the built-up pollutants
are washed off into the drainage system. Wash-off is the process of erosion or solution of pollutants
from an impervious surface during the period of runoff in urban areas.
The calibration of stormwater quality models is normally more difficult than quantity models.
Modeling is used to predict, analyze and manage urban water quality and pollution, as urban runoff
quality monitoring needs substantial resources [58]. In SWMM, the stormwater quality is modeled using
different build-up and wash-off equations, such as exponential function, power function, or saturation
equation [59]. There is usually a wide range of variation in urban stormwater quality data between
rainfall events. The buildup and wash-off equations consist of many parameters which make the
stormwater quality difficult to calibrate without on-site data [23]. The spatial and temporal variation
of the quality modeling process also make it difficult to use the buildup and wash-off equations on
a catchment scale [60,61]. Moreover, the urban stormwater quantity model is usually calibrated first,
followed by calibrating the quality module [62].
For the quality calibration of the model, total suspended sediments (TSS) and (NOx+NH3)
(referred to as TN) were selected. We used nitrate, nitrite (NOx), and ammonia-N (NH3) as the total
nitrogen. Dissolved inorganic nitrogen (DIN) includes ammonia (NH3), nitrite (NO2), and nitrate
(NO3). The waterbodies are greatly affected by these constituents because they are easily available to be
Water 2019, 11, 1415 8 of 23

taken up by simple organisms [63], and may result in eutrophication, hypoxia, and loss of biodiversity
and habitat [64]. As TSS and TN are widespread pollutants in residential areas [53], they were selected
for the model quality calibration.

2.7. Goodness-of-Fit Test

Different evaluation methods are often used for measuring the accuracy of the predicted stormwater
runoff against the observed values. Three main methods selected to be used for the current research
model evaluation are root mean square error (RMSE), Nash–Sutcliffe coefficient (NSC), and the
regression method. The Normalized Objective Function (NOF) can be used for the root mean square
error. The RMSE and NOF are explained as follows
v
u
N
t
1 X
RMSE = (Pi − Oi )2 (1)
N
i=1

RMSE
NOF = (2)
O
where Pi and Oi are the predicted and observed values at time step i, respectively, and N is the number
of observations during the flow period. O is the mean of observed values. The ideal value for NOF is 0,
however, it cannot be expected to occur; otherwise it would be a perfect model. So, values between 0.0
and 1.0 are acceptable for NOF when field specific data are available for calibration [65].
Nash–Sutcliffe coefficient (NSC) is explained as follows [66]

(Oi − Pi )2
Pn
NSC = 1 − Pi=1 2
(3)
n
i=1 (Oi − O)

where Pi is the predicted value, and Oi is the observed value for the n observations, and O is the mean
of observed values. The optimal condition for the model occurs when the NSC is 1.
Regression method:
In the regression method, the linear regression line is fitted between the modeled and observed
values. Generally, the best calibration requires that r2 be as close to 1.0 as possible [67].

2.8. Assigning LID Controls to the Model

Having calibrated and validated the model, the selected LIDs were assigned into the model.
The LIDs are assigned to each subcatchment through the LID control editor of the subcatchment.
The LID Control editor is used to define a low impact development control in each subcatchment
which can be placed throughout a catchment to store, infiltrate, and evaporate subcatchment runoff.
The design of the LID control is made on a per unit area basis so that it can be placed in any number of
subcatchments at different sizes or number of replicates.
The common types of LID being defined in SWMM include bioretention cell, rain garden, green
roof, infiltration trench, permeable pavement, vegetative swale, and rain barrel. Vegetated Swale and
Rain Garden have been the selected LIDs to be assigned in the final model. According to the literature,
the two LIDs have a high capacity of runoff reduction and pollutants removal in the urban areas.
The rain garden and vegetated swale are able to reduce TSS and TN in urban runoff up to 89% and 58%,
respectively [68]. In this study, the selected LIDs occupy 7% of each subcatchment. Swale and Rain
Garden treats 40% and 30% of the impervious area of each subcatchment, respectively. The final model
was tested by the calibrated rainfall as well as the TSS and TN to check the model efficiency. after that
the model was executed using the selected rainfall scenarios to check efficiency of the developed model
for different rainfall intensities.
Water 2019, 11, 1415 9 of 23

2.9. Selecting Rainfall Scenarios

In order to test the final model with different rainfall scenarios, nine sets of rainfall were selected
as follow. The intensity–duration–frequency (IDF) curves for the area had already been prepared by
DID and published in the Urban Stormwater Management Manual for Malaysia [69]. These curves
have been developed based on 100-year hydrological data for the return periods of 2, 5, 10, 20, 50,
and 100 years. They have been particularly developed to be used by researchers working on the
respective study areas. In IDF curves, the rainfall durations are depicted on the “X” axis and the
intensities are
Water 2019, 11, depicted
x FOR PEERon the “Y” axis. The rainfall frequencies are illustrated in the form
REVIEW 10 ofof
25 diagonal
curves. The return periods of 5 years, 10 years, and 20 years were selected for the three selected
durations, namely 1, 1.5, and 2 hours. For instance, to achieve the first rainfall intensity, the 1-hour
to achieve
duration the first
is found onrainfall
the “X”intensity,
axis andthethe
1-hour duration isof
intersection found on the “X”
the vertical axis
line andthe
with the return
intersection of of
period
the vertical line with the return period of 5 is found. Then the intersection of the horizontal line with the
5 is found. Then the intersection of the horizontal line with the “Y” axis of this point will be the
“Y” axis of this point will be the respective rainfall intensity of that return period and duration. Likewise,
respective rainfall intensity of that return period and duration. Likewise, the process was repeated
the process was repeated for other durations and with other return periods. Therefore, for each return
for other durations and with
period, three durations were other
appliedreturn
to findperiods. Therefore,
the corresponding for each
rainfalls. return
Thus, period,
total nine threerainfalls
different durations
werewere
applied to find
achieved. the corresponding
The achieved rainfalls.in Thus,
rainfalls are depicted Table 1total nine different
and Figure 3. rainfalls were achieved.
The achieved rainfalls are depicted in Table 1 and Figure 3.
Table 1. Rainfall scenarios for three return periods and three durations.
Table 1. Rainfall scenarios for three return periods and three durations.
Return Period Duration (hour)
Duration (hour) Rainfall
Rainfall(mm)
(mm)
Return Period
5 1 72
5 5 1
1.5 72
81
5 5 1.5
2 8190
10 5 12 9080
1010 1
1.5 8090
1010 1.5
2 9096
2010 12 9690
2020 1
1.5 9099
2020 1.5
2 99110
20 2 110

Figure
Figure 3. Rainfall
3. Rainfall scenariosfor
scenarios forthree
three return
return periods,
periods,namely
namely5, 10, andand
5, 10, 20 years.
20 years.

