Environmental Modelling & Software 22 (2007) 394e405

www.elsevier.com/locate/envsoft

A review of models for low impact urban stormwater drainage

A.H. Elliott a,*, S.A. Trowsdale b
a

National Institute of Water and Atmospheric Research, PO Box 11-115, Hamilton, New Zealand
b
Landcare Research, Private Bag 92170, Auckland, New Zealand

Received 3 April 2005; received in revised form 2 December 2005; accepted 13 December 2005
Available online 3 March 2006

Abstract
Low-impact development urban stormwater drainage systems (LID) are an increasingly popular method to reduce the adverse hydrologic and
water quality effects of urbanisation. In this review, ten existing stormwater models are compared in relation to attributes relevant to modelling
LID. The models are all based on conventional methods for runoff generation and routing, but half of the models add a groundwater/baseflow
component and several include infiltration from LID devices. The models also use conventional methods for contaminant generation and treatment such as buildup-washoff conceptual models and first order decay processes, although some models add treatment mechanisms specific to
particular types of LID device. Several models are capable of modelling distributed on-site devices with a fine temporal resolution and continuous simulation, yet the need for such temporal and spatial detail needs to be established. There is a trend towards incorporation of more types of
LID into stormwater models, and some recent models incorporate a wide range of LID devices or measures. Despite this progress, there are
many areas for further model development, many of which relate to stormwater models in general, including: broadening the range of contaminants; improving the representation of contaminant transport in streams and within treatment devices; treating baseflow components and runoff
from pervious surfaces more thoroughly; linkage to habitat and toxicity models; linkage to automated calibration and prediction uncertainty
models; investigating up-scaling for representation of on-site devices at a catchment level; and catchment scale testing of model predictions.
2006 Elsevier Ltd. All rights reserved.
Keywords: Stormwater; Catchment; Model; Review; Urban drainage; Low impact

1. Introduction
Worldwide, there is a well documented decline in habitat
and water quality of urban streams. Urbanisation is typically
accompanied by increases in impervious surfaces such as roofs
and roads, construction of hydraulically efficient drainage systems, compaction of soils, and modifications to vegetation.
This results in increased flood flows (Leopold, 1968) and
stream erosion (Hammer, 1972), and the potential for
decreased baseflow (Paul and Meyer, 2001; Schueler, 1994).
Urbanisation also leads to water contamination from

* Corresponding author. Tel.: 64 (9) 8567026; fax: 64 (9) 8560151.

E-mail address: s.elliott@niwa.co.nz (A.H. Elliott).
1364-8152/$ - see front matter 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.envsoft.2005.12.005

suspended sediments, heavy metals, hydrocarbons, nutrients,

and pathogens (Burton and Pitt, 2001; Hall, 1984).
In the last two decades, new urban water management
approaches have been developed to deliver improved environmental, economic, social and cultural outcomes. We term such
an approach LID (low impact development), but alternative acronyms are SUDS (sustainable urban drainage systems), WSUD
(water sensitive urban design), and LIUDD (low impact urban
design and development, a term used in New Zealand). In this review, we focus on stormwater aspects of LID, with limited attention to broader issues of integrated urban water cycle
management. The scope is also limited to the effects of stormwater on water quality and quantity, rather than visual, social
and economic impacts.
LID devices are designed to detain, store, infiltrate, or treat
urban runoff, and so reduce the impact of urban development

A.H. Elliott, S.A. Trowsdale / Environmental Modelling & Software 22 (2007) 394e405

