Journal of Hydrology 612 (2022) 128179

Contents lists available at ScienceDirect

Journal of Hydrology
journal homepage: www.elsevier.com/locate/jhydrol

Research papers

Modeling bioretention hydrology: Quantifying the performance of

DRAINMOD-Urban and the SWMM LID module
W.A. Lisenbee a, c, *, J.M. Hathaway a, R.J. Winston b
a
University of Tennessee, Department of Civil and Environmental Engineering, 415 John D. Tickle Building, 851 Neyland Drive, Knoxville, TN 37996, USA
b
Ohio State University, Department of Food, Agricultural, and Biological Engineering and Department of Civil, Environmental, and Geodetic Engineering, 230B
Agricultural Engineering Building, 590 Woody Hayes Drive, Columbus, OH 43210, USA
c
Pennsylvania State University, Department of Ecosystem Science and Management, 415 Agricultural Sciences and Industries Building, University Park, PA 16802, USA

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

This manuscript was handled by Nandita Basu, Bioretention systems have become a leading infiltration-based Low Impact Development (LID) practice to reduce
Editor-in-Chief, with the assistance of Claire J. urban stormwater runoff volumes and peak flows. Although these systems have performed well in many site-
Oswald, Associate Editor scale field studies, modeling of bioretention systems has received less attention. Additional studies are needed
which calibrate various models to field measurements to investigate and optimize the performance of individual
Keywords:
LID practices and effectively scale local interventions to the watershed. DRAINMOD-Urban has been successfully
Biofilter
applied to bioretention at the site-scale due to its advanced soil–water accounting using the soil–water charac­
Rain garden
Stormwater teristic curve and its ability to explicitly model underdrains and internal water storage (IWS) zones. At the same
Green infrastructure time, the U.S. EPA Stormwater Management Model (SWMM) has become one of the most widely used urban
Low impact development drainage models. The latest version, SWMM5, included dedicated LID modules including a routine for bio­
Hydrologic modeling retention modeling.
In this study, DRAINMOD-Urban and the SWMM LID module were compared through detailed analysis of the
internal processes of each model as well as through model calibration and output investigation. The objective
was to identify the strengths and weaknesses of each model and compare the performance of both models to a
single bioretention cell. Both SWMM and DRAINMOD-Urban were evaluated in calibrated and uncalibrated
scenarios since urban drainage models often remain uncalibrated for planning scenario analysis. Following
calibration, DRAINMOD-Urban was superior for replicating drainage hydrographs (NSE = 0.60) while SWMM
produced better overflow hydrographs (NSE = 0.57). Specifically, SWMM often output a maximum drainage rate
that caused rectangular drainage hydrographs, but DRAINMOD-Urban was better able to match the shape of
measured drainage hydrographs. While the DRAINMOD-Urban model output was in good agreement with
measured drainage and overflow event volumes when calibrated (drainage NSE = 0.83, overflow NSE =
0.57–0.66), SWMM was closer to measured volumes even when uncalibrated (drainage NSE = 0.70–0.93,
overflow NSE = 0.59–0.81). This study improved existing knowledge of the SWMM LID module by calibrating to
field-collected data from a single bioretention cell for the first time in literature. Furthermore, the results of this
study indicate an opportunity for model coupling that could combine the strengths and weaknesses of each
model and improve bioretention cell modeling.

1. Introduction
Bioretention systems have been widely accepted within the storm­
water engineering community due to numerous, geographically diverse
field studies demonstrating substantial volumetric reductions and
generally good pollutant removal (Tirpak et al., 2021; Olszewski and
Davis, 2013; Brown and Hunt, 2012; Davis et al., 2012; Brown and Hunt,
2011; Davis, 2008; Hunt et al., 2006). However, most bioretention
studies have centered on field monitoring and laboratory assessments
(Liu et al., 2014; Davis et al., 2009). Computational models have been
slower to develop for bioretention systems despite the advantage they
provide in quantifying bioretention performance as a function of design

* Corresponding author at: Pennsylvania State University, Department of Ecosystem Science and Management, 415 Agricultural Sciences and Industries Building,
University Park, PA 16802, USA.
E-mail addresses: wlisenbee@psu.edu (W.A. Lisenbee), hathaway@utk.edu (J.M. Hathaway), winston.201@osu.edu (R.J. Winston).