3. Results

3.1. Sensitivity Analysis

Water 2019, 11, 1415 10 of 23

3. Results

3.1. Sensitivity Analysis

Before
Water 2019,conducting
11, x FOR PEERcalibration
REVIEW and validation, the sensitivity analysis was performed. Figure 11 of 25 4 depicts
the results of sensitivity analysis. As can be observed, the most sensitive parameters for the model calibration
are % imperviousness (Percent of the land area which is impervious), % zero imperviousness (Percent of
Before conducting calibration and validation, the sensitivity analysis was performed. Figure 4 depicts
the impervious area with no depression storage), Dstore-impervious (Depth of depression storage on the
the results of sensitivity analysis. As can be observed, the most sensitive parameters for the model
impervious portion of the subcatchment), and Dstore-pervious (Depth of depression storage on the pervious
calibration are % imperviousness (Percent of the land area which is impervious), % zero imperviousness
portion of the
(Percent ofsubcatchment)
the imperviouswhich haveno
area with been shortened
depression as D-Store
storage), impervious-pervious.
Dstore-impervious (Depth ofThis means that
depression
the model
storageisonmostly sensitive portion
the impervious to theseofthree parameters and
the subcatchment), they
and need to be adjusted
Dstore-pervious (Depthaccordingly
of depression tostorage
get a good
calibration. However,
on the pervious care of
portion should be taken thatwhich
the subcatchment) the sensitivity
have beenofshortened
the model as to the three
D-Store parameters is not the
impervious-pervious.
This
same. means
The peakthat the is
runoff model is mostly
directly sensitive
proportional toto
thethese threein
changes parameters and they
imperviousness andneed
zerotoimperviousness,
be adjusted
whereas it is inversely proportional to the D-Store impervious-pervious. This means that themodel
accordingly to get a good calibration. However, care should be taken that the sensitivity of the to
peak runoff
the three parameters is not the same. The peak runoff is directly proportional to the changes in
will increase with the increase of % imperviousness and % zero imperviousness. Likewise, the peak runoff
imperviousness and zero imperviousness, whereas it is inversely proportional to the D-Store impervious-
will decline with the reduction of the two parameters. On the other hand, the peak runoff will decline with
pervious. This means that the peak runoff will increase with the increase of % imperviousness and % zero
the increase of D-Store
imperviousness. impervious-pervious,
Likewise, the peak runoff willwhile it will
decline withincrease with the
the reduction D-Store
of the impervious-pervious
two parameters. On the
reduction. It should
other hand, the peakalso be noted
runoff that thewith
will decline peaktherunoff
increaseis more sensitive
of D-Store to the reductionwhile
impervious-pervious, of the it D-Store
will
impervious-pervious
increase with the D-Store than the increase. In otherreduction.
impervious-pervious words, the peak runoff
It should also be will
notedonly
thatdecrease
the peak 5% with
runoff is the
30%more sensitive
increase in the to D-Store
the reduction of the D-Store impervious-pervious
impervious-pervious, than the 12%
whereas it will increase increase.
withIntheother
30%words, the of
reduction
peakimpervious-pervious.
D-Store runoff will only decrease 5% with the 30% increase in the D-Store impervious-pervious, whereas it
will increase 12% with the 30% reduction of D-Store impervious-pervious.

Figure
Figure 4. 4. Sensitivityanalysis
Sensitivity analysis results
resultsfor SWMM
for model.
SWMM model.

3.2. 3.2.
Calibration and
Calibration andValidation
Validation

To To calibratethe
calibrate themodel,
model, the
thequantity
quantity and quality
and qualitydatadata
collected on 20 September
collected 2018 was2018
on 20 September used.was
Basedused.
on the sensitivity analysis performed and the goodness-of-fit test, the best result was achieved after running
Based on the sensitivity analysis performed and the goodness-of-fit test, the best result was achieved
the model for many times. To calibrate the model, the data series was applied to the model and the
after running the model for many times. To calibrate the model, the data series was applied to
respective parameters were set accordingly. The modeled values are compared with the observed ones.
the Then,
model theand theevaluation
model respective parameters
criteria were set
(goodness-of-fit accordingly.
test) is applied toThe modeled
the results. values
If the are compared
error between the
withmodeled
the observed ones. values
and observed Then,isthe model
in the evaluation
acceptable range, thecriteria
model(goodness-of-fit test) isthe
is calibrated, otherwise applied
processtois the
results. If theFigure
repeated. error 5between thethe
illustrates modeled
result ofand observed
model values
calibration. As is in obvious
it is the acceptable range,
from Figure the model
5, there is a is
discrepancy
calibrated, betweenthe
otherwise theprocess
modeledis and the observed
repeated. values
Figure at the beginning
5 illustrates of theof
the result hydrograph which couldAs it
model calibration.
be due to some uncertainties in modeling. The modeled and observed values
is obvious from Figure 5, there is a discrepancy between the modeled and the observed values are in good match before the
at the
hydrograph peak, at which point the model starts underestimating the flow a little bit. This trend continues
beginning of the hydrograph which could be due to some uncertainties in modeling. The modeled
and observed values are in good match before the hydrograph peak, at which point the model starts
underestimating the flow a little bit. This trend continues up to the middle of the hydrograph falling
Water 2019, 11, 1415 11 of 23

Water 2019, 11, x FOR PEER REVIEW 12 of 25

limb. From this point onward, the model overestimates the flow a little bit. Nevertheless, it is obvious
from Figure 5 that, overall, the modeled and observed hydrographs are a very good match. This could
up to the middle of the hydrograph falling limb. From this point onward, the model overestimates the flow
alsoa be confirmed
little by the itresults
bit. Nevertheless, of goodness-of-fit
is obvious test.
from Figure 5 that, As it the
overall, is obvious
modeledfrom the goodness-of-fit
and observed hydrographs test,
the are
evaluation parameters, namely NOF, NSC, and r2 are 0.05, 0.93, and 0.93, respectively, which show
a very good match. This could also be confirmed by the results of goodness-of-fit test. As it is obvious
a very
fromgood match between
the goodness-of-fit test,the
the modeled
evaluationand observed
parameters, NOF, NSC, and 𝑟 are 0.05, 0.93, and 0.93,
values.
namely
respectively, which show a very good match between the modeled and observed values.

Figure 5. Calibration
Figure 5. Calibrationofofthe
themodel forthe
model for therainfall
rainfallonon
20 20 September
September 20182018
event.event.

ForFor
thethe
model
modelvalidation,
validation, the
the data
data onon 1 August 2017was
August 2017 wasused.
used.This
This data
data waswas selected
selected fromfrom
the the
historical data collected
historical data collected from from the SMART control center. Figure 6 depicts the result of model validation.
SMART control center. Figure 6 depicts the result of model validation. As
can be seen, the modeled and observed values are in good match from the starting
As can be seen, the modeled and observed values are in good match from the starting point up to point up to the middle
theof the rising limb. However, the model overestimates the flow at the peak. This might be due to some
middle of the rising limb. However, the model overestimates the flow at the peak. This might be
uncertainties in the observed data. After this, the modeled and observed hydrographs follow each other,
due to some uncertainties in the observed data. After this, the modeled and observed hydrographs
although there is a little bit underestimation in the falling limb. Overall, the modeled and observed
follow each other,
hydrographs although
follow therethroughout
each other is a little bit
theunderestimation
validation process inand
the the
falling limb. Overall,
validation the modeled
is acceptable. The
andgoodness-of-fit
observed hydrographs follow
test also shows thateach other throughout
the validation the validation
result is quite acceptable,process
and the and
two the validation is
hydrographs
acceptable. Thematch
satisfactorily goodness-of-fit
each other.test alsobe
As can shows
seen, that the validation
the evaluation result is
parameters quite acceptable,
(goodness-of-fit test),and the two
namely
hydrographs and 𝑟 are 0.08,
NOF, NSC,satisfactorily 0.74,each
match and other.
0.77, respectively, whichthe
As can be seen, show an acceptable
evaluation match (goodness-of-fit
parameters between the
modeled
test), namely and observed
NOF, NSC,values.
and r2 are 0.08, 0.74, and 0.77, respectively, which show an acceptable match
Water 2019, 11, x FOR PEER REVIEW 13 of 25
between the modeled and observed values.

Figure
Figure 6. Validation
6. Validation ofofthe
themodel
model for
for the
therainfall
rainfallonon
1 August 20172017
1 August event.
event.