(e.g. Wong et al., 2002; NZWERF, 2004). LID devices include

structural measures such as wetlands, ponds, swales, rainwater
tanks, bioretention devices, vegetated filter strips, and filter
strips. LID approaches also include non-structural measures
such as alternative layouts of roads and buildings to minimise
imperviousness and to maximise the use of pervious soils and
vegetation, contaminant source reduction, and programmes of
education to modify activities. LID particularly emphasises
on-site small-scale control of stormwater sources. Many
design guidelines for such devices are now available (e.g.,
CIRIA, 2000).
Despite an increasing awareness and knowledge of these issues and potential solutions, the transition to more sustainable
urban drainage design has been slow. This may reflect, among
other factors, a dearth of LID drainage design tools that operate effectively at the necessary range of scales. The availability
of effective LID modelling software could act to encourage
wider uptake of LID principles (Beecham, 2002). Tools can
make design and application of LID more efficient, and demonstrate outcomes that can be used for education and policy
development. The challenge is to translate complex and highly
variable natural processes into a computerised system or tool
that allows straightforward evaluation of LID drainage measures at a range of scales applicable to urban management.
This paper explores the current range of LID assessment
tools, discusses their strengths and weaknesses and puts forward future research needs, with the aims of aiding model selection, increasing awareness of the available models, and
encouraging model development.
1.1. Previous reviews
Zoppou (2001) reviewed both quantity and quality aspects
of urban stormwater models. He provided an overview of
stormwater modelling approaches, with a concise mathematical description of common methods for flow routing and contaminant generation and transport. He also described several
stormwater models. Burton and Pitt (2001, Appendix H) reviewed catchment and receiving water modelling in relation
to stormwater. They classified the catchment models primarily
according to the complexity, ranging from simple methods
(based on export coefficients or event-mean concentrations
multiplied by runoff volume) through to complex models
that are typically spatially distributed and processed based.
McAlister et al. (2003) reviewed urban stormwater quality
models, and stressed the importance of using suitably small
temporal resolution and continuous simulation over one or
more years. Beecham (2002) presented key features of four
models for water sensitive urban design, but did not compare
and contrast the models or discuss their suitability. Other reviews of contaminant models (e.g. the review of sediment
models by Merritt et al., 2003) are not focussed on urban
stormwater or LID.
These previous reviews provide a valuable background on
the features of a range of models, methods for representing
key processes, and categorisations of the models. However,
none of them focus specifically on the ability of the models

395

to represent LID. In this review the focus is urban stormwater models and LID.

1.2. Review process and structure

We identified approximately 40 models for urban stormwater from previous published reviews, journal abstracting
services, internet searches, conference proceedings, and modelling practitioners. We then selected 10 models that are currently available and have not been superseded, have
sufficient documentation in English, and are more than a conventional stormwater drainage/hydrology model. Our assessment is based on versions of the models available in
February 2005, and the range of models and the features of
the models may have changed since that time.
The ten models (Table 1) were compared in relation to
the following attributes:
e The intended uses of the model including research, public
education, developing device sizing rules, catchment planning, and conceptual to detailed design. All these levels of
model use are relevant to LID.
e Temporal resolution and scale. The temporal resolution refers to the smallest computational timestep of the model.
We also distinguish between models intended only for
a single rainfall event and those intended for simulation
of a long-term sequence of events (continuous simulation)
as discussed in Singh (1995).
e Catchment and drainage network representation, and
spatial resolution and scale. The catchment and drainage representation refers to the types of element that
are used to represent the catchment, soil column and
groundwater, drainage network, and treatment or flow
control devices. We categorise models according to
whether they are lumped (a single catchment element),
quasi-distributed (where the model is broken into
a number of elements such as subcatchments), or fully
distributed (usually grid or mesh-based) as discussed in
Singh (1995). In each of these categories, we include
cases where the catchment element is broken down
into a number of land uses, surface types or stormwater
treatment categories. Spatial scale refers to the size of
the modelled area.
e Representation of runoff generation, routing to the drainage network, routing within the drainage network, and
groundwater movement.
e Types of contaminant included in the models, and methods
used to represent processes of contaminant generation,
transport and treatment.
e LID devices or technologies specifically included in the
model, or able to be simulated indirectly using the model.
The devices assessed range from on-site non-structural
controls such as reduction of imperviousness, to regional
scale wetlands.
e User interface and integration with other software such as
automated calibration software or receiving-water models.