https://doi.org/10.1016/j.jhydrol.2022.128179
Received 2 February 2022; Received in revised form 18 May 2022; Accepted 4 July 2022
Available online 11 July 2022
0022-1694/© 2022 Elsevier B.V. All rights reserved.
W.A. Lisenbee et al.
prior to implementation. Specifically, bioretention modeling is benefi­
cial to designers by informing bioretention cell design, optimizing de­
signs for local conditions and objectives, and scaling impacts from a
single cell to the larger watershed.
There are many hydrologic models that have developed tools or
routines for bioretention (Lisenbee et al., 2021). Low Impact Develop­
ment (LID) models may include bioretention as one of many LID prac­
tices but mostly examine lumped performance of LID installations
compared to traditional development under uncalibrated and calibrated
scenarios (Kaykhosravi et al., 2018; Li et al., 2017; Elliott and Trow­
sdale, 2007). Often these models focus on runoff volume reduction to
describe the performance of bioretention cells. However, hydrologic
pathways within a bioretention cell (drainage, overflow, exfiltration,
and evapotranspiration) should also be evaluated since they play a role
in determining the fate of pollutants in bioretention and impact water­
shed hydrology.
Unfortunately, hydrologic pathways are not well-represented in
many current bioretention models (Lisenbee et al., 2021). Underdrains
and restrictions to drainage, such as internal water storage (IWS) zones,
are not considered in some models and the hydrologic changes from
these features are computationally simplified in other models despite
widespread use in field applications (Lynn et al., 2018; Liu et al., 2014;
Brown and Hunt, 2011). Further, many models use infiltration processes
that assume uniform saturation of the media, while bioretention systems
are variably saturated and unsaturated during and following rain events
(Tu et al., 2020; Barbu and Ballestero, 2015; Akan, 2013; Brown et al.,
2013). Finally, the hydrologic flow paths in a bioretention model could
be calibrated to better represent field-measured drainage and overflow
from a bioretention cell but these studies have been limited to column
studies thus far (Gülbaz and Kazezyılmaz-Alhan, 2017; Liu and Fassman-
Beck, 2017). Additional studies are needed to determine how these
limitations in model processes may or may not affect the model
outcomes.
The U.S. EPA Stormwater Management Model (SWMM) is one of the
most widely used models for bioretention, especially since the release of
SWMM5 which includes dedicated LID modules (Rossman, 2010).
SWMM has many applications for catchment hydrology, provides
several methods for hydrologic and hydraulic processes, and reports
hydrographs from flows in sewers, open channels, or LID features.
SWMM has been widely applied to watershed studies investigating
lumped LID benefits compared to traditional stormwater management
(Bai et al., 2018; Avellaneda et al., 2017; Sun et al., 2014; McCutcheon
and Wride, 2013; Bosley, 2008). Fewer studies have calibrated SWMM
to observed total runoff reduction specifically provided by bioretention
at the watershed scale (Bai et al., 2018; Avellaneda et al., 2017; Li and
Lam, 2015; Rosa et al., 2015). However, bioretention models must first
be evaluated at the site scale to ensure they provide reliable estimations
of performance at the watershed scale.
Some studies have focused on performance of a single bioretention
cell (or rain garden) in SWMM compared to column or pilot-scale studies
(Gülbaz and Kazezyılmaz-Alhan, 2017; Liu and Fassman-Beck, 2017; Li
and Lam, 2015; McCutcheon and Wride, 2013). Only a few recent
studies have evaluated SWMM compared to bioretention cells installed
in field but were primarily focused on sensitivity analysis (Fassman-Beck
and Saleh, 2021; Platz et al., 2020). Thus, evaluation of the SWMM LID
module performance for site-scale bioretention studies is needed to
better understand how the model represents internal flow processes of a
bioretention cell.
Further studies have compared SWMM with other bioretention
models such as HYDRUS-1D (Tu et al., 2020; Lynn et al., 2018), Win­
SLAMM (Tiveron et al., 2018), HM-RWB (Gülbaz and Kazezyılmaz-
Alhan, 2017), RECARGA (Sun et al., 2011) and HydroCAD (Lucas,
2010). These studies have established that SWMM is capable of
modeling bioretention cells under a variety of conditions, but it has
limitations such as limited underdrain placement and design, simplistic
infiltration routines using a bucket model approach, and scarce
2
Journal of Hydrology 612 (2022) 128179
examples of calibration and validation against field LID practices
(Fassman-Beck and Saleh, 2021; Lisenbee et al., 2021). The research
herein combined important aspects of each of the aforementioned
studies: site-scale evaluation of the SWMM LID module, calibration of
hydrologic pathways in the SWMM LID module using field measured
data, and comparison of SWMM to another promising bioretention
model, DRAINMOD-Urban.
DRAINMOD is an agricultural drainage model that has performed
well when applied to urban bioretention systems by addressing some of
the limitations of other models related to infiltration modeling and
representation of the IWS zone (Winston, 2015; Hathaway et al., 2014;
Brown et al., 2013). Another advantage of DRAINMOD is the detailed
soil–water accounting procedure that estimates water level fluctuations
in the cell using the soil water characteristic curve (SWCC). The SWCC is
a better representation of available pore space in the bioretention cell at
various depths compared to the common assumption that storage ca­
pacity is simply the difference between water content at saturation and
field capacity (Brown et al., 2013). This is important in variably satu­
rated bioretention systems, especially those employing IWS zones which
elevate the internal water level. In addition to sophisticated soil-
moisture accounting, the SWCC is used in DRAINMOD for determining
the volume drained and the Green-Ampt infiltration parameters at
various internal water levels within the bioretention cell.
DRAINMOD-Urban is a version of DRAINMOD adapted for urban
hydrologic response times. DRAINMOD-Urban produces hydrographs at
time steps as small as one minute for each hydrologic pathway in the
bioretention cell (inflow, overflow, infiltration, drainage, exfiltration,
and ET). DRAINMOD-Urban has shown good prediction (NSE = 0.60) of
observed drainage hydrographs from a bioretention cell (Lisenbee et al.,
2020). While DRAINMOD-Urban appears to be well-suited for modeling
bioretention, it also has some disadvantages compared to SWMM.
SWMM has an easy-to-use graphical user interface (GUI) and graphing
functions for built-in visualization of outputs. Conversely, DRAINMOD-
Urban outputs text files that must undergo rigorous post-processing with
an external program. Furthermore, SWMM is a watershed-scale model
that can capture dynamics between catchments, sewers, hydraulic
structures, and other bioretention cells upstream and downstream of any
given bioretention cell. It also incorporates routing procedures to ac­
count for travel time and friction losses between these objects. At pre­
sent, DRAINMOD-Urban can only model a single bioretention cell. This
site-scale modeling allows users to evaluate the performance of a spe­
cific bioretention cell or design feature, whereas watershed-scale
modeling is useful for determining effects on a larger catchment.
Watershed-scale models for bioretention often lump parameters and
make simplifications to allow for quick analysis of many contributing
factors without cumbersome input requirements. To combine the best of
these methods, site-scale models can be incorporated as add-in tools for
larger watershed models.
This study addressed the need to understand the performance of the
SWMM LID module in representing an individual bioretention cell by
calibrating SWMM to field-measured bioretention cell hydrology. The
resulting model performance was compared with that of DRAINMOD-
Urban. Evaluating hydrographs produced by each model was impor­
tant to consider the effect bioretention has on flow dynamics of urban
watersheds, combined sewer overflow (CSO) mitigation, and treatment
trains that consist of multiple LID strategies which require accurate
hydrographs (Liu and Fassman-Beck, 2017). This study identified
strengths and weaknesses of each model’s ability to represent event
volumes and hydrographs of drainage and overflow pathways in a bio­
retention cell. Acknowledging advantages and applications best suited
to each model bears potential for combining these strengths in future
bioretention modeling efforts.
W.A. Lisenbee et al.
2. Material & methods
2.1. Site description
A single site-scale model was created in both DRAINMOD-Urban and
SWMM and calibrated to the Ursuline College (UC) bioretention cell
located near Cleveland, Ohio, USA. This site was used for previous
evaluations of DRAINMOD and DRAINMOD-Urban performance
(Lisenbee et al., 2020; Winston, 2015). A 3600 m2 drainage area
comprised largely of parking lot (77% impervious) produced storm­
water runoff that entered the UC cell which had a surface area of 182 m2
(Fig. 1). The bioretention media layer (60 cm) in the UC cell was 87%
sand, 4% silt, and 9% clay according to sieve analysis (ASTM, 2007). The
UC cell had a 30 cm ponding zone that was topped with a 7.5 cm layer of
mulch. Below the bioretention media was a choking sand layer (15 cm)
and a gravel storage layer (30 cm) which contained an underdrain with
an upturned elbow, creating a 60 cm deep IWS zone (Fig. 2). The bio­
retention media was not expected to experience clogging or compaction
over the monitoring period due to recent construction, a forebay to
collect fine sediment entering the cell, and the growth of vegetation and
roots.
Precipitation was measured on-site with a tipping bucket rain gauge
at 1-minute intervals. Inflow to the UC cell was unable to be directly
measured on-site so catchment properties were entered into SWMM to
model an inflow hydrograph that was used in both SWMM and
DRAINMOD-Urban models. The combined drainage and overflow were
measured with a 60-degree, sharp-crested, v-notch weir, and a bubbler
flow module attached to an ISCO 6712 sampler that collected data every
two minutes (Fig. 1). Drainage and overflow were separated in SWMM
using a rating curve based on the internal water level (Winston et al.,
2016). A total of 12 drainage events and four overflow events were
measured over the seven-month monitoring period. Additionally, volu­
metric soil moisture was measured in the UC cell using time domain
reflectometer (TDR) probes at 15, 30, 60, and 90 cm depths (Fig. 1). The
15 and 30 cm depths described the bioretention media whereas the 60
and 90 cm depths described moisture in the IWS and sand/gravel storage
zones.
2.2. Model description
Both DRAINMOD-Urban and SWMM are long-term, continuous
Fig. 1. The inflow, outflows, and monitoring equipment for the Ursuline College (UC) bioretention cell (blue area) and its drainage area from a parking lot (green).
(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
3
Journal of Hydrology 612 (2022) 128179
simulation models which account for antecedent moisture conditions.
SWMM also has the benefit of being applicable to the larger urban
watershed, employing routing procedures such as the kinematic wave
equation which accounts for runoff travel time to the bioretention cell.
The routing procedures in SWMM also allow for outflow to be routed to
another LID component or pipe network downstream.
To allow direct comparison to DRAINMOD-Urban, the bioretention
LID module in SWMM was evaluated as its own subwatershed. This
reduced the SWMM model to a single bioretention cell similar to
DRAINMOD-Urban and provided valuable analysis of the capabilities of
the LID module for bioretention cell modeling. Before modeling these
bioretention cells in SWMM, the governing equations used for various
hydrologic pathways in SWMM and DRAINMOD-Urban were compared
in the following sections. For specific model inputs, see Table 4. More
information on model processes can be found in model documentation
(Rossman and Huber, 2016; DRAINMOD 6.1 Help File, 2013; Skaggs
et al., 2012) and the bioretention modeling review paper (Lisenbee
et al., 2021).
2.2.1. Inflow
SWMM was used to model a continuous runoff hydrograph from the
catchment area at 1-min intervals. For DRAINMOD-Urban, this runoff
hydrograph was entered as inflow to the bioretention cell without
modification. In the SWMM simulations, the same method was used to
determine bioretention inflow with the addition of a potential evapo­
transpiration (PET) file using the Penman-Monteith method which
reduced the inflow to the bioretention cell by 10%.
2.2.2. Overflow
DRAINMOD-Urban has a surface storage input that represents the
bioretention ponding zone. If ponding exceeds this level, then overflow
is equal to the sum of direct precipitation on the bioretention cell and
inflow minus infiltration. SWMM includes a ponding zone (denoted
berm height) in the LID module for bioretention but also includes a
vegetation volume fraction to account for the space that vegetation
occupies in the ponding zone. Overflow is calculated in SWMM as the
water level above a maximum freeboard for each given timestep.
2.2.3. Evapotranspiration
Both DRAINMOD-Urban and SWMM accept user-input PET but
DRAINMOD-Urban can also calculate PET using the Thornthwaite
W.A. Lisenbee et al.
Fig. 2. Comparison of the drainage configuration used in DRAINMOD-Urban (a) and SWMM (b) to represent the internal water storage zone (IWS) present in the
Ursuline College (UC) bioretention cell. *Image
adapted from Lisenbee et al., 2020.
method. In DRAINMOD-Urban, PET is used to represent ET from the
system when soil water is not limiting. If ET is limited by soil water
conditions, then ET is equal to the upward flux of water as a function of
internal water level (determined from the SWCC soil input). When this
upward flux is not enough to meet ET demand, water is removed from
the root zone.
In SWMM, ET is calculated for the surface, soil, and storage layers