In order to calibrate the quality model for TSS and TN parameters derived from the 20 September 2018
In order to calibrate the quality model for TSS and TN parameters derived from the 20 September 2018
event sampling, the catchment was divided into two categories— residential and commercial—consisting
event sampling, the catchment was divided into two categories—residential and commercial—consisting of
of 70% and 30% of the catchment, respectively. The residential and commercial percentage of the study
area were derived through site inspection and google earth.
Similarly, there are influential parameters to be adjusted to calibrate the quality model as well. The
most important parameters affecting the model calibration are the buildup and wash-off functions and the
input parameters (C1, C2, C3, and C4). Pollutant accumulation can be represented by different types of
buildup functions on an urban catchment, including power function, exponential function, or saturation
Water 2019, 11, 1415 12 of 23

70% and 30% of the catchment, respectively. The residential and commercial percentage of the study area
were derived through site inspection and google earth.
Similarly, there are influential parameters to be adjusted to calibrate the quality model as well.
The most important parameters affecting the model calibration are the buildup and wash-off functions
and the input parameters (C1 , C2 , C3 , and C4 ). Pollutant accumulation can be represented by different
types of buildup functions on an urban catchment, including power function, exponential function,
or saturation equation, while the wash-off is simulated using exponential function, rating curve
equation or event mean concentration (EMC). Among various types of pollutant buildup and wash-off
functions, the exponential function was selected for both buildup and wash-off after testing the model
for different functions. The exponential functions for buildup and wash-off in SWMM are explained
as follows.
Buildup follows an exponential growth curve that approaches a maximum limit asymptotically,
 
B = C1 1 − e−C2 t (4)

where C1 = maximum buildup possible (mass per unit of area or curb length), C2 = buildup rate
constant (1/days), and t is the time.
The wash-off load (W) in units of mass per hour is proportional to the product of runoff raised to
some power and to the amount of buildup remaining, i.e.,

W = C3 qC4 B (5)

where C3 = wash-off coefficient, C4 = wash-off exponent, q = runoff rate per unit area (inches/hour
or mm/hour), and B = pollutant buildup in mass units. The buildup here is the total mass (not per
area or per curb length) and both buildup and wash-off mass units are the same as used to express the
pollutant’s concentration. The input parameters (C1 -C4 ) for quality modeling are depicted in Table 2.

Table 2. Results of buildup and wash-off input parameters to calibrate the quality model.

Build-Up Wash-Off
Land Use Pollutant
C1 C2 C3 C4
TSS 1.5 0.3 0.4 0.9
Residential
TN 0.002 0.05 12 1.7
TSS 12 0.3 1.5 0.6
Commercial
TN 0.1 0.7 0.3 3.5

Having adjusted the parameters (C1 –C4 ) after many runs, a satisfactory match between the
observed and calibrated values for TSS was achieved, as shown in Figure 7. As can be seen, there is
a good match between the modeled and the observed values in the rising limb. However, there is
an inconsistency at the peak, in which one observed point is higher than the modeled peak. This might
be due mainly to the uncertainty in sampling, data collection, or the model itself. In contrast, the model
overestimates the TSS a little bit in the falling limb, although the modeled and observed graphs converge
at the end. As the quality modeling is always more challenging than quantity one, these discrepancies
are quite normal. The evaluation parameters in Table 3, namely NOF, NSC, and r2 for TSS, are 0.14,
0.81, and 0.84, respectively, which show a satisfactory match between the modeled and observed
values. Overall, it is clear from the calibrated model and the result of test in Table 3 that the modeled
TSS graph follows the observed TSS graph and the calibration is satisfactory.
uncertainty in sampling, data collection, or the model itself. In contrast, the model overestimates the TSS a
little bit in the falling limb, although the modeled and observed graphs converge at the end. As the quality
modeling is always more challenging than quantity one, these discrepancies are quite normal. The
evaluation parameters in Table 3, namely NOF, NSC, and 𝑟 for TSS, are 0.14, 0.81, and 0.84, respectively,
which show a satisfactory match between the modeled and observed values. Overall, it is clear from
Water 2019, 11, 1415
the
13 of 23
calibrated model and the result of test in Table 3 that the modeled TSS graph follows the observed TSS
graph and the calibration is satisfactory.

Figure 7.
Figure 7. Total suspended sediments
Total suspended sediments (TSS)
(TSS)calibration
calibrationresult.
result.

ForForTN,
TN,it itwas
waseven
evenmoremoredifficult
difficult to
to calibrate
calibrate the
the model
model as TNTN is is more
moresensitive
sensitivetotothethemodel
model
parameters.
parameters. HardHard effort
effort waswas
made made to adjust
to adjust the parameters
the parameters in manyin many
runs toruns to get
get the bestthe bestFigure
result. result.8
illustrates
Figure 8the result of the
illustrates TN calibration.
result of TN Ascalibration.
is obvious, there
As isisobvious,
an incompatibility
there is anat the beginning and
incompatibility atinthe
the
middle of theand
beginning graphs.
in theFor the rest
middle of the
of the calibration,
graphs. For thethe modeled
rest graph follows
of the calibration, the the observed
modeled graphonefollows
and the
twothegraphs
observedconverge
one andvery wellgraphs
the two at the end, although
converge very there
well atisthe
an end,
overestimation
although therejust is
before the last point.
an overestimation
Thejustevaluation
before theparameters
last point. Thein Table 3, namely
evaluation NOF, NSC,
parameters and3,𝑟namely
in Table for TN, areNSC,
NOF, 0.74,r2and
0.14, and 0.74,
for TN,
respectively, which
are 0.14, 0.74, and are
0.74,inrespectively,
the acceptable range
which and
are in show a good match
the acceptable rangebetween
and show theamodeled
good match andbetween
observed
values. Overall,and
the modeled Figure 8 and values.
observed Table 3 Overall,
show that the result
Figure 8 andofTable
calibration
3 showwasthatquite satisfactory.
the result of calibration was
Water 2019, 11, x FOR PEER REVIEW 15 of 25
quite satisfactory.

Figure
Figure 8. Total nitrogen
8. Total nitrogen(TN)
(TN)calibration
calibrationresult.
result.

3.3.
3.3. Goodness-of-Fit
Goodness-of-Fit Test
Test Results
Results
The
The goodness-of-fit
goodness-of-fit test
test results
results forfor the
the quantity
quantity andand quality
quality calibration
calibration and
and validation
validation aredepicted
are depictedin
in Table
Table 3. 3.
AsAs shown
shown in in Table
Table 3, 3,
thethe results
results of of quantity
quantity calibration
calibration areare quite
quite satisfactory.
satisfactory. Although
Although for for
TSSTSS
and
and TN, the results are not the same as quantity calibration, the three evaluated parameters still
TN, the results are not the same as quantity calibration, the three evaluated parameters still show a good show
match between the predicted and the observed values in both TSS and TN, based on the criteria for the
goodness-of-fit test.

Table 3. The goodness-of-fit test results to evaluate the reliability of the model.
 Water 2019, 11, 1415 14 of 23

a good match between the predicted and the observed values in both TSS and TN, based on the criteria
for the goodness-of-fit test.

Table 3. The goodness-of-fit test results to evaluate the reliability of the model.

Flow Flow TSS

Parameter TN Calibration
Calibration Validation Calibration
NOF 0.05 0.08 0.14 0.14
NSC 0.93 0.74 0.81 0.74
r2 0.93 0.77 0.84 0.74

3.4. The Results of Assigning LIDs to the Model

Having finished the calibration and validation of the model, the selected LIDs were assigned to the
model. Then, the model was executed for the same rainfall event that was calibrated (20 September 2018).
The aim was to check the performance of the model after assigning the LIDs. The results of the model
with the applied LIDs are shown in Table 4 and Figures 9 and 10.