396

Table 1
Name and introductory information for the selected models (unless stated otherwise, the source code is not available)
Versions

References

Primary author or
organisation

Cost (USD)

Primary intended use

MOUSE

First: 1985.
DHI, 2002aed; http://www.
Latest: MOUSE 2004 dhisoftware.com/mouse

DHI Water and

Environment

w5000 for basic flow module.

Further modules comparable.
MIKE STORM, a version with
reduced capabilities, costs less

Detailed simulation of urban drainage.

Widespread use outside USA

MUSIC (Model for

Urban Stormwater
Improvement Conceptualisation)

First: 2000. Latest:

Version 2.0, 2003

Chiew and McMahon, 1999;

Monash University and
MUSIC Development
the CRC for Catchment
Team, 2003;
Hydrology, Australia
Wong et al., 2002; http://www.
toolkit.net.au/music

w300

Conceptual design for drainage systems,

with emphasis on treatment devices.
Popular in Australia

P8-UCM

First: 1990. Latest:

Version 2.4, 2002

Palmstrom and Walker, 1990;

http://wwwalker.net/

Free

Estimation of urban stormwater

pollutant load

PURRS (Probabilistic
Urban Rainwater and Wastewater
Reuse Simulator)

First: c. 1999. Latest: Coombes, 2002;

Version 6.5, 2004
http://rambler.newcastle.edu.
au/wcegak/Coombes/

Peter Coombes, Newcastle w800

University, Australia

Single site water use model. Originally

for research but now includes commercial
users, especially for rain tanks

RUNQUAL

Latest 1999

Douglas Haith, Cornell

University

Free, including
source code

Preliminary planning or education

SLAMM (Source
Loading and Management Model)

Latest: Winslamm 8.7, Pitt, 1998; PV & Associates

2004

Bob Pitt, University of

Alabama.

200

Planning tool for load of contaminants

StormTac

Latest: 2004

Larm, 2000, 2003; http://

www.stormtac.com

Thomas Larm, SWECO

VIAK

2500

Management of lake catchments and

conceptual design of stormwater
treatment. Applied in Scandinavia

SWMM (Storm Water

Management Model); XP-SWMM;
PCSWMM; MIKESWMM

First: 1970. Latest:

SWMM5 in 2004

Various for USEPA.

Rossman 2004; Huber and
Dickinson 1988.
Proprietary versions are XPSWMM
(http://www.xpsoftware.com.au/
products/xpswmm.htm),
PCSWMM (http://
www.computationalhydraulics.
com/) and MIKE-SWMM (http://
www.dhisoftware.
com/mikeswmm/)

USEPA version is free,

including code

Detailed model for planning and

preliminary design. Widely used

UVQ (Urban
Volume and Quality)

First: 2000

Mitchell et al. 2003;

Mitchell and Diaper (2006)

CSIRO and Monash

University, Australia

Available for a
small charge

Integrated water cycle, water re-use,

Used mainly for research in Australia

Water Balance
Model (WBM)

First: 2004

http://www.waterbalance.ca

Greater Vancouver RegionalWeb-based. Basic model free:

District
ongoing licence payment
for full model

http://wri.eas.cornell.edu/
products/software/runqual/

William W.
Walker Jr.