consecutively such that any remaining PET is available to the subse­
quent layer. The surface ET is calculated as the minimum of the PET and
the ponding depth at a given timestep. The ET from the soil layer is
calculated from the moisture content and the depth of the soil. Both the
4
Journal of Hydrology 612 (2022) 128179
soil and storage ET are set to zero when surface infiltration is occurring.
2.2.4. Infiltration
The Green-Ampt method was used to represent surface infiltration
through the bioretention cell in SWMM and DRAINMOD-Urban. This
method is a simplification of Richards’ equation which employs certain
assumptions such as one-dimensional, vertical flow and total saturation
behind a sharp wetting front. These assumptions are not always valid in
bioretention cells which operate under variably saturated and unsatu­
rated conditions. For the LID module in SWMM, the Green-Ampt equa­
tion was adjusted to account for ponding depth which is an important
W.A. Lisenbee et al.
component of bioretention cells.
DRAINMOD-Urban requires the user to input the SWCC of the bio­
retention media to account for variable soil moisture; this is especially
important for cells with an IWS zone where the internal water level may
be close to the surface. The SWCC provides a more detailed relationship
between moisture content fluctuations and depth to the internal water
level than traditional methods of assuming an initial moisture deficit
(which is the procedure used in SWMM; Brown et al., 2013). This was
especially true when the water level was close to the surface as it often is
in a bioretention cell during a storm event. For SWMM, infiltration ca­
pacity recovery is modeled using the hydraulic conductivity in Green-
Ampt instead of a soil-accounting method like DRAINMOD-Urban
(Bosley, 2008). SWMM assumes all soil moisture is evenly distributed
throughout the soil layer (Rossman and Huber, 2016), which may not be
an appropriate assumption (Barbu and Ballestero, 2015).
DRAINMOD-Urban and SWMM require different inputs although
they both employ the Green-Ampt equation (Table 4). To estimate
infiltration parameters, the SWMM reference manual provides a table of
suggested bioretention soil parameters based on sand, loamy sand, and
sandy loam soil textures using the SPAW model (Rossman and Huber,
2016; Saxton and Rawls, 2006). In DRAINMOD, infiltration parameters
are derived automatically from the measured SWCC entered in the
model, but they can be adjusted manually if desired (Skaggs et al.,
2012). Overall, deriving infiltration parameters from a measured SWCC
is hypothesized to improve infiltration estimation in DRAINMOD-Urban
compared to SWMM which does not require sophisticated soil charac­
teristic parameterization but instead relies on empirical estimation
based on soil texture.
2.2.5. Percolation
After the initial surface infiltration, percolation refers to flow
through soil layers within the bioretention cell. DRAINMOD-Urban
employs the Green-Ampt parameters derived from the SWCC and satu­
rated hydraulic conductivity (Ksat) for each soil layer in the model are
used to find an effective Ksat (Keff) at each time step based on the internal
water table depth. In SWMM, Darcy’s law is used to simulate percolation
through subsequent soil layers in the bioretention cell. This is applied in
the same manner as SWMM’s groundwater routine using the coefficient,
HCO, to describe the exponential decrease in hydraulic conductivity
with decreasing moisture content (Rossman and Huber, 2016). The
equation used for the percolation rate through each soil layers is:
f2 = Ksat e(− HCO(ϕ− θ))
(1)
Fig. 3. Schematic of DRAINMOD water balance including variables for the Hooghoudt drainage equation. .
Adapted from Lisenbee et al., 2020
5
Journal of Hydrology 612 (2022) 128179
where Ksat is the saturated hydraulic conductivity of the soil layer,
HCO is the conductivity slope, ϕ is the soil porosity and θ is the soil
moisture. If the native soil layer is saturated, the percolation is limited to
the saturated hydraulic conductivity of the bioretention media. If the
moisture content of the soil layer falls below field capacity, then the
percolation rate is set to zero (Rossman and Huber, 2016).
2.2.6. Drainage
DRAINMOD-Urban represents drainage in bioretention systems
(Fig. 3) using the Hooghoudt equation (Eqn. (2)) that accounts for flow
convergence near drains:
q = 4Ke m(2de + m)/L2 (2)
where q is the drainage rate (cm/h), m is the midpoint water level
above the drain, Ke is the equivalent lateral hydraulic conductivity of the
profile (cm/h), de is the equivalent depth from the drain to the restric­
tive layer (cm), and L is the drain spacing (cm).
DRAINMOD-Urban’s soil-moisture accounting is also used to esti­
mate the change in internal water level. When water begins to pond on
the soil surface and the soil is fully saturated, the Kirkham equation
(1957) is used to calculate drainage instead of the Hooghoudt equation
(Skaggs et al., 2012). However, the drainage rate could be limited by
other hydraulic constraints which are accounted for through the
drainage coefficient parameter which sets the maximum drainage ca­
pacity. DRAINMOD-Urban also provides a “controlled drainage” option
that uses weir settings to control drawdown in the bioretention cell and
represent an IWS zone, a common design in practice.
SWMM can model underdrains in the LID module but the drainage
inputs required are not measurable quantities, and the drain advisor that
assists users in determining the correct values for these inputs is limited
in scope (EPA SWMM Help File 5.1, 2017). According to the SWMM LID
user manual, SWMM models the underdrain using an empirical power
law weir equation unless the maximum drainage rate is reached:
q = CD (h)η (3)
Where q is the drainage rate (ft/s), h is the hydraulic head (ft), CD is
the underdrain discharge coefficient, and η is the underdrain discharge
exponent (Rossman and Huber, 2016). The hydraulic head on the
underdrain is defined by the offset height, which is an input in the LID
module; drainage does not begin until the water level in the storage
layer reaches the drain offset height. The maximum drainage limit is
equal to the drainage coefficient (as represented in DRAINMOD-Urban)
if the exponent is set to zero. If the underdrain is not limiting the
W.A. Lisenbee et al.
hydraulics, the drainage coefficient can be any number larger than the
Ksat. In this case, the drainage rate is equal to the rate of percolation
entering the underdrain from the adjoining soil layer minus the seepage
rate as long as the maximum drainage limit (represented by the drainage
coefficient) is not reached.
Other drainage restrictions can be incorporated by adjusting the
drainage exponent to 0.5 to represent the standard orifice equation
(Rossman and Huber, 2016). Under this scenario, the drainage coeffi­
cient can be used to represent slotted pipes (that act as orifices) or an
orifice at the outlet of the underdrain. These adjustments to Equation (3)
offer more flexibility in how the drainage hydraulics are calculated by
considering multiple drains, valves, cap orifices, slotted pipes, etc.
which are not considered in many bioretention models. However, an
upturned elbow (for creation of an IWS zone) is one drainage configu­
ration that is still not readily modeled in the SWMM LID module. One
study has shown how to represent an elevated outlet in SWMM by
creating a rating curve using HYDRUS to represent both unsaturated
zones and saturated sections of the bioretention media (Lynn et al.,
2018). However, this study was conducted using the original SWMM
framework.
Ours is the first published study to model an IWS zone in the SWMM5
LID module using the following model setup. In the SWMM LID module,
the soil layer (or bioretention media) can only be placed above the
storage layer and the underdrain is typically placed at the top of the
storage layer. In this study, for comparison to DRAINMOD-Urban, the
cell configuration had to be adjusted in SWMM to account for storage
available in the IWS zone. The storage layer depth was input as the
depths of gravel and sand layers (45 cm), and the depth of bioretention
media remained at 60 cm. For this study, the underdrain was placed at
the top of the IWS zone (60 cm above the bottom of cell) to accommo­
date the proper underdrain invert elevation. Therefore, the underdrain
was placed in the soil layer 15 cm above the storage layer to match as-
built conditions (Fig. 2).
This representation of the IWS zone in SWMM is simplified to fit
within the constraints of the LID module, i.e. using an elevated under­
drain instead of an upturned elbow configuration. We anticipate
increased use of SWMM for IWS bioretention designs because IWS is
becoming a standard design in some states, and SWMM is a prominent
bioretention modeling tool (NCDEQ, 2020). Studying model behavior
with this IWS setup will demonstrate the best possible model outcomes
using the model as is and highlight areas for improvement in the LID
module. Updating the SWMM model to address IWS is paramount to
broaden the scope of bioretention practices that can be modeled using
SWMM.
2.2.7. Exfiltration
In SWMM, exfiltration from the bottom of the storage layer into the
underling soil is simply set to a user-supplied saturated hydraulic con­
ductivity of the underlying soil. DRAINMOD-Urban uses Darcy’s law
with the Dupuit-Forchheimer assumptions to calculate vertical seepage
at the transition from the storage layer to the underlying soil.
DRAINMOD-Urban also requires seepage inputs such as the piezometric
head of the aquifer, the thickness of the restricting layer and the vertical
conductivity of the restricting layer, which are often used as calibration
parameters since they are difficult to measure.
2.3. Modeling methods
2.3.1. Inputs
Understanding model inputs aids the user in determining if the
model utilizes mechanistic or empirical functions to estimate bio­
retention hydrology. Furthermore, the complexity and number of
required inputs can discourage use of a model especially for designers
that are less familiar with hydrologic modeling or have limited access to
descriptive data for the system (Lisenbee et al., 2021). The right balance
must be struck between accurately representing the system and reducing
6
Journal of Hydrology 612 (2022) 128179
the effort required to measure or estimate inputs.
SWMM requires fewer inputs than DRAINMOD-Urban, but more
inputs are simply estimated with guidance from the model or left as
defaults (Table 1). We described “estimated” inputs as those using model
defaults or documentation whereas “measured” inputs come from field
or laboratory testing of site-specific properties. Soil properties play a big
role in both bioretention models. In SWMM, a total of 15 inputs are
required for the LID module and six of these are soil parameters
(Table 4). DRAINMOD-Urban has fewer estimated parameters than
SWMM but requires more detailed collection of soil characteristics such
as the SWCC and hydraulic conductivity of each layer which require
extensive laboratory testing (Tables 1 & 4).
2.3.2. Calibration parameters
The calibration parameters used in DRAINMOD-Urban were the
saturated hydraulic conductivity (Ksat) of the bioretention media, Ksat of
the sand layer, the drainage coefficient, the piezometric head of the
aquifer, and the thickness of the restricting layer (Table 2). The cali­
bration parameters used in SWMM were Ksat and soil suction of the
bioretention media, the conductivity slope (HCO), the void ratio of the
storage layer and the seepage rate (Table 2). The Ksat was sensitive in
both DRAINMOD-Urban and SWMM to the amount of infiltration
allowed through the bioretention cell. The void ratio for the storage
layer and the soil suction also affected the infiltration in SWMM but
were not as sensitive as Ksat. The parameters that controlled the exfil­
tration from the bioretention cell were the piezometric head of the
aquifer and the thickness of the restricting layer in DRAINMOD-Urban
and the seepage rate in SWMM.
There were two unique calibration parameters: the drainage coeffi­
cient and the HCO parameter. The initial drainage coefficient in
DRAINMOD-Urban was increased to avoid restricting drain flow since
measured hydrographs did not show any evidence of drainage restric­
tion. For comparison, the same was done in SWMM. The HCO parameter
is suggested to fall in the range of 30 to 60 based on its soil texture (EPA
SWMM Help File 5.1, 2017). Several studies have shown much smaller
HCO values than suggested by SWMM documentation in bioretention
simulations (Lynn et al., 2018; Liu and Fassman-Beck, 2017; Rosa et al.,
2015). Therefore, although the HCO was calculated to be 49.4 based on
SWMM recommendations, several smaller HCO values (ranging from 7
to 60) were tested during calibration.
2.3.3. Calibration process
When calibrating both models to measured drainage and overflow
hydrographs, it was important to simultaneously evaluate the effect of
calibration on event and total volumes. Sometimes these processes
couple such that calibrating hydrographs also improves event volumes.
However, an attempt to match volumes sometimes distorts the hydro­
graph shape (Lisenbee et al., 2020). Therefore, calibrations were per­
formed to optimize fit to either hydrographs (hydrograph-calibrated) or
volumes (volume calibrated) in DRAINMOD-Urban and SWMM. Results
for model performance for matching both hydrographs and volume were
recorded regardless of the output being optimized. An additional
assessment was completed to document performance of each model in
an uncalibrated state, which is the way bioretention models are
commonly used in practice for preliminary design analysis due to
limited data availability or modeling experience. We believe that the
Table 1
Summary of inputs required for both SWMM and DRAINMOD-Urban (DM-
Urban) for the Ursuline College (UC) bioretention cell (BRC).
Summary of Inputs DRAINMOD-Urban SWMM
Total Required Inputs (minus inflow) 26 24
Total Estimated 12 14
Total Measured 14 10
Total Calibration Parameters 5 5
W.A. Lisenbee et al. Journal of Hydrology 612 (2022) 128179