Table 4. Best management practices (BMP) removal and peak runoff reduction after assigning low
impact developments (LIDs) to the model.

Rainfall on LID Removal (%) Peak Runoff Reduction

Water 2019, 11, x FOR PEER
20 REVIEW
September 2018 16 of 25
TN (%) TSS (%) (%)
(mm)
12.50 29 41 23
12.50 29 41 23
Rainfall Observed TSS TSS without LID TSS with LID
160 0

140 1

2
120
3

Rainfall (mm)
100
TSS (mg/L)

4
80
5
60
6
40
7

20 8

0 9
17:20 17:30 17:40 17:50 18:00 18:10 18:20 18:30 18:40 18:50 19:00 19:10 19:20 19:30 19:40 19:50
Time (min)

Figure9.9.TSS
Figure TSS reduction
reduction after assigning LIDs
LIDs to
to the
the model.
model.

Rainfall Observed TN TN without LID TN with LID

10 0

9 1

8 2

7 3
Rainfall (mm)

6 4
TN (mg/L)

5 5

4 6

3 7

2 8
 20 8

0 9
17:20 17:30 17:40 17:50 18:00 18:10 18:20 18:30 18:40 18:50 19:00 19:10 19:20 19:30 19:40 19:50
Time (min)
Water 2019, 11, 1415 15 of 23
Figure 9. TSS reduction after assigning LIDs to the model.

Rainfall Observed TN TN without LID TN with LID

10 0

9 1

8 2

7 3

Rainfall (mm)
6 4
TN (mg/L)

5 5

4 6

3 7

2 8

1 9

0 10
17:20 17:30 17:40 17:50 18:00 18:10 18:20 18:30 18:40 18:50 19:00 19:10 19:20 19:30 19:40 19:50 20:00
Time (min)
Figure10.
Figure 10.TN
TNreduction
reduction after
after assigning LIDs to
to the
the model.
model.

Theresults
The resultsdepicted
depictedininTable
Table 44 show
show that TN and TSS TSS reduced
reducedby by29%
29%andand41%,
41%,respectively,
respectively,after
after
assigningLIDs
assigning LIDstotothe
themodel.
model. The
The peak
peak runoff also reduced
reduced by by about
about23%.
23%.Figures
Figures99and and1010also
alsoclearly
clearly
show
showthat
thatTSS
TSSand
andTNTN reductions
reductionswere satisfactory
were afterafter
satisfactory assigning the LIDs
assigning to thetomodel.
the LIDs As canAs
the model. becan
seen
from the table
be seen fromand
the figures, LIDs
table and have LIDs
figures, been more
have efficient
been more to reduce TSS
efficient toand TN compared
reduce TSS and TN to peak runoff.
compared
This means
to peak that more
runoff. ThisLID techniques
means that moreare LID
required in the study
techniques area to effectively
are required in the studycontrol
areaboth runoff and
to effectively
pollutants.
control bothHowever, the pollutants.
runoff and cost of LIDHowever,
implementation
the costwill alsoimplementation
of LID be an importantwillissue.
alsoInbe
other words, the
an important
issue. In other words, the number and placement of LIDs should be optimized in order to have
the maximum runoff and pollutant reduction with the minimum cost. The role of LIDs is to make
the impervious surfaces more permeable to reduce runoff and remove pollutants through filtration,
chemical, and biological processes. In this study, both the table and the figures confirm that LID
practices are of high importance in all developed urban areas to mitigate the effects of imperviousness
on both runoff quantity and quality. Thus, it can be concluded that LID techniques could reduce urban
runoff and improve stormwater quality substantially.

4. Discussion

4.1. Summary and Findings

Urbanization disturbs the natural landscapes; vegetation covered surfaces are replaced with
impervious surfaces. Thus, impervious surfaces such as roof tops, parking lots, and roads, decreases
infiltration capacity, and increases runoff generation in urban areas [5]. Low impact development
(LID), such as bioretention cells and vegetated swales, are techniques implemented at or near the
source of runoff generation to make the ground more permeable and mitigate stormwater runoff and
its pollutants [22]. Particularly, the LID techniques have been developed to mimic the predevelopment
hydrologic conditions and promote the storage, infiltration and evapotranspiration processes [24].
The main roles of LID practices include runoff reduction (peak and volume), infiltration increase,
groundwater recharge, stream protection, and water quality enhancement by removing pollutants
through mechanics such as filtration, chemical sorption, and biological processes [70].
The EPA SWMM was used in this study to develop a model to be able to manage stormwater
runoff and pollutants at the source using LID controls. To develop SWMM, a wide range of climatic
and field data are required to calibrate and validate the model with rainfall and flow being the main
one. Based on this, the required data was collected from a rainfall event of the study area. Vegetated
Water 2019, 11, 1415 16 of 23

swale and rain garden were selected as LID practices for this purpose. Rain garden and swale are
suitable for residential and commercial areas [35,71]. These two LID techniques have been proved to
be highly efficient in urban areas for both runoff and pollutant reduction [35,68]. The specific aim of
the study was to investigate the efficiency of the selected LIDs in reducing the stormwater runoff and
its pollutants in higher intensity and longer duration rainfalls in urban areas.
The model sensitivity to different parameters was first checked to find out the most sensitive
parameters for model calibration. Several parameters impact the model calibration. Previous studies
have found out that different parameters impact the model calibration. The main parameters affecting
the model calibration are subcatchments area, percent impervious area, width, slope, infiltration
parameters, Manning’s roughness coefficients for pervious and impervious surfaces, depression
storage depth for pervious and impervious surfaces, percent zero, and internal routing parameters [48].
In this study, the most sensitive parameters for the model calibration were found out to be %
imperviousness (percent of the land area which is impervious), % zero imperviousness (percent of the
impervious area with no depression storage), Dstore-impervious (depth of depression storage on the
impervious portion of the subcatchment), and Dstore-pervious (depth of depression storage on the
pervious portion of the subcatchment).
Appropriate calibration of rainfall–runoff models for urban catchments is necessary to ensure
reliable assessment of stormwater modeling results [72]. Rainfall–runoff models can be calibrated
over a set of single storm events or continuous storm events. Calibrating the model with a single
event provides better time to peak and overall hydrograph shape compared to continuous calibration,
but continuous event calibration gives more accurate estimation of the total runoff volume [73].
Single storm event calibration is a rapid process and usually does not require a great deal of observed
data [72]. A large number of previous studies can be found in the literature that calibrated their models
over a single storm event [43,73,74]. In the current study, the model was also calibrated using a single
storm event since the peak ruoff was the target of assessment. Moreover, the model is a quantity–quality
one, and calibrated was performed for both quantity and quality. There was no continues quality data
available to perfom continues calibration. Therefore, the quantity–quality calibration was conducted
over a single-event, collected individually on 20 September 2018.
The results of quantity and quality calibration and validation are depicted in Figures 5–8.
Figures 5 and 6 illustrate the flow calibration and validation, respectively. Figures 7 and 8 depict
the TSS and TN calibration, respectively. As for the flow, the figures and evaluation test (Table 3)
show that there is a good match between the modeled and the observed values. The stormwater
quality is generally more difficult than quantity, as urban runoff quality monitoring needs substantial
resources [58]. However, in our case, the quality modeling was also satisfactory. As it is obvious
from Figures 7 and 8 and Table 3, there is a satisfactory match between the modeled and observed
values. The results of goodness-of-fit test in this study have been compared with the results of two
previous studies. Chow, Yusop, Toriman, and Technology [67] developed a model and presented the
results of SWMM calibration and validation for modeling runoff quantity and quality in tropical areas.
Mancipe-Munoz, Buchberger, Suidan, and Lu [72] also presented the calibration and validation of
a rainfall–runoff SWMM 5 in their study. The comparison has been presented in Table 5.