Planning-level assessment of
water quantity. Strong support in British
Columbia

A.H. Elliott, S.A. Trowsdale / Environmental Modelling & Software 22 (2007) 394e405

Model

A.H. Elliott, S.A. Trowsdale / Environmental Modelling & Software 22 (2007) 394e405

All the models are capable of long-term continuous simulation (with the exception of StormTac which is based on mean
annual average values). Simulations covering 10 years or more
are computationally feasible, provided that the level of spatial
detail is not excessive (say, less than 200 spatial elements) and
provided that stability constraints do not necessitate very small
timesteps (in the order of one second).
MOUSE, SWMM, and MUSIC are most suited for prediction of flow rates from small catchments, while the daily or annual models (RUNQUAL, SLAMM, StormTac and UVQ) are
unsuited for this purpose. MUSIC has a smallest timestep of
6 min, so it has limited applicability for predicting flow rates
from areas smaller than about 0.01 km2.
For modelling of small on-site LID devices and small
catchments, sub-hourly timesteps may be required as the timescale of variation in the runoff and associated treatment processes is likely to be in the order of minutes (McAlister
et al., 2003). MOUSE and SWMM are suited for this purpose,
while MUSIC (with its smallest timestep of 6 min) is marginally suited. On the other hand, for the purpose of establishing
contaminant loads and annual water balance (rather than the
timing of loads and flow rates), longer timesteps may be adequate. This aspect of temporal resolution requirements warrants further systematic evaluation. In the absence of such
studies, there is likely to be a trend to the use of small timesteps, even though small timesteps may not actually be
required to adequately address a particular management issue.

2. Comparison of models in relation to selected attributes

Basic information on the models (such as cost and availability) is given in Table 1. In this section we compare the
models in relation to the set of attributes described above
and discuss how this relates to representation of LID, but we
leave a discussion of the collective limitations of the models
until Section 3.
2.1. Potential uses of the model
The range of uses for each of the selected models is summarised in Fig. 1. Two of the models (MOUSE and
SWMM) are suitable for a wide range of uses, yet they are
too complex to be used by the general public or non-modelling
planners. Other models (StormTac and PURRS) have a comparatively restricted range of uses. The remaining six models
have a moderate range of uses, mostly clustered around planning or preliminary design, which may reflect an attempt to
encourage LID by targeting the level of use in which broad
principles start to be converted to designs or planning measures for specific catchments or sites. MUSIC, P8 and WBM
are best suited for conceptual or preliminary design at either
a subdivision or catchment scale.
2.2. Temporal resolution and scale

2.3. Catchment and drainage network representation,

and spatial resolution and scale

The temporal resolution of the models ranges from annual

average to sub-hourly (Fig. 2). Models with the capability for
small timesteps can also usually be run with longer timesteps.
If dynamic-wave flow routing is used, timesteps in the order of
a second may be required to maintain numerical stability, but
this is probably finer than required for representing the runoff
generation, contaminant generation, or contaminant transport
and treatment processes.

te

io
ct
se
le
ls
er
ia

at
m
d

an

si
g
yo

ut

de
d
Si

te

la

le
ai
et
D

si
or
n
io
is
iv

bd
su
of

of
gn
si
de
d
le

ai
et
D

sy
ge
na
ai
dr

na
l
io
re
g

a
of
gn
si
de

ar
y
in
im
Pr
el

st

te
n
io
is
iv
bd
su

on
gi
re
of
gn

si
de
ar
y
in
im
el

Pr

or

ls
co
nt
ro
al

hm
tc
ca
in
e

us
of

ng
ni
an
Pl

el
ev
D

si

s
tie
ci
s/
en
t

es
ic
ev
rd
fo
nd
la

zi
si
g
in
op

rc
h
ea
es
R

em

Four models (MOUSE, MUSIC, SWMM and P8) are spatially distributed with a link-node drainage network. Each of
the catchment elements in these models is associated with
a node of the network, and treatment or flow control devices
are also placed at nodes. The nodes are linked by drainage

es
ru
l
ng

n
io
uc
at
ed
ic
bl
Pu

397

MOUSE
MUSIC
P8
PURRS
RUNQUAL
SLAMM
StormTac
SWMM
UVQ
WBM

Fig. 1. Potential uses for the selected models. Grey shading indicates that the model is marginally suited to that use.