Table 2
Comparison of calibration parameters for uncalibrated, volume-calibrated, and hydrograph-calibrated simulations using SWMM and DRAINMOD-Urban (DM-Urban)
for the Ursuline College (UC) bioretention cell.
Calibration PIEZO. HEAD* RESTRICT. LAYER* KSAT-sand KSAT-media DRAIN HCO VOID* SOIL SUCTION SEEP.* (cm/
Method (cm) (cm) (cm/hr) (cm/hr) COEFF. (cm) hr)

DM-Urban
Uncalibrated 53 55 15 16 25
Volume-Calibrated 23 26.5 30 17 120
Hydrograph- 12 20 45 35 300
Calibrated
SWMM
Uncalibrated 16.8 300 49.4 0.51 6.10 0.44
Volume-Calibrated 13.3 300 23.0 0.30 25.4 1.3
Hydrograph- 16.8 300 23.0 0.30 11.0 0.44
Calibrated

*PIEZO. Head = Piezometric head, RESTRICT. LAYER = thickness of the restricting layer, VOID = void ratio of the storage layer, SEEP. = Seepage.

uncalibrated scenario was valuable to provide insight on the model
performance in cases where calibration data may be constrained.
This process resulted in three calibration states for each model: un­
calibrated, volume-calibrated, and hydrograph-calibrated. Optimization
of modeled volumes and hydrographs was determined using the Nash-
Sutcliffe Efficiency (NSE) and percent bias (PBIAS) as goodness-of-fit
tests. Other statistics such as the index of agreement (d), its relative
error counterparts (d1) and the relative NSE (E1) were calculated as
supplements to the NSE and PBIAS during calibration (Moriasi et al.,
2007). After these statistics were used to narrow the number of viable
scenarios, drainage and overflow hydrographs were visually inspected,
and characteristics such as peak flow, time to peak, and event duration
were compared. These assessments aided in determining a final
hydrograph-calibrated simulation for the UC cell.
3. Results & discussion
3.1. DRAINMOD-Urban performance
3.1.1. Hydrographs
When calibrating DRAINMOD-Urban to measured hydrographs
(hydrograph-calibration), good performance was achieved with NSE =
0.60 for drainage hydrographs at a 2-minute timestep (Table 3). When
the UC cell was calibrated to achieve the best fit to measured drainage
and overflow volumes (volume-calibration), the hydrograph perfor­
mance was reduced to NSE = 0.31, even smaller than NSE = 0.39 for the
uncalibrated simulation (Table 3). According to Skaggs et al. (2012), an
NSE greater than 0.4 is acceptable model performance at a daily time­
step. Thus, for a much smaller 2-minute timestep, both the uncalibrated
and volume-calibrated drainage hydrographs could have acceptable
model performance at 0.39 and 0.31 NSE. For overflow hydrographs,
however, all calibrations performed poorly. The lower NSE statistics for
overflow were likely due to the emphasis on matching peak flow by the
NSE statistic (Broekhuizen et al., 2020; Lisenbee et al., 2020; Krause
et al., 2005). For instance, modeled timing and duration of overflow
were very close to measured, but peak flow was overestimated in all
DRAINMOD-Urban simulations (Fig. 5).
Visualization of drainage and overflow hydrographs is a key step in
evaluating DRAINMOD-Urban performance (Saraswat et al., 2015;
Moriasi et al., 2007). The performance of DRAINMOD-Urban drainage
hydrographs varies depending on the calibration method. The volume-
calibration had the worst drainage hydrograph fit based on visual
assessment of a representative storm event (Fig. 4c). This simulation had
a truncated peak that was much lower than measured, and the duration
of the first two peaks was extended to compensate for this loss in vol­
ume. Although the peak and duration of the uncalibrated drainage
hydrograph was still poorly represented, the shape of the hydrograph
was better than the rectangular shape observed under the volume-
calibration (Fig. 4a). Also, notably, in both the uncalibrated and
volume-calibrated simulations, the third peak was not detected for the
event depicted in Fig. 4. The hydrograph-calibrated drainage hydro­
graph for this representative event was visually improved in multiple
ways. The peak drainage was closer to measured (albeit still under­
estimated), the timing of rising and falling limbs was much closer to
measured, and the third peak corresponded well to measured timing and
peak drainage (Fig. 4e). Drainage hydrographs clearly benefited from
hydrograph-calibration in DRAINMOD-Urban. For overflow hydro­
graphs, the improvements across calibrations were less evident. In
general, the peaks were reduced, and the timing was improved under
hydrograph-calibration. This is demonstrated in the representative
overflow event, particularly for the second smaller peak (Fig. 5a, c, e).