Table 5. Comparison of the results of goodness-of-fit test with previous studies.

Flow Calibration Flow Validation TSS

Parameter
This Study Chow [67] Munoz [72] This Study Chow [67] Munoz [72] This Study Chow [67] Munoz [72]
NOF 0.05 0.06 - 0.08 - - 0.14 0.24 -
NSC 0.93 0.99 0.60 0.74 - 0.62 0.81 0.75 -
r2 0.93 0.99 - 0.77 - - 0.84 0.77 -

According to Table 5, the model in this study has outperformed the other two studies in both flow
calibration–validation and TSS calibration although the two parameters, namely NSC and r2 , are a little
bit lower than [67] in flow calibration. In terms of TSS, the model in this study has performed much
Water 2019, 11, 1415 17 of 23

better than [67] although TSS calibration is more complex than quantity. In other words, the model
calibration and validation in this study have been quite satisfactory for both quantity and quality.
Table 4 and Figures 9 and 10 present the results of LID control assigning to the model for the same
rainfall that the model was calibrated. The results show a reduction of 23%, 41%, and 29% reduction for
the peak runoff, TSS, and TN, respectively. As mentioned in the introduction, rain garden (bioretention)
is relatively highly efficient for both runoff and pollutant reduction [35]. Bioretention cells could reduce
the average peak flows by at least 45% in Maryland and North Carolina [36]. Bioretentions are also
capable of reducing sediment and nutrient from 0% to 99% [35]. Swales have also been shown to have
an average retention of 14% to 98% for nutrients and TSS, and up to 93% for metals [24]. According to
this, the peak runoff reduction and pollutant removal of LIDs in this study are not compatible with their
utmost capabilities. This might be mainly due to the fewer units of LIDs used in this study. In other
words, more LIDs are required to be implemented in the study area to improve LID efficiency for both
peak runoff reduction and pollutants removal. However, the cost of LID implementation should also
be considered in this regard. Optimization methods could be applied to find the optimal number and
placement of LIDs for the maximum flow and pollutants reduction with the minimum cost.
The results of model performance for the selected rainfall scenarios have also been depicted
in Table 6 and Figure 11. The table shows that the model performance is quite efficient in terms of TSS
and TN removal for different rainfall amounts. The LID removal efficiency is up to 40% and 62% for
TN and TSS, respectively. The results for TSS and TN show that the LID removal efficiency for TSS and
TN is independent of the rainfall amount and the model is highly efficient for TSS and TN removal for
both short and long storm events.
As for the peak runoff reduction, it could be concluded from Tables 4 and 6 and Figure 11 that
in smaller rainfall amounts of up to 70 mm, the model performed well, and the peak runoff reduction
reached up to 27%. In rainfall amounts of between 70 and 90 mm, the model performance was
moderately good, and the reduction of peak runoff reached up to 19%. In the case of higher intensity
rainfalls, when the rainfall was higher than 90 mm, the model performance, in terms of runoff reduction,
was poor. That is mainly because in high intensity rainfalls the soil is saturated, and the input and
output are equal. In this case, no more runoff infiltrates into the soil and the LIDs and the soil beneath
act as filters only to remove pollutants. Generally, the model performed satisfactorily for rainfall of up
to 90 mm and for the return period of up to 10 years.

Table 6. Results of model performance for the selected rainfall scenarios.

Return Period Duration Rainfall LID Removal (%) Peak Runoff

(T) (hour) (mm) TN (%) TSS (%) Reduction (%)

5 1 72 40 61 27
5 1.5 81 40 61 19
5 2 90 40 62 18
10 1 80 40 61 18
10 1.5 90 40 61 9
10 2 96 39 60 0
20 1 90 38 58 0
20 1.5 99 38 59 0
20 2 110 39 60 0

Nevertheless, for the return period of more than 10 years and for the rainfall amount of more than
90 mm, the designed LIDs for the catchment cannot handle the surcharge amount of runoff in urban
areas, and either we need more LIDs to be installed or a combination of conventional BMPs and LIDs
is required to tackle the excess runoff in impervious surfaces to avoid any flooding in urban areas.
The results achieved in this study for high intensity rainfalls are well consistent with the findings of
previous studies on LID runoff reduction efficiency in urban areas, e.g., Hood, Clausen, and Warner [34],
and Damodaram, Giacomoni, Prakash Khedun, Holmes, Ryan, Saour, and Zechman [27].
Based on the results achieved, the final word on the application of this study could be that
LID practices (here swale and rain garden) are vital in all developed urban areas and they can
 Water 2019, 11, 1415 18 of 23

substantially reduce runoff and enhance water quality. They can also replace the conventional
stormwater management systems in lower rainfall events (less than 90 mm). However, in higher
rainfall events, they need to be coupled with conventional stormwater management systems to tackle
the2019,
Water runoff,
11, x but
FOR for
PEERpollutant
REVIEW removal, they maintain good efficiency in high intensity rainfall
20 ofas
25 well.
Furthermore, the cost of implementation is also a concern and should be considered in urban planning.

Figure 11. Peak

Figure Peak runoff
runoff reduction
reduction for
forthe
theselected
selectedrainfall
rainfallscenarios.
scenarios.

4.2.
4.2. Future
Future Research
Research Directions
Directions
Finally,future
Finally, futurestudies
studiesarearesuggested
suggestedto to consider
consider the
the following
following issues
issues in
in the
the context
contextofofstormwater
stormwater
modeling. Uncertainty reduction is of high importance in modeling. There are various sourcesofof
modeling. Uncertainty reduction is of high importance in modeling. There are various sources
uncertainties
uncertainties totobebeconsidered
consideredininstormwater
stormwaterrunoffrunoff modeling
modeling from
from data
data collection
collection toto model
modelcalibration.
calibration.
Thesources
The sourcesofofuncertainties
uncertaintiescan canbe
be classified
classified as input uncertainties,
uncertainties, parameter
parameteruncertainties,
uncertainties,and andmodel
model
structure
structure uncertainties.
uncertainties. Therefore,
Therefore, these
these uncertainties
uncertainties could
could be addressed
be addressed in future
in future studies.
studies. Another
Another issue
issue to be considered is that other pollutants such as heavy metals and pathogens
to be considered is that other pollutants such as heavy metals and pathogens could be investigated in could be investigated
in further
further researches
researches to studyto study the removal
the removal efficiency
efficiency of LIDsofmore
LIDsaccurately.
more accurately.
It is alsoItsuggested
is also suggested
that Pilot
that Pilot
testing to be testing
carried to outbeincarried outto
laboratory ininvestigate
laboratorythe to effect
investigate
of LIDsthe
oneffect of LIDs
hydrology andon hydrology
water and
quality more
precisely. Moreover, the land cover in urban catchments is not homogeneous,
water quality more precisely. Moreover, the land cover in urban catchments is not homogeneous, which affect the physical
processes. Thus,
which affect thethe spatial-temporal
physical processes.distribution of precipitation in
Thus, the spatial-temporal urban areas
distribution ofisprecipitation
important and should
in urban
beareas
addressed in future
is important andstudies.
should be addressed in future studies.