A.H. Elliott, S.A. Trowsdale / Environmental Modelling & Software 22 (2007) 394e405

ly

b-

ly

b-

ur
D
is
tri
b

ut
ed

su

ho
D
is
tri
b

ut
ed

ur
ho
d
m
pe
Lu

ho

ur

su
or
ly

ily
da
ut
ed
D
is
tri
b

Lu

Lu

m
pe

m
pe

da

an

ily

nu

al

or

av

ev

er

en

ag

ho

ur

ly

398

MOUSE
MUSIC
P8
PURRS
RUNQUAL
SLAMM
StormTac
SWMM
UVQ
WBM

Fig. 2. Spatial and temporal resolution of the selected models.

elements (pipes or channels). In MOUSE, the catchment element is divided into a number of different contaminant-generating
surfaces or land uses, whereas in the other link-node models the
catchment element is taken to be homogeneous in relation to
contaminant generation. UVQ has a novel representation of
the catchment, using three nested spatial components (property
or unit block, neighbourhood or land use, and catchment) so that
a range of scales and associated types of water management can
be addressed. These quasi-distributed models allow for explicit
representation of the spatial distribution of LID devices.
The remaining models (PURRS, RUNQUAL, SLAMM,
StormTac and WBM) treat the catchment in a lumped fashion
with no drainage network, except that the catchment may be
broken into a number of land uses or surface classes (with
the exception of PURRS).
All the models except for StormTac divide the catchment
elements (such as subcatchments) into pervious and impervious components for runoff generation.

N
ur
ve

ua
lr
SC

pt

F
(e urth
.g e
.G rr
re uno
en ff
-A ge
m n
pt er
G
) at
ro
io
un
n
dw
op
tio
at
er
ns
/
b
R
a
se
ou
flo
tin
w
g
to
dr
ai
na
ge
R
ou
ne
tin
tw
g
or
th
k
ro
ug
h
H
d
yd
ev
ro
ic
es
lo
gi
c
ro
ut
in
H
g
yd
in
ra
dr
ul
ai
ic
na
ro
ge
ut
ne
in
g
tw
or
k

al
ai
nf

en
t
ffi
ci
C

on
ce

fc
oe
un
of
R

Routing

um
be
r

l-r
un
of

Runoff generation

Over half of the models (MOUSE, MUSIC, PURRS,

SWMM, UVQ and WBM) include soil moisture stores (up
to three) in each catchment element, and five models include
a groundwater store in each catchment element (Fig. 3).
Most of the models have no inherent limit on the spatial
extent of the modelled area, and could be set up for a range
of spatial scales encountered in urban areas, ranging from
a single site (w100 m2) up to medium catchments
(w10 km2). An exception is PURRS, which is only intended
for single sites. However, the timestep in some of the models
(such as P8) means that predictions of flow rates would not be
reliable for small catchments.
The maximum number of elements in the quasi-distributed
models is also of interest for those cases where the modeller
wishes to incorporate considerable spatial detail or complexity
(as may occur for modelling large catchments or on-site LID).
P8 has a limit of 192 catchment elements. MUSIC would be
impracticable to use with more than about 100 catchment

MOUSE
MUSIC
P8
PURRS
RUNQUAL
SLAMM
StormTac
SWMM
UVQ
WBM

Fig. 3. Runoff generation and routing methods for the selected models.

A.H. Elliott, S.A. Trowsdale / Environmental Modelling & Software 22 (2007) 394e405