Table 3
Goodness-of-fit tests for measured drainage and overflow volumes and hydrographs compared to uncalibrated, volume-calibrated, and hydrograph-calibrated sim­
ulations in DRAINMOD-Urban (DM-Urban) and SWMM.
Goodness of Fit Tests

NSE PBIAS NSE PBIAS d d1 E1

DM-Urban Volumes Hydrographs

Uncalibrated Drainage 0.50 41.3 0.39 − 1.62 0.88 0.60 0.30
Overflow − 1.77 72.8 − 1.60 − 73.01 0.38 0.64 − 0.06
Volume-Calibrated Drainage 0.83 –23.4 0.31 16.7 0.86 0.60 0.32
Overflow 0.57 − 58.4 − 1.82 − 96.8 0.37 0.64 − 0.01
Hydrograph- Calibrated Drainage 0.83 − 46.6 0.60 5.19 0.93 0.75 0.52
Overflow 0.66 67.0 − 0.10 − 18.5 0.68 0.74 0.36
SWMM Volumes Hydrographs
Uncalibrated Drainage 0.86 29.5 0.34 − 28.76 0.92 0.76 0.46
Overflow 0.70 24.9 0.54 24.87 0.85 0.76 0.47
Volume-Calibrated Drainage 0.93 16.6 0.47 − 6.29 0.92 0.77 0.52
Overflow 0.81 16.9 0.12 − 0.26 0.73 0.74 0.38
Hydrograph- Calibrated Drainage 0.70 43.1 0.42 − 42.90 0.92 0.75 0.45
Overflow 0.59 30.7 0.57 30.67 0.86 0.76 0.48

7
W.A. Lisenbee et al. Journal of Hydrology 612 (2022) 128179

Table 4
Comparison of initial uncalibrated inputs required for both SWMM and DRAINMOD-Urban (DM-Urban) for the Ursuline College (UC) bioretention cell (BRC). Cali­
bration parameters are denoted and the change in these parameters can be found in Table 2.
Input Parameters DRAINMOD-Urban SWMM Measured (M) or Calibration
Estimated (E) Parameters

Catchment Geometry Catchment Area (m2) 3600 M

Properties
Catchment Width (m) 43.6 M
% Slope 4.49 M
Land Cover % Imperviousness 77 M
Impervious Manning’s n 0.01 E
Pervious Manning’s n 0.1 E
Depression Storage (cm) 0.127 E
% of Impervious Area w/o Depression 25 E
Storage
Runoff User-defined Inflow From SWMM n/a
BRC Design Geometry BRC Area (m2) 182 M
Properties
BRC Area:Catchment Area Ratio 19.7 M
BRC Width (m) 7.01 M
Ponding Ponding Layer Depth (cm) 30 30 M
Layer
Surface Slope 0 E
Drainage Underdrain Diameter (cm) 10 M
Underdrain Flow (Drainage) Coefficient 25 300 E DM-Urban
Underdrain Flow Exponent 0 E
Underdrain Offset Height (cm) 60 E
Depth from Soil Surface to Drain (cm) 107 M
Drain Spacing (cm) 597 E
Weir Depth from Soil Surface (for IWS 52 M
zone)
BRC Media BRC Media Depth (cm) 60 45 M
BRC Surface Roughness (Manning’s) 0.1 E
Initial Water Table Depth (cm) 112.5 E
Input Parameters DRAINMOD-Urban SWMM Measured (M) or Calibration
Estimated (E) Parameters
BRC Design Storage Storage Layer Depth (cm) 45 60 M
Properties Layer Storage Void Ratio 0.514 E SWMM
Clogging Factor 0 E
Piezometric Head (cm) 53 E DM-Urban
Thickness of Restricting Layer (cm) 55 E DM-Urban
Seepage Rate (cm/hr) 0.437 0.437 M SWMM
Climate Precipitation 1-min measured M
data
Temperature Daily min/max M
Heat Index 50 E
User-input PET n/a Penman- E
Monteith
Soil BRC Media Soil Texture Loamy Sand E
Porosity 0.331 0.331 M
Initial Moisture Deficit 0.154 E
Field Capacity Moisture Content 0.177 0.177 M
Wilting Point Moisture Content 0.02 0.047 E
Suction Head (cm) 6.10 E SWMM
Saturated Hydraulic Conductivity (cm/ 16 16.8 M DM-Urban/ SWMM
hr)
Soil-Water Characteristic Curve required M
Other Saturated Hydraulic Conductivity of 60 E
Mulch
Saturated Hydraulic Conductivity of 15 E DM-Urban
Sand
Saturated Hydraulic Conductivity of 200 E
Gravel
Hydraulic Conductivity Slope (HCO) 49.4 E SWMM
Vegetation Root Layer Depth (cm) 30 E
Vegetation Volume Fraction 0.01 E

3.1.2. Volumes The hydrograph-calibrated DRAINMOD-Urban model described in

Higher NSEs were observed for event volumes than for hydrographs
for all calibration sets except for the uncalibrated overflow which
showed poor performance for both volumes and hydrographs (Table 3).
This was expected because event volumes were summed over the entire
event as opposed to hydrographs which were evaluated at each 2-minute
interval (Moriasi et al., 2007). Small shifts in the modeled hydrograph
compared to the measured hydrograph led to disparities in the hydro­
graph NSE but may have had little impact on the modeled volume.
Lisenbee et al. (2020) achieved an NSE = 0.83 for drainage event vol­
umes. The NSE of drainage volumes under the volume-calibration was
actually the same as the hydrograph-calibration (0.83) but with slightly
better PBIAS. The uncalibrated DRAINMOD-Urban simulation has the
worst performance for drainage and overflow event volumes among the
three calibration methods (Table 3). Drainage and overflow volumes
were mostly overestimated when DRAINMOD-Urban was uncalibrated
except for a few storms with the greatest measured volumes (Fig. 6a &

8
W.A. Lisenbee et al. Journal of Hydrology 612 (2022) 128179

Fig. 4. A representative drainage event was used to demonstrate modeled and measured drainage hydrographs from both DRAINMOD-Urban (a, c, e) and SWMM (b,
d, f) under three simulations: uncalibrated (a, b), volume-calibrated (c, d), and hydrograph-calibrated (e, f).

b). Overflow event volumes in DRAINMOD-Urban showed good per­
formance under the volume and hydrograph calibrations even when
overflow hydrograph performance was poor (Table 3). Unexpectedly,
overflow event volumes under the volume-calibration had a lower NSE
compared to the hydrograph-calibration. However, both volume and
hydrograph calibrations substantially improved drainage and overflow
volumes suggesting that, for DRAINMOD-Urban, calibration is highly
recommended. Both calibration techniques (i.e., volume or hydrograph
calibrated) yielded reasonable fit between drainage volumes and
measured volumes, though they were still slightly overestimated (Fig. 6c
& e). With only four overflow events for comparison, it was hard to
detect trends in overflow volumes across calibration methods (Fig. 6).
3.2. SWMM model performance
3.2.1. Hydrographs
The uncalibrated SWMM model simulated overflow hydrographs
(NSE = 0.54) almost as well as SWMM under hydrograph-calibration
(NSE = 0.57). Drainage hydrographs showed greater improvement
with hydrograph-calibration (NSE = 0.42) from an uncalibrated model
(NSE = 0.34; Table 3). However, when visually examining drainage
event hydrographs, the difference between the uncalibrated (b) and
hydrograph-calibrated (f) simulations were less obvious, with the NSE
likely improving due to improved prediction of drainage timing (Fig. 4).
When Ksat and seepage rate were adjusted in SWMM to better match
drainage and overflow volumes (volume-calibration), the maximum

9
W.A. Lisenbee et al. Journal of Hydrology 612 (2022) 128179

Fig. 5. A representative overflow event was used to demonstrate modeled and measured overflow hydrographs from both DRAINMOD-Urban (a, c, e) and SWMM (b,
d, f) under three simulations: uncalibrated (a, b), volume-calibrated (c, d), and hydrograph-calibrated (e, f).

drainage rate was underestimated, unlike with other calibration
methods (Fig. 4). This change was not represented by the cumulative
NSE which only changed from 0.42 to 0.47 from hydrograph-calibration
to volume-calibration.
One noticeable characteristic of SWMM drainage hydrographs is that
many reach a maximum drainage limit where the peak drainage is
truncated. The peak drainage rate could be adjusted in SWMM by
modifying the drainage coefficient, Ksat, or seepage rate, but these
changes had no effect on the rectangular shape. This behavior is
explained by the percolation equation in SWMM (Eqn. (1)). When the
water level exceeds the top of the storage layer and the soil moisture of
the bioretention media is between field capacity and saturation, the
drainage can be calculated using the HCO parameter (Eqn. (1)). This
explains the sensitivity of this parameter in SWMM despite an ambig­
uous effect on event hydrographs. When the soil layer is saturated, the
percolation rate (Eqn. (1)) is equal to the Ksat and the drainage rate
becomes the Ksat minus the seepage rate.
This effect was observed in eight out of 12 event hydrographs
simulated, but when examining soil moisture measured in the bio­
retention cell over the course of the monitoring period, the soil layer was
rarely fully saturated. The measured soil mositure remained near field
capacity during the majority of time the bioretention cell was draining.
The maximum measured moisture content in the bioretention media was
0.312 cm3/cm3 and in the IWS/storage layer was 0.354 cm3/cm3
compared to a measured porosity of 0.331 cm3/cm3 for the bioretention
media (Supplementary material). One explanation for the lack of