5. 5. Conclusions
Conclusions
AA quantity–quality
quantity–quality model
model waswas developed
developed using
using thethe
USUS EPAEPA SWMM.
SWMM. TheThe model
model aimed
aimed to to assess
assess the
impact of LIDofon
the impact LIDstormwater quantity
on stormwater and quality
quantity in a in
and quality subcatchment
a subcatchment in Kuala Lumpur,
in Kuala Lumpur,Malaysia. The
Malaysia.
required rainfall-flow
The required data anddata
rainfall-flow quality
anddata weredata
quality collected
were in the fieldinfrom
collected the afield
real from
event aonreal
20 September
event on
2018.
20 September 2 catchment was
The 18 km2018. The 18 km2 catchment
divided into was35 divided
subcatchments using SWMM model.
into 35 subcatchments The model
using SWMM was
model.
calibrated
The model was calibrated and validated both for quantity and quality using the real data fromA
and validated both for quantity and quality using the real data from 20 September 2018 event.
sensitivity analysis
20 September 2018was also performed
event. A sensitivity beforehand
analysis wasto find
alsoout the most sensitive
performed beforehand parameters of the
to find out themodel.
most
The developed
sensitive parametersmodel was
of the to simulate LID techniques at the catchment scale by applying vegetated
model.
swale and
Therain gardenmodel
developed as efficient
was to practices
simulate forLID
urban areas. The
techniques impacts
at the of LIDscale
catchment practices on water
by applying quantity
vegetated
and water quality were evaluated using both the collected field data and the selected
swale and rain garden as efficient practices for urban areas. The impacts of LID practices on water rainfall scenarios
derived from
quantity andthe IDF curves
water qualityfor
weretheevaluated
study area. using both the collected field data and the selected rainfall
Based on the model performance and
scenarios derived from the IDF curves for the the results
studyachieved,
area. the following conclusions could be made.
 The most sensitive parameters of the model are % imperviousness, % zero imperviousness, and D-
Store impervious-pervious. That is to say, a slight change in the imperviousness, the depression
Water 2019, 11, 1415 19 of 23

Based on the model performance and the results achieved, the following conclusions could be made.

• The most sensitive parameters of the model are % imperviousness, % zero imperviousness,
and D-Store impervious-pervious. That is to say, a slight change in the imperviousness,
the depression storage or the depth of depression storage will significantly change the simulated
runoff and the peak flow. However, the model is more sensitive to D-Store impervious-pervious
rather than the other two parameters. It was also noted that the peak runoff will be more
affected when D-Store impervious-pervious decreased rather than increased. It means that a slight
reduction in the depth of depression storage, will substantially increase the peak runoff.
• In terms of water quality, the developed model performed well. The LID removal efficiency
reached up to 40% for TN and up to 62% for TSS, respectively. The LID removal efficiency of the
model was independent of the rainfall intensity and duration, taking into account the current
research rainfall scenarios.
• As for the peak runoff reduction, in smaller rainfall of up to 70 mm the model performed well
and the peak runoff reduction reached up to 27%. In rainfall amounts between 70 and 90 mm,
the model performance was moderately good, and the reduction of peak runoff reached up to
19%. In the case of higher intensity rainfalls when the rainfall was higher than 90 mm, the model
performance in terms of runoff reduction was poor.
• Overall, the model performed satisfactorily for rainfall of up to 90 mm and for the return period
of up to 10 years. Nevertheless, for the return period of more than 10 years and for the rainfall
amount of more than 90 mm, the designed LIDs for the catchment cannot handle the surcharge
amount of runoff in urban areas.
• The LIDs applied for the catchment in this study are more effective in peak runoff reduction
during lower intensity rainfall events. Therefore, it would be more efficient to combine the LID
techniques with other conventional stormwater management practices to control urban flooding
in case of high intensity storm events. However, the LID removal efficiency for TSS and TN was
quite satisfactory in all selected rainfall scenarios. The LID applied in this study performed well
in improving water quality in both low and high intensity rainfall events.

Thus, the model confirms the significant role of LID in reducing peak runoff and improving water
quality in urban stormwater events.

Author Contributions: A.R.R. completed this study as part of his PhD thesis under the supervision of Z.I.
and M.H.N. The paper was written by A.R.R., while technical support and revisions were provided by Z.I.,
M.H.N., and M.A.D. during model development and calibration as well as reviewing the paper final version.
A.H.R. assisted in data collection and revision, whereas S.M.S. reviewed and provided feedback.
Funding: This research was supported by the RP013A-15SUS grant.
Acknowledgments: The authors would like to thank the Malaysian Department of Irrigation and Drainage (DID) and
the SMART control center for providing the research with the required data. The ArcGIS computations were performed
in the Computer Laboratory, Block J, Faculty of Engineering, University of Malaya, 50603 Kuala Lumpur, Malaysia.
Conflicts of Interest: The authors declare no conflicts of interest.

References
1. Han, L.; Zhou, W.; Li, W.; Li, L. Impact of urbanization level on urban air quality: A case of fine particles
(PM 2.5) in Chinese cities. Environ. Pollut. 2014, 194, 163–170. [CrossRef] [PubMed]
2. Long, H.; Liu, Y.; Hou, X.; Li, T.; Li, Y. Effects of land use transitions due to rapid urbanization on ecosystem
services: Implications for urban planning in the new developing area of China. Habitat Int. 2014, 44, 536–544.
[CrossRef]
3. Chen, Y.; Samuelson, H.W.; Tong, Z.J. Integrated design workflow and a new tool for urban rainwater
management. J. Environ. Manag. 2016, 180, 45–51. [CrossRef] [PubMed]
4. Eckart, K.; McPhee, Z.; Bolisetti, T. Performance and implementation of low impact development–A review.
Sci. Total. Environ. 2017, 607, 413–432. [CrossRef] [PubMed]
Water 2019, 11, 1415 20 of 23