elements, as the data need to be entered manually and large

files are created. The level of detail in SWMM and MOUSE
may be limited by computational constraints, especially for
long-term simulation and dynamic flood routing. UVQ can
potentially be used for spatially complex models, due to the
hierarchical spatial configuration and daily timestep.
2.4. Runoff generation
Most of the models are similar in the way that runoff from
impervious areas is generated, and this is the dominant effect
of urbanisation on runoff generation. The models mostly use
simple conventional rainfall-runoff methods for generating
runoff from pervious areas (Fig. 3). MOUSE includes an initial-and-continuing loss option while SWMM includes
a Green-Ampt infiltration option. MUSIC uses daily calculations to determine the volumes of runoff for each runoff component, then temporal disaggregation based on the rainfall
pattern to break each of these components into a sub-daily
time distribution. There is no clear advantage of one of the
runoff generation methods over another in relation to modelling of LID. They are all likely to need calibration or development of suitable regional parameters, and all are somewhat
coarse in relation to treatment of the effects of vegetation
(on soil moisture and interception).
Several models also include a baseflow runoff component,
which is relevant to LID because maintenance of the baseflow
is often a goal of LID. MOUSE, MUSIC, P8 and UVQ use
a single linear lumped groundwater reservoir in each subcatchment, while SWMM has an additional unsaturated zone and
the groundwater reservoir is non-linear.
2.5. Flow routing
Eight of the models have no routing of flow between the
point of runoff generation and the modelled drainage network
(Fig. 3). This is appropriate for six of the models which are not
intended to resolve fine-scale temporal flow variations.
MOUSE and SWMM do include routing of runoff to the drainage network, including a range of conventional methods such
as reservoir routing, unit-hydrographs, and time-area routing.
In MUSIC, such routing would have to be represented approximately using a drainage link, and the minimum timestep of
6 min would have the effect of smoothing of flows for small
catchments. PURRS does not include routing, but the user is
advised to select a timestep comparable to the on-site time
of concentration which partially accounts for runoff attenuation. Such lags could be important in relation to modelling
the flow rates from small or medium size catchments, which
is particularly relevant to modelling of on-site LID devices.
However, resolving such lags and temporal detail in the flows
is probably of secondary importance in relation to predicting
the effects of LID on contaminant loads. In SWMM5, overland flow can also be routed between sub-areas within a subcatchment or between subcatchments, and this capability
could be useful for representing LID (for example, by allowing

399

runoff from roofs to pass over a pervious depression storage

area).
All the models bar one (Stormtac) route flow through
devices (Fig. 3). Simple level-pool routing is used for routing
flows through devices in all the models, but the models vary in
the details such as the types of outflow, specifications of the
outflow rates, and specification of the device dimensions.
Models such as SWMM and MOUSE are the most flexible
in this regard. MUSIC allows for flow to bypass devices, but
reservoirs must have vertical sides and the outlets have a simple configuration. In SWMM and MOUSE, the outflow can be
controlled by downstream water levels if the hydraulic flow
routing option is used. MUSIC, PURRS, and UVQ allow for
time-varying abstraction of water from devices for irrigation
or household use. These capabilities are of direct relevance
to modelling particular types of LID device.
MUSIC and P8 use hydrologic routing in the drainage network (time-lag, linear reservoirs, or MuskinghameCunge
routing). Two models (MOUSE and SWMM) are capable of
dynamic-wave hydraulic routing, and they are also capable
of simpler routing methods such as kinematic wave routing
or hydrologic routing. For representation of LID, fully dynamic flood routing would often not be necessary, except perhaps in the lower reaches of a catchment where backwater
effects are more likely or for off-line detention facilites.
2.6. Contaminant range, generation, transport,
and treatment
2.6.1. Range of contaminants
The contaminants included in each of the models are shown
in Fig. 4. Two of the models (PURRS and WBM) deal only
with flow. The remaining models can be used to model sediment, nutrients, heavy metals, and other sediment-related toxic
contaminants, although heavy metals are included explicitly in
only half of these models. In some models (such as P8) the
main emphasis is on sediment, while other contaminants are
modelled through their association with sediments plus a dissolved fraction. Only MOUSE has the capability to address
dissolved oxygen in streams, and only MOUSE deals with
pathogenic organisms or associated bacterial indicators (but
even then only in a generalised and simple fashion).
In many cases, a given model is not set up to simulate a particular contaminant explicitly, but can be used to represent that
contaminant either by using generic contaminant generation
and treatment options, modelling another contaminant that
has similar behaviour, or by specifying the association of the
contaminant with sediment.
2.6.2. Contaminant generation
The models use a range of methods for contaminant generation, which are discussed in Zoppou (2001). The methods include: buildup-washoff (MOUSE, RUNQUAL, SLAMM, P8
and SWMM5); characteristic concentrations (MUSIC, StormTac, UVQ, SLAMM, RUNQUAL and XP-SWMM), sometimes with a stochastic component (MUSIC, SLAMM and
XP-SWMM); empirical power rating curves for concentration