10
W.A. Lisenbee et al. Journal of Hydrology 612 (2022) 128179

Fig. 6. Measured and modeled drainage (a, c, e) and overflow (b, d, f) volumes from both DRAINMOD-Urban (red triangle) and SWMM (blue circle) under three
simulations: uncalibrated (a, b), volume-calibrated (c, d), and hydrograph-calibrated (e, f). (For interpretation of the references to colour in this figure legend, the
reader is referred to the web version of this article.)

saturation in the UC bioretention cell was that it was oversized
compared to typical design recommedations in the region, fully treating
29.5 mm in the ponding zone without overflow (Winston et al., 2016). In
a related study, the SWMM-modeled soil moisture reached saturation
multiple times in the 95% predictive uncertaintiy bounds using Monte
Carlo anlysis but the extent of observed soil moisture values across seven
bioretention cell never reached saturation (Avellaneda et al., 2017).
Another explanation is the potential for preferential flow pathways that
are not modeled but could drain the bioretention cell faster than
assumed with an average soil moisture (Liu and Fassman-Beck, 2017;
Carpenter and Hallam, 2010).
Lastly, the cumulative NSE for overflow hydrographs dropped sub­
stantially under the volume-calibration compared to the hydrograph-
calibration and uncalibrated simulations, but other goodness-of-fit
tests do not demonstrate as drastic a decline in performance (Table 3).
Again, this is likely due to the NSE being highly influenced by peak
matching. The representative event overflow hydrograph produced by
SWMM in Fig. 4 (b, d, f) do not considerably vary across calibration
methods except for a larger second peak in the volume-calibration which
explained some of the change in NSE.
3.2.2. Volumes
Statistical evaluation of model fit (Table 3) demonstrates impressive
performance from SWMM in estimating event volumes even when un­
calibrated. All SWMM simulations had an NSE for drainage and overflow
volumes greater than 0.59 which is considered satisfactory model

11
W.A. Lisenbee et al. Journal of Hydrology 612 (2022) 128179

performance at a monthly time step and even stronger model perfor­
mance at the event level (Moriasi et al., 2007). However, the uncali­
brated SWMM simulation did appear to underestimate event overflow
volumes (Fig. 6) which was also found in a study where an uncalibrated
SWMM model underestimated weekly surface runoff of an LID water­
shed (Rosa et al., 2015).
Drainage and overflow event volumes had the best performance
(NSE = 0.93 and NSE = 0.81, respectively) when SWMM was calibrated
specifically for event volumes (as opposed to hydrographs). Similarly, a
previous study of a pilot-scale bioretention system found excellent
model performance (NSE = 0.97) for SWMM overflow volumes over a
range of rainfall depths (58 events, approx. 0.3–100 mm) (Li and Lam,
2015). The NSE for event drainage volumes decreased from 0.86 (un­
calibrated) to 0.70 under hydrograph-calibration of the SWMM model
due to overestimation (Table 3, Fig. 6a and e). Similarly, the NSE of
overflow event volumes was reduced from an uncalibrated 0.70 to 0.59
under hydrograph-calibration. For all simulations, the measured
drainage and overflow volumes closely matched the SWMM model
outputs as seen by the proximity to the 1:1 line in Fig. 6.
3.3. Comparison of SWMM and DRAINMOD-Urban
Both SWMM and DRAINMOD-Urban performed reasonably for
modeling the UC bioretention cell, but different advantages existed for

Fig. 7. Decision flowchart to determine which model and calibration method is better suited to different modeling objectives with relative NSE model perfor­
mance ratings.