5. Shuster, W.D.; Bonta, J.; Thurston, H.; Warnemuende, E.; Smith, D.R. Impacts of impervious surface on
watershed hydrology: A review. Urban Water J. 2005, 2, 263–275. [CrossRef]
6. Yang, L.; Tian, F.; Niyogi, D. A need to revisit hydrologic responses to urbanization by incorporating the
feedback on spatial rainfall patterns. Urban Clim. 2015, 12, 128–140. [CrossRef]
7. Stovin, V.R.; Moore, S.L.; Wall, M.; Ashley, R.M. The potential to retrofit sustainable drainage systems to
address combined sewer overflow discharges in the T hames T ideway catchment. Water Environ. J. 2013, 27,
216–228. [CrossRef]
8. Qin, H.P.; He, K.M.; Fu, G. Modeling middle and final flush effects of urban runoff pollution in an urbanizing
catchment. J. Hydrol. 2016, 534, 638–647. [CrossRef]
9. Wang, J.; Zhang, P.; Yang, L.; Huang, T. Cadmium removal from urban stormwater runoff via bioretention
technology and effluent risk assessment for discharge to surface water. J. Contam. Hydrol. 2016, 185, 42–50.
[CrossRef] [PubMed]
10. Zhang, P.; Cai, Y.; Wang, J. A simulation-based real-time control system for reducing urban runoff pollution
through a stormwater storage tank. J. Clean. Prod. 2018, 183, 641–652. [CrossRef]
11. Paule-Mercado, M.C.A.; Salim, I.; Lee, B.-Y.; Memon, S.; Sajjad, R.U.; Sukhbaatar, C.; Lee, C.-H. Monitoring
and quantification of stormwater runoff from mixed land use and land cover catchment in response to land
development. Ecol. Indic. 2018, 93, 1112–1125. [CrossRef]
12. Zhang, K.; Deletic, A.; Bach, P.M.; Shi, B.; Hathaway, J.M.; McCarthy, D.T. Testing of new stormwater
pollution build-up algorithms informed by a genetic programming approach. J. Environ. Manag. 2019, 241,
12–21. [CrossRef] [PubMed]
13. Steele, J.A.; Blackwood, A.D.; Griffith, J.F.; Noble, R.T.; Schiff, K.C. Quantification of pathogens and markers
of fecal contamination during storm events along popular surfing beaches in San Diego, California. Water Res.
2018, 136, 137–149. [CrossRef] [PubMed]
14. Chen, W.-B.B. Optimal Allocation of Stormwater Pollution Control Technologies in a Watershed. Ph.D. Thesis,
The Ohio State University, Columbus, OH, USA, 2006.
15. Maksimovic, C.; Tucci, C.E. Urban Drainage in Specific Climates; Volume I: Urban Drainage in Humid Tropics;
Unesco: Paris, France, 2001; Volume 1.
16. Stec, A.; Słyś, D. Effect of development of the town of Przemysl on operation of its sewerage system.
Ecol. Chem. Eng. S 2013, 20, 381–396. [CrossRef]
17. Paule-Mercado, M.; Lee, B.; Memon, S.; Umer, S.; Salim, I.; Lee, C.-H. Influence of land development on
stormwater runoff from a mixed land use and land cover catchment. Sci. Total Environ. 2017, 599, 2142–2155.
[CrossRef] [PubMed]
18. Ahiablame, L.; Shakya, R. Modeling flood reduction effects of low impact development at a watershed scale.
J. Environ. Manag. 2016, 171, 81–91. [CrossRef] [PubMed]
19. Leitão, J.; Carbajal, J.; Rieckermann, J.; Simões, N.; Marques, A.S.; de Sousa, L. Identifying the best locations
to install flow control devices in sewer networks to enable in-sewer storage. J. Hydrol. 2018, 556, 371–383.
[CrossRef]
20. Marques, G.F.; de Souza, V.B.; Moraes, N.V. The economic value of the flow regulation environmental service
in a Brazilian urban watershed. J. Hydrol. 2017, 554, 406–419. [CrossRef]
21. Qin, H.-P.; Li, Z.-X.; Fu, G. The effects of low impact development on urban flooding under different rainfall
characteristics. J. Environ. Manag. 2013, 129, 577–585. [CrossRef]
22. Elliott, A.; Trowsdale, S.A. A review of models for low impact urban stormwater drainage. Environ. Model. Softw.
2007, 22, 394–405. [CrossRef]
23. Tuomela, C.; Sillanpää, N.; Koivusalo, H. Assessment of stormwater pollutant loads and source area contributions
with storm water management model (SWMM). J. Environ. Manag. 2019, 233, 719–727. [CrossRef] [PubMed]
24. Ahiablame, L.M.; Engel, B.A.; Chaubey, I. Effectiveness of low impact development practices: Literature
review and suggestions for future research. Water Air Soil Pollut. 2012, 223, 4253–4273. [CrossRef]
25. Ahiablame, L.M.; Engel, B.A.; Chaubey, I. Effectiveness of low impact development practices in two urbanized
watersheds: Retrofitting with rain barrel/cistern and porous pavement. J. Environ. Manag. 2013, 119, 151–161.
[CrossRef] [PubMed]
26. Autixier, L.; Mailhot, A.; Bolduc, S.; Madoux-Humery, A.-S.; Galarneau, M.; Prévost, M.; Dorner, S.
Evaluating rain gardens as a method to reduce the impact of sewer overflows in sources of drinking water.
Sci. Total Environ. 2014, 499, 238–247. [CrossRef] [PubMed]
Water 2019, 11, 1415 21 of 23

27. Damodaram, C.; Giacomoni, M.H.; Prakash Khedun, C.; Holmes, H.; Ryan, A.; Saour, W.; Zechman, E.M.
Simulation of Combined Best Management Practices and Low Impact Development for Sustainable
Stormwater Management 1. J. Am. Water Resour. Assoc. 2010, 46, 907–918. [CrossRef]
28. Kok, K.H.; Sidek, L.M.; Abidin, M.; Basri, H.; Muda, Z.; Beddu, S. Evaluation of green roof as green technology
for urban stormwater quantity and quality controls. In IOP Conference Series: Earth and Environmental Science;
IOP Publishing Ltd.: Bristol, UK, 2013; Volume 16, p. 012045.
29. Newcomer, M.E.; Gurdak, J.J.; Sklar, L.S.; Nanus, L. Urban recharge beneath low impact development and
effects of climate variability and change. Water Resour. Res. 2014, 50, 1716–1734. [CrossRef]
30. Vijayaraghavan, K.; Joshi, U.; Balasubramanian, R. A field study to evaluate runoff quality from green roofs.
Water Res. 2012, 46, 1337–1345. [CrossRef]
31. Zhang, X.; Zhang, M. Modeling effectiveness of agricultural BMPs to reduce sediment load and
organophosphate pesticides in surface runoff. Sci. Total Environ. 2011, 409, 1949–1958. [CrossRef]
32. Drake, J.; Bradford, A.; Van Seters, T. Hydrologic performance of three partial-infiltration permeable
pavements in a cold climate over low permeability soil. J. Hydrol. Eng. 2013, 19, 04014016. [CrossRef]
33. Liu, Y.; Ahiablame, L.M.; Bralts, V.F.; Engel, B.A. Enhancing a rainfall-runoff model to assess the impacts of
BMPs and LID practices on storm runoff. J. Environ. Manag. 2015, 147, 12–23. [CrossRef]
34. Hood, M.J.; Clausen, J.C.; Warner, G.S. Comparison of Stormwater Lag Times for Low Impact and Traditional
Residential Development 1. J. Am. Water Resour. Assoc. 2007, 43, 1036–1046. [CrossRef]
35. Dietz, M.E. Low impact development practices: A review of current research and recommendations for
future directions. Water Air Soil Pollut. 2007, 186, 351–363. [CrossRef]
36. Hunt, W.; Smith, J.; Jadlocki, S.; Hathaway, J.; Eubanks, P. Pollutant removal and peak flow mitigation by
a bioretention cell in urban Charlotte, NC. J. Environ. Eng. 2008, 134, 403–408. [CrossRef]
37. Jackisch, N.; Weiler, M. The hydrologic outcome of a Low Impact Development (LID) site including
superposition with streamflow peaks. Urban Water J. 2017, 14, 143–159. [CrossRef]
38. Wilson, C.; Hunt, W.; Winston, R.; Smith, P. Comparison of runoff quality and quantity from a commercial
low-impact and conventional development in Raleigh, North Carolina. J. Environ. Eng. 2014, 141, 05014005.
[CrossRef]
39. Bosley, I.; Kern, E. Hydrologic Evaluation of Low Impact Development Using a Continuous, Spatially-Distributed
Model. Master’s Thesis, Virginia Polytechnic Institute and State University, Blacksburg, VA, USA, 2008.
40. Brown, R.; Line, D.; Hunt, W. LID treatment train: Pervious concrete with subsurface storage in series with
bioretention and care with seasonal high water tables. J. Environ. Eng. 2011, 138, 689–697. [CrossRef]
41. Lenhart, H.A.; Hunt, W.F., III. Evaluating four storm-water performance metrics with a North Carolina
coastal plain storm-water wetland. J. Environ. Eng. 2010, 137, 155–162. [CrossRef]
42. Mayer, A.L.; Shuster, W.D.; Beaulieu, J.J.; Hopton, M.E.; Rhea, L.K.; Roy, A.H.; Thurston, H.W. Environmental
reviews and case studies: Building green infrastructure via citizen participation: A six-year study in the
Shepherd Creek (Ohio). Environ. Pract. 2012, 14, 57–67. [CrossRef]
43. Palla, A.; Gnecco, I. Hydrologic modeling of low impact development systems at the urban catchment scale.
J. Hydrol. 2015, 528, 361–368. [CrossRef]
44. Jia, H.; Wang, X.; Ti, C.; Zhai, Y.; Field, R.; Tafuri, A.N.; Cai, H.; Shaw, L.Y. Field monitoring of a LID-BMP
treatment train system in China. Environ. Monit. Assess. 2015, 187, 373. [CrossRef]
45. Rossman, L.A. Storm Water Management Model User’s Manual; Version 5.0; National Risk Management
Research Laboratory, Office of Research and Development, Environmental Protection Agency: Cincinnati,
OH, USA, 2010.
46. Zoppou, C. Review of urban storm water models. Environ. Model. Softw. 2001, 16, 195–231. [CrossRef]
47. Duan, H.-F.; Li, F.; Yan, H. Multi-objective optimal design of detention tanks in the urban stormwater
drainage system: LID implementation and analysis. Water Resour. Manag. 2016, 30, 4635–4648. [CrossRef]
48. Eckart, K.; McPhee, Z.; Bolisetti, T. Multiobjective optimization of low impact development stormwater
controls. J. Hydrol. 2018, 562, 564–576. [CrossRef]
49. Yao, L.; Wei, W.; Chen, L. How does imperviousness impact the urban rainfall-runoff process under various
storm cases? Ecol. Indic. 2016, 60, 893–905. [CrossRef]
50. Zhang, G.; Hamlett, J.M.; Reed, P.; Tang, Y. Multi-objective optimization of low impact development designs
in an urbanizing watershed. Open J. Optim. 2013, 2, 95. [CrossRef]
Water 2019, 11, 1415 22 of 23