A.H. Elliott, S.A. Trowsdale / Environmental Modelling & Software 22 (2007) 394e405

yg
ox
d

ra
m
pe
Te

Pa
th
o

ge

ns

tu
r

ol
ve
BO
D
,D
is
s

H
ea

vy

nt
s
N
ut
rie

Se
d

im
en

m
et
al
s,

ot
he

rt
ox

en

ic
s

400

MOUSE
MUSIC
P8
PURRS
RUNQUAL
SLAMM
StormTac
SWMM
UVQ
WBM

Fig. 4. Contaminants included in the selected models. Grey shading indicates models where the contaminant can be modelled only coarsely or indirectly.

as a function of flow rate (P8 and SWMM5); and unit area

loadings (StormTac). In some models the method depends
on the contaminant, and some models include a range of
methods. SWMM5 allows for a user-defined BMP reduction
efficiency to be applied to the contaminant sources, separate
from the treatment that occurs in devices. P8 and MOUSE
model two or more sediment fractions plus a dissolved fraction, and other contaminants are modelled through their association with each of the size fractions. In SWMM and MUSIC,
the concentrations in baseflow can be specified separately
from stormflow concentrations.
All the methods for representing contaminant generation
rely on empirical parameters relating to concentrations, yields,
or buildup-washoff processes. This reflects the limited knowledge of the processes and process rates for contaminant generation. The particular method has little bearing on the
suitability of the models to represent LID, except that methods
with a range of particle sizes permit more detailed representation of contaminant removal processes in devices.
2.6.3. Contaminant transport and treatment processes
In MOUSE, SWMM, and P8 contaminants are transported
through the network by treating the links as well-mixed reservoirs with first order decay, except that MOUSE can also use
the advection-dispersion equation and P8 allows for second order decay. For dissolved oxygen and biological oxygen demand,
MOUSE allows for re-aeration and interaction with bed sediments. In UVQ, contaminants are transported conservatively
in the links with no lags (the timing is not relevant as only daily
calculations are performed). In MUSIC, contaminants are transported conservatively through links but with either a time lag or
dispersion based on MuskinghameCunge routing. MOUSE
includes a sediment erosion component intended to represent
sediment deposits in pipes. For most contaminants, simple transport methods are likely to be adequate when the drainage system
is dominated by pipes.
In those models with infiltration devices (see Section 2.7),
the removal of contaminants due to infiltration is determined
from the product of the infiltration rate and the concentration