12
W.A. Lisenbee et al.
each model. A decision flowchart was created based on this study to
signify model performance under different model objectives (Fig. 7).
The following sections describe advantages for each model when eval­
uating event volumes and hydrographs.
3.3.1. Hydrograph performance
With respect to reproducing drainage hydrographs, a hydrograph-
calibrated DRAINMOD-Urban (NSE = 0.60) performed better than a
hydrograph-calibrated SWMM (NSE = 0.42). DRAINMOD-Urban also
better mimicked drainage hydrograph shape by visual inspection which
was attributed to better timing and response during peak drainage
(Fig. 4). SWMM produced rectangular drainage hydrographs (i.e.,
truncated, elongated.
duration peaks) when the bioretention cell was saturated which
occurred more often in SWMM than observed in soil moisture mea­
surements collected in the UC cell.
Conversely, SWMM performed better than DRAINMOD-Urban for
overflow hydrographs (Table 3). The overflow peaks predicted by
SWMM were smaller and closer to measured peaks than those predicted
by DRAINMOD-Urban (Fig. 4). It has been shown that peaks have a large
effect on the NSE statistic (Broekhuizen et al., 2020; Lisenbee et al.,
2020; Krause et al., 2005). The calculation of overflow in each model
was very similar, simply identifying when the water level exceeds the
specified maximum ponding depth and attributing excess stormwater to
overflow. Therefore, the better prediction of overflow hydrographs by
SWMM may be due to differences in other hydrologic processes such as
percolation and exfiltration or the infiltration capacity at the surface.
3.3.2. Hydrograph characteristics
The peak flow, time to peak, and event duration were also used to
characterize measured and modeled hydrographs. For each of these
hydrograph characteristics, the measured and simulated values from
SWMM and DRAINMOD-Urban for each drainage or overflow event
Fig. 8. Measured hydrograph attributes (peak flow, time to peak and duration) for drainage (a, b, c) and overflow (d, e, f) compared to modeled hydrograph at­
tributes from hydrograph-calibrated DRAINMOD-Urban (DM-Urban, red triangles) and SWMM (blue circles) simulations. (For interpretation of the references to
colour in this figure legend, the reader is referred to the web version of this article.)
13
Journal of Hydrology 612 (2022) 128179
were compared against the 1:1 line to find patterns in under- and
overprediction (Fig. 8). This was only performed for the hydrograph-
calibration as that simulation provided drainage and overflow hydro­
graphs closest to those measured for both DRAINMOD-Urban and
SWMM.
Peak simulated drainage rate reached a maximum threshold in both
models (Fig. 8a). For SWMM, the maximum drainage rate was 8.3 L/s
which is equal to the measured Ksat (16.8 cm/hr) minus the seepage rate
(0.44 cm/hr). This maximum drainage rate occurred during rainfall
events with depths greater than 25 mm; therefore, most drainage events
(eight out of 12) reached this maximum causing a rectangular shape and
overestimated peaks in the drainage hydrographs. For DRAINMOD-
Urban, the maximum drainage rate was 4.8 L/s (or 9.5 cm/hr over the
bioretention area) which occurred during events with rainfall depths
greater than 42 mm (six out of 12 drainage events). During DRAINMOD-
Urban calibration, this maximum drainage rate was shown to vary with
changes in the bioretention media Ksat but was not simply equal to the
Ksat minus the seepage rate as it was with SWMM. In fact, the maximum
drainage rate for DRAINMOD-Urban was smaller than the calibrated Ksat
used for the bioretention media. For overflow, the peak flow was over­
estimated by both models, but SWMM tended to be closer to measured
overflow peaks (Fig. 8d).
The drainage time to peak was slightly underestimated by both
SWMM and DRAINMOD-Urban (Fig. 8b). The average difference be­
tween the time to peak measured and simulated by DRAINMOD-Urban
was only 19 min with no difference larger than an hour. In SWMM,
the average difference between measured and modeled time to peak was
35 min with a maximum difference of 3.3 h. This is attributed to the
plateaued peaks in many SWMM drainage hydrographs which reached a
maximum drainage flow rate faster than a single peak flow. Another
consideration is that the IWS of the UC cell is not represented in SWMM
as well as in DRAINMOD-Urban. A column study of free drainage versus
IWS showed a delay in peak drainage under IWS conditions (Liu and
W.A. Lisenbee et al.
Fassman-Beck, 2017). The time to peak for overflow hydrographs was
estimated almost perfectly by both models (Fig. 8e). The duration of
drainage hydrographs seemed to be underpredicted for some storms
events with DRAINMOD-Urban and SWMM, but overall, both models
had good results compared to measured values (Fig. 8c). For the four
overflow hydrographs evaluated, it was difficult to detect a pattern
among the duration predicted by DRAINMOD-Urban and SWMM
(Fig. 8f).
Although studies evaluating drainage and overflow hydrographs
from the SWMM LID module are rare, one study showed similar rect­
angular hydrographs from SWMM compared to a column study under a
variety of rainfall intensities and durations (Gülbaz and Kazezyılmaz-
Alhan, 2017). In this study, SWMM underestimated peak drainage rates;
however, the drainage rates from the UC cell were two or three orders of
magnitude higher than those in this column study. The most notable
finding by Gulbaz and Kazezyilmaz-Alhan (2017) was that there was no
change in the peak flow of SWMM-predicted drainage hydrographs
across different rainfall intensities and durations. Therefore, SWMM was
predicting the columns would reach saturation and thus their maximum
peak drainage regardless of variations in inflow. Another study showed
SWMM drainage hydrographs with very sharp peaks, slow rising limbs,
truncated receding limbs, and mismatched duration compared to
measured drainage from a bioretention cell column (Liu and Fassman-
Beck, 2017). These hydrograph characteristics do not align with find­
ings from this research due to the lack of an IWS zone in this column.
3.3.3. Volume performance
SWMM consistently predicted event volumes better than
DRAINMOD-Urban as shown through goodness-of-fit tests; the excep­
tion was drainage volumes under the hydrograph-calibration (Table 3).
The best fit for overflow volumes from DRAINMOD-Urban occurred
under hydrograph-calibration (NSE = 0.66) which had the worst over­
flow volume performance in SWMM (NSE = 0.59).
In DRAINMOD-Urban, drainage volumes improved as the model was
calibrated to match the measured and modeled hydrographs, but in
SWMM, drainage volume performance decreased as the hydrograph
performance improved. Therefore, the user must decide whether to
calibrate SWMM to achieve more accurate volumes or hydrographs (i.e.,
timing and rate of flow); in DRAINMOD-Urban, calibrating to the
hydrograph improved both factors.
3.3.4. Effect of model processes
The differences in hydrographs and event volumes created by
DRAINMOD-Urban and SWMM can be explained by internal processes of
the models. The soil water distribution in each model led to substantial
differences in modeling infiltration through the media, drainage calcu­
lations, and representation of the IWS layer.
While DRAINMOD-Urban and SWMM both use the Green-Ampt
equation for surface infiltration, the Green-Ampt equation in the LID
module of SWMM was modified to account for the effect of ponding
(Rossman and Huber, 2016). Meanwhile, DRAINMOD-Urban accounts
for ponding in its soil water distribution procedures (Skaggs et al.,
2012). This means that surface infiltration could be very similar in these
two models if there is adequate infiltration capacity in the rest of the
bioretention cell, which is determined by how easily water flows
through the soil layer. If this becomes limited, then it will influence
ponding and eventually overflow.
For percolation through the bioretention media, DRAINMOD-Urban
uses Green-Ampt with a Keff at different water level depths which is
recalculated at each time step as the water moves through the soil
media. SWMM uses the percolation equation ((1)) which often is
simplified to the Ksat when the soil is assumed to be saturated as seen in
the truncated drainage hydrographs (Fig. 4). When not saturated, the
percolation in SWMM depends on the average soil moisture and HCO
parameter which is a highly uncertain, empirical parameter that is
estimated by most users. The change in water holding capacity in the soil
14
Journal of Hydrology 612 (2022) 128179
layer through each of these methods affects the rate and amount of water
reaching the storage layer and the underdrain.
Once water reaches the underdrain, differences in drainage equa­
tions affect modeled drain output (as described in section 3.2.6).
Without drainage restrictions, SWMM assumes all percolation that rea­
ches the underdrain minus the user-defined seepage rate becomes
drainage. DRAINMOD-Urban uses the Keff from the soil layer to calculate
the drainage rate with more physically based equations. SWMM over­
estimated drainage peak flow while DRAINMOD-Urban often under­
estimated drainage peak flow (Fig. 8). By overestimating the drainage
rate in the bioretention cell, SWMM provides more infiltration capacity
which reduced the volume and intensity of stormwater converted to
overflow such that overflow peaks were closer to measured peaks than
with DRAINMOD-Urban (Fig. 5). Therefore, the improved overflow
performance in SWMM could be an artefact of the drainage design and
leaves the question: if drainage is improved will overflow perform as
well?
Another consideration regarding drainage and overflow prediction is
how the IWS zones are represented in each model. DRAINMOD-Urban
simply allows for controlled drainage at a given weir height which
directly matches to the IWS configuration in the UC cell (and most
bioretention cell with an IWS zone). The drainage configuration in
SWMM had to be adjusted to represent the IWS which increased the
storage zone to include the sand layer and raised the height of the
underdrain to the top of the IWS layer (Fig. 2). This IWS configuration
could affect the drainage calculated by SWMM but warrants more
investigation. With the exception of the volume calibration in SWMM,
overestimation in drainage volumes also led to an underestimation of
overflow volumes (Fig. 7). Therefore, the water balance under the un­
calibrated and hydrograph-calibrated simulations could be better rep­
resented in SWMM if IWS configurations were improved in the model.
We recommend more testing of underdrain configurations in the SWMM
LID module to better understand the connection of the IWS model rep­
resentation to the drainage and overflow outputs.
4. Summary and conclusions
The intended application of a bioretention model has a significant
impact on model selection and how or if the model is calibrated. If the
objective of the application is to understand event drainage and over­
flow volumes from bioretention, then SWMM is a good choice. SWMM is
capable of modeling event volumes with good to excellent fit in both
uncalibrated and calibrated simulations ranging from NSE = 0.70 to
0.93 for drainage and NSE = 0.59 to 0.81 for overflow.
The performance of both models suffered when attempting to model