51. Huber, W.C.; Dickinson, R.E.; Barnwell, T.O., Jr.; Branch, A. Storm Water Management Model; Version 4;
United States Environmental Protection Agency: Washington, DC, USA, 1988.
52. Sun, N. Development of a Stormwater Model for Testing the Hydrological Effectiveness of Green Infrastructure
Implementation Scenarios in Urban Sewersheds; State University of New York College of Environmental Science
and Forestry: New York, NY, USA, 2012.
53. Water, M. WSUD Engineering Procedures: Stormwater; Csiro Publishing: Collingwood, Australia, 2005.
54. Rossman, L.A. SWMM (Stormwater Management Model) Version 5 User Manual; USEPA (Environmental
Protection Agency): Washington, DC, USA, 2004.
55. Liong, S.; Chan, W.; Lum, L. Knowledge-based system for SWMM runoff component calibration.
J. Water Resour. Plan. Manag. 1991, 117, 507–524. [CrossRef]
56. Warwick, J.; Tadepalli, P. Efficacy of SWMM application. J. Water Resour. Plan. Manag. 1991, 117, 352–366.
[CrossRef]
57. Zhang, G. Development of a Multi-Objective Optimization Framework for Implementing Low Impact
Development Scenarios in an Urbanizing Watershed. Ph.D. Thesis, The Pennsylvania State University,
University Park, PA, USA, 2009.
58. Zhu, H.; Xu, Y.; Yan, B.; Guan, J. Snowmelt runoff: A new focus of urban nonpoint source pollution.
Int. J. Environ. Res. Public Health 2012, 9, 4333–4345. [CrossRef] [PubMed]
59. Rossman, L.; Huber, W. Storm Water Management Model Reference Manual Volume II–Hydraulics; Technical
Report EPA/600/R-17/111; US EPA Office of Research and Development, Water Systems Division: Cincinnati,
OH, USA, 2016. Available online: https://nepis.epa.gov/Exe/ZyPDF.cgi (accessed on 15 June 2018).
60. Bonhomme, C.; Petrucci, G. Should we trust build-up/wash-off water quality models at the scale of urban
catchments? Water Res. 2017, 108, 422–431. [CrossRef] [PubMed]
61. Wijesiri, B.; Egodawatta, P.; McGree, J.; Goonetilleke, A. Understanding the uncertainty associated with
particle-bound pollutant build-up and wash-off: A critical review. Water Res. 2016, 101, 582–596. [CrossRef]
[PubMed]
62. Chen, J. Integrated Urban Stormwater Management Modeling. Ph.D. Thesis, University of Toronto, Toronto,
ON, Canada, 2004.
63. Seitzinger, S.P.; Sanders, R.; Styles, R. Bioavailability of DON from natural and anthropogenic sources to
estuarine plankton. Limnol. Oceanogr. 2002, 47, 353–366. [CrossRef]
64. Galloway, J.N.; Aber, J.D.; Erisman, J.W.; Seitzinger, S.P.; Howarth, R.W.; Cowling, E.B.; Cosby, B.J. The nitrogen
cascade. Bioscience 2003, 53, 341–356. [CrossRef]
65. Kornecki, T.S.; Sabbagh, G.J.; Storm, D.E. EVALUATION Of RUNOFF, EROSION, AND PHOSPHORUS
MODELING SYSTEM—SIMPLE 1 . J. Am. Water Resour. Assoc. 1999, 35, 807–820. [CrossRef]
66. Nash, J.E.; Sutcliffe, J.V. River flow forecasting through conceptual models part I—A discussion of principles.
J. Hydrol. 1970, 10, 282–290. [CrossRef]
67. Chow, M.; Yusop, Z.; Toriman, M.E. Modelling runoff quantity and quality in tropical urban catchments
using Storm Water Management Model. Int. J. Environ. Sci. Technol. 2012, 9, 737–748. [CrossRef]
68. Martin-Mikle, C.J.; de Beurs, K.M.; Julian, J.P.; Mayer, P.M. Identifying priority sites for low impact
development (LID) in a mixed-use watershed. Landsc. Urban Plan. 2015, 140, 29–41. [CrossRef]
69. Department of Irrigation and Drainage (DID). Urban Stormwater Management Manual for Malaysia; Department
of Irrigation and Drainage: Malaysia Kuala Lumpur, Malaysia, 2012.
70. Hunt, W.F.; Traver, R.G.; Davis, A.P.; Emerson, C.H.; Collins, K.A.; Stagge, J.H. Low impact development
practices: Designing to infiltrate in urban environments. In Effects of Urbanization on Groundwater: An Engineering
Case-Based Approach for Sustainable Development; American Society of Civil Engineers (ASCE): Reston, VA, USA,
2010; pp. 308–343.
71. Barrett, M.E.; Walsh, P.M.; Malina, J.F., Jr.; Charbeneau, R.J. Performance of vegetative controls for treating
highway runoff. J. Environ. Eng. 1998, 124, 1121–1128. [CrossRef]
72. Mancipe-Munoz, N.A.; Buchberger, S.G.; Suidan, M.T.; Lu, T. Calibration of rainfall-runoff model in urban
watersheds for stormwater management assessment. J. Water Resour. Plan. Manag. 2014, 140, 05014001.
[CrossRef]
 Water 2019, 11, 1415 23 of 23

73. Tan, S.B.; Chua, L.H.; Shuy, E.B.; Lo, E.Y.-M.; Lim, L.W. Performances of rainfall-runoff models calibrated
over single and continuous storm flow events. J. Hydrol. Eng. 2008, 13, 597–607. [CrossRef]
74. Fang, T.; Ball, J.E. Evaluation of spatially variable control parameters in a complex catchment modelling
system: A genetic algorithm application. J. Hydroinform. 2007, 9, 163–173. [CrossRef]

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).

View publication stats