a high temporal resolution time series, performing much better for
aggregated event volumes. However, drainage or overflow hydrographs
are necessary for applications considering flow dynamics such as stream
response, combined sewer systems, or flooding. The hydrograph-
calibrated DRAINMOD-Urban performs better than SWMM for
drainage hydrographs which is not surprising when comparing drainage
equations employed by each model. SWMM performs better than
DRAINMOD-Urban for overflow hydrographs primarily by more closely
predicting overflow event peak flow rates.
DRAINMOD-Urban benefits more from calibration of drainage and
overflow hydrographs in comparison with SWMM. Interestingly, in
SWMM, calibration to optimize one hydrologic objective (event volumes
or hydrographs), led to diminished performance of the other objective.
This presents a choice for SWMM users: are volumes or hydrographs
more important for drainage and/or overflow given the application? For
DRAINMOD-Urban, the hydrograph-calibration improves both the vol­
umes and hydrographs of drainage and overflow eliminating a decision
between better predicted volumes or hydrographs.
Future bioretention research requires further calibration of
DRAINMOD-Urban and SWMM for additional bioretention sites, espe­
cially with data sets containing more than four overflow events. More
W.A. Lisenbee et al.
overflow events need to be analyzed to improve the confidence of sta­
tistical conclusions about overflow performance in each model. This
study only considered one bioretention cell, but additional sites need to
be modeled to understand how these models behave across a variety of
conditions such as storm size, drainage configurations, and other bio­
retention design features. In particular, the IWS representation in
SWMM needs to be investigated further and improved in the model to
increase the applicability of the LID module to all bioretention systems.
Sensitivity analysis could also improve understanding of how mea­
surement or estimation of model inputs affect the accuracy of output
hydrographs. In future work, utilizing DRAINMOD-Urban as an add-on
tool for SWMM would improve drainage hydrographs in SWMM and
allow for better representation of bioretention cells in watershed-scale
models.
CRediT authorship contribution statement
W.A. Lisenbee: Conceptualization, Methodology, Investigation,
Software, Formal analysis, Writing – original draft. J.M. Hathaway:
Conceptualization, Funding acquisition, Resources, Supervision,
Writing – review & editing. R.J. Winston: Methodology, Investigation,
Validation, Writing – review & editing.
Declaration of Competing Interest
The authors declare that they have no known competing financial
interests or personal relationships that could have appeared to influence
the work reported in this paper.
Acknowledgements:
Funding for this research was provided by the National Science
Foundation grant no. 1553475.
Appendix A. Supplementary data
Supplementary data to this article can be found online at https://doi.
org/10.1016/j.jhydrol.2022.128179.
References
Akan, A.O., 2013. Preliminary Design Aid for Bioretention Filters. J. Hydrol. Eng. 18 (3),
318–323.
Avellaneda, P.M., Jefferson, A.J., Grieser, J.M., Bush, S.A., 2017. Simulation of the
Cumulative Hydrological Response to Green Infrastructure. Water Resour. Res. 53
(4), 3087–3101.
Bai, Y., Zhao, N.a., Zhang, R., Zeng, X., 2018. Storm Water Management of Low Impact
Development in Urban Areas Based on SWMM. Water 11 (1), 33.
Barbu, I.A., Ballestero, T.P., 2015. Unsaturated Flow Functions for Filter Media Used in
Low-Impact Development—Stormwater Management Systems. J. Irrig. Drain. Eng.
141 (1), 04014041.
Bosley, E.K., 2008. Hydrologic Evaluation of Low Impact Development Using a
Continuous. Spatially-Distributed Model. Civil Engineering. Master of Science.
Broekhuizen, I., Leonhardt, G., Marsalek, J., Viklander, M., 2020. Event Selection and
Two-Stage Approach for Calibrating Models of Green Urban Drainage Systems.
Hydrol. Earth Syst. Sci. 24 (2), 869–885.
Brown, R.A., Hunt, W.F., 2011. Impacts of Media Depth on Effluent Water Quality and
Hydrologic Performance of Undersized Bioretention Cells. J. Irrig. Drain. Eng. 137
(3), 132–143.
Brown, R.A., Hunt, W.F., 2012. Improving Bioretention/Biofiltration Performance with
Restorative Maintenance. Water Sci. Technol. 65 (2), 361–367.
Brown, R.A., Skaggs, R.W., Hunt, W.F., 2013. Calibration and Validation of DRAINMOD
to Model Bioretention Hydrology. J. Hydrol. 486, 430–442.
Carpenter, D., Hallam, L., 2010. Influence of Planting Soil Mix Characteristics on
Bioretention Cell Design and Performance. J. Hydrol. Eng. 15 (SPECIAL ISSUE: Low
Impact Development, Sustainability Science, and Hydrological Cycle), 404–416.
Davis, A.P., 2008. Field Performance of Bioretention: Hydrology Impacts. J. Hydrol. Eng.
13 (2), 90–95.
Davis, A.P., Hunt, W.F., Traver, R.G., Clar, M., 2009. Bioretention Technology: Overview
of Current Practice and Future Needs. J. Environ. Eng. 135 (3), 109–117.
Davis, A.P., Traver, R.G., Hunt, W.F., Lee, R., Brown, R.A., Olszewski, J.M., 2012.
Hydrologic Performance of Bioretention Storm-Water Control Measures. J. Hydrol.
Eng. 17 (5), 604–614.
15
Journal of Hydrology 612 (2022) 128179
DRAINMOD 6.1 Help File (2013). North Carolina State University.
Elliott, A., Trowsdale, S., 2007. A Review of Models for Low Impact Urban Stormwater
Drainage. Environ. Modell. Software 22 (3), 394–405.
Fassman-Beck, E., Saleh, F., 2021. Sources and Impacts of Uncertainty in Uncalibrated
Bioretention Models Using SWMM 5.1.012. J. Sustain. Water Built Environ. 7 (3).
Gülbaz, S., Kazezyılmaz-Alhan, C.M., 2017. An Evaluation of Hydrologic Modeling
Performance of EPA SWMM for Bioretention. Water Sci. Technol. 76 (11),
3035–3043.
Hathaway, J.M., Brown, R.A., Fu, J.S., Hunt, W.F., 2014. Bioretention Function under
Climate Change Scenarios in North Carolina, USA. J. Hydrol. 519, 503–511.
Hunt, W.F., Jarrett, A.R., Smith, J.T., Sharkey, L.J., 2006. Evaluating Bioretention
Hydrology and Nutrient Removal at Three Field Sites in North Carolina. J. Irrig.
Drain. Eng. 132 (6), 600–608.
Kaykhosravi, S., Khan, U., Jadidi, A., 2018. A Comprehensive Review of Low Impact
Development Models for Research, Conceptual, Preliminary and Detailed Design
Applications. Water 10 (11), 1541.
Krause, P., Boyle, D.P., Base, F., 2005. Comparison of Different Efficiency Criteria for
Hydrological Model Assessment. Adv. Geosci. 5, 89–97.
Li, C., Fletcher, T.D., Duncan, H.P., Burns, M.J., 2017. Can Stormwater Control Measures
Restore Altered Urban Flow Regimes at the Catchment Scale? J. Hydrol. 549,
631–653.
Li, Z.Y., Lam, K.M., 2015. Statistical Evaluation of Bioretention System for Hydrologic
Performance. Water Sci. Technol. 71 (11), 1742–1749.
Lisenbee, W., Hathaway, J., Negm, L., Youssef, M., Winston, R., 2020. Enhanced
Bioretention Cell Modeling with DRAINMOD-Urban: Moving from Water Balances to
Hydrograph Production. J. Hydrol. 582, 124491.
Lisenbee, W.A., Hathaway, J.M., Burns, M.J., Fletcher, T.D., 2021. Modeling Bioretention
Stormwater Systems: Current Models and Future Research Needs. Environ. Modell.
Software 144, 105146.
Liu, R.F., Fassman-Beck, E., 2017. Hydrologic Experiments and Modeling of Two
Laboratory Bioretention Systems under Different Boundary Conditions. Front.
Environ. Sci. Eng. 11 (4).
Liu, J., Sample, D., Bell, C., Guan, Y., 2014. Review and Research Needs of Bioretention
Used for the Treatment of Urban Stormwater. Water 6 (4), 1069–1099.
Lucas, W., 2010. Design of Integrated Bioinfiltration-Detention Urban Retrofits with
Design Storm and Continuous Simulation Methods. J. Hydrol. Eng. 15 (6), 486–498.
Lynn, T.J., Nachabe, M.H., Ergas, S.J., 2018. SWMM5 Unsaturated Drainage Models for
Stormwater Biofiltration with an Internal Water Storage Zone. J. Sustain. Water Built
Environ. 4 (1), 04017018.
McCutcheon, M., Wride, D., 2013. Shades of Green: Using SWMM LID Controls to
Simulate Green Infrastructure. JWMM.
Moriasi, D.N., Arnold, J.G., Van Liew, M.W., Bingner, R.L., Harmel, R.D., Veith, T.L.,
2007. Model Evaluation Guidelines for Systematic Quantification of Accuracy in
Watershed Simulations. Trans. ASABE 50 (3), 885–900.
NCDEQ, 2020. NCDEQ Stormwater Design Manual. Bioretention Cell, North Carolina
Department of Environmental Quality. https://deq.nc.gov/about/divisions/ener
gy-mineral-and-land-resources/stormwater/stormwater-program/stormwater-desi
gn.
Olszewski, J.M., Davis, A.P., 2013. Comparing the Hydrologic Performance of a
Bioretention Cell with Predevelopment Values. J. Irrig. Drain. Eng. 139 (2),
124–130.
Platz, M., Simon, M., Tryby, M., 2020. Testing of the Storm Water Management Model
Low Impact Development Modules. J. Am. Water Resour. Assoc. 56 (2), 283–296.
Rosa, D.J., Clausen, J.C., Dietz, M.E., 2015. Calibration and Verification of SWMM for
Low Impact Development. J. Am. Water Resour. Assoc. 51 (3), 746–757.
Rossman, L.A., 2010. Modeling Low Impact Development Alternatives with SWMM.
JWMM.
Rossman, L.A., Huber, W.C., 2016. Storm Water Management Model Reference Manual
Volume III – Water Quality. National Risk Management Laboratory, Office of
Research and Development, U.S. Environmental Protection Agency, Cinncinati, OH.
Saraswat, D., Frankenberg, J.R., Pai, N., Ale, S., Daggupati, P., Douglas-Mankin, K.R.,
Youssef, M.A., 2015. Hydrologic and Water Quality Models: Documentation and
Reporting Procedures for Calibration, Validation, and Use. Trans. ASABE 58 (6),
1787–1797.
Saxton, K.E., Rawls, W.J., 2006. Soil Water Characteristic Estimates by Texture and
Organic Matter for Hydrologic Solutions. Soil Sci. Soc. Am. J. 70 (5), 1569.
Skaggs, R.W., Youssef, M.A., Chescheir, G.M., 2012. Drainmod: Model Use, Calibration,
and Validation. Trans. ASABE 55 (4), 1509–1522.
Sun, Y.-W., Wei, X.-M.-C., Pomeroy, C.A., 2011. Global Analysis of Sensitivity of
Bioretention Cell Design Elements to Hydrologic Performance. Water Sci. Eng. 4 (3),
246–257.
Sun, Y.-W., Li, Q.-Y., Liu, L., Xu, C.-D., Liu, Z.-P., 2014. Hydrological Simulation
Approaches for BMPs and LID Practices in Highly Urbanized Area and Development
of Hydrological Performance Indicator System. Water Sci. Eng. 7 (2), 143–154.
EPA SWMM Help File 5.1 (2017). Washington, DC, U.S. Environmental Protection
Agency.
Tirpak, R.A., Afrooz, A.N., Winston, R.J., Valenca, R., Schiff, K., Mohanty, S.K., 2021.
Conventional and Amended Bioretention Soil Media for Targeted Pollutant
Treatment: A Critical Review to Guide the State of the Practice. Water Res. 189,
116648.
Tiveron, T., Gholamreza-Kashi, S., Joksimovic, D., 2018. A USEPA SWMM Integrated
Tool for Determining the Suspended Solids Reduction Performance of Bioretention
Cells. J. Water Manage. Modeling.
W.A. Lisenbee et al. Journal of Hydrology 612 (2022) 128179

Tu, M.C., Wadzuk, B., Traver, R., 2020. Methodology to Simulate Unsaturated Zone Winston, R.J., Dorsey, J.D., Hunt, W.F., 2016. Quantifying Volume Reduction and Peak
Hydrology in Storm Water Management Model (SWMM) for Green Infrastructure Flow Mitigation for Three Bioretention Cells in Clay Soils in Northeast Ohio. Sci.
Design and Evaluation. PLoS ONE 15 (7), e0235528. Total Environ. 553, 83–95.
Winston, R. J. (2015). Resilience of Green Infrastructure under Extreme Conditions.

16