in the device. This removal processes is very relevant to

several LID devices. The infiltrated contaminants do not reemerge into the drainage system and the groundwater contamination is not considered, except that in P8 a removal efficiency in groundwater can be applied.
Most of the models include contaminant treatment in ponds,
determined using sediment settling theory (SLAMM, P8,
MOUSE, and in a simple manner, RUNQUAL), first or second
order decay (MUSIC and P8), removal fractions or output concentrations (UVQ and StormTac), or user-specified functions of
variables such as flow rate (SWMM). All these methods rely on
user inputs for rate parameters or sediment settling rates.
In MUSIC, a number of devices apart from ponds can be
modelled. Treatment in biofilter storage areas and swales is
modelled using a series of well-mixed reactors with first order
quasi-steady removal kinetics, plus removal with infiltration.
Treatment in gross pollutant traps is modelled with a concentration rating curve (inflow concentration versus outflow concentration). For buffer strips, concentration reduction efficiencies
are calculated as fixed (hard-coded) functions of hydraulic loading. The filtration efficiency of bioretention filter media is determined using fixed empirical relations based on the retention time
in the medium and particle size of the medium. The user can also
define a concentration rating curve for any device.
Some other models include treatment in LID devices and
measures apart from ponds and infiltration. In RUNQUAL,
vegetated filter strips are modelled by assuming complete
removal of sediment for strips of 30 m length with performance
reduced in proportion to the length for shorter filters, while dissolved contaminants are not removed. In SLAMM the removal
of sediment through street sweeping and catchbasin cleaning is
calculated from empirical equations. In UVQ the user specifies
a removal efficiency or required output quality for each device.
2.7. LID devices or practices
The ability of the models to incorporate LID devices and
measures, a key consideration in this review, is summarised
in Fig. 5. In many cases a model can be used to represent

nk
s
Bi
Fi o-r
ltr et
at en
io ti
n on
de ,
vi rai
Pe
ce n
rm
s ga
ea
rd
en
bl
e
s
pa
G
vi
re
n
en
g
ro
of
s

401

ta
n
ai
R

un
on
R

Im
pe
r

vi
ou
sn
es
s
Po
re
nd
du
s
ct
an
io
d
n
w
So
et
la
il
nd
pr
ot
s
ec
tio
R
n
e
co du
nt ct
am io
in n o
an f
tg
In
en
fil
tra
er
at
tio
io
n
n
tre
n
ch
O
nes
si
/b
te
or
de
es
te
n
Sw
tio
n
al
ta
es
nk
s

A.H. Elliott, S.A. Trowsdale / Environmental Modelling & Software 22 (2007) 394e405

MOUSE
MUSIC
P8
PURRS*
RUNQUAL
SLAMM
StormTac
SWMM
UVQ
WBM*

Fig. 5. LID devices and measures included in the selected models. Grey shading indicates that the model does not explicitly address the device, but could be used to
model the device. Models with an asterisk do not address water quality.

a device indirectly, even though it does not cater explicitly for

that type of device. For example, a rain tank with constant
withdrawal, an orifice outlet, and an overflow can be modelled
with MOUSE by using suitable combinations of devices and
outlet types (Kettle et al., 2004). Many standard hydrologic/
hydraulic models can be used to model the hydrological implications of LID, particularly the storm flow implications: they
can predict the effect of changes in imperviousness on storm
flows; most of them include detention ponds; they can represent the flow-retarding effects of swales either explicitly
with a detailed drainage network or in an approximate way
by adjusting the catchment routing parameters; and on-site
detention can be modelled either as small ponds or approximately by increasing the depression storage in the runoff
generation sub-model (e.g. Department of Environmental
Resources, 1999; Guther et al., 1996; Kandasamy and
OLoughlin, 1995).
All of the models in this review can be used to investigate
the effects of reducing imperviousness. They can also be used
to represent the effects of soil protection or improvement, by
altering infiltration parameters.
All the water quality models can be used to represent the
reduction of contaminant generation by altering the storm concentrations or yields, or mix of land uses. However, none of
the models deal specifically with contaminant reduction due
to the use of source-reduction practices such as altered construction or vehicle emissions controls to reduce contaminant
inputs to the urban system, which is part of the LID philosophy. Hence some major opportunities to limit contaminant
generation may not be able to be simulated explicitly.
All the models except PURRS include ponds or wetlands,
reflecting the widespread use of these devices. The models
do vary in the details of flow routing and contaminant treatment in such devices (Sections 2.5 and 2.6), which is a factor
to consider when selecting a model in catchments where ponds
and wetlands are an important component.

Most of the models can be used for infiltration devices, but

users of SWMM and MOUSE would need to represent the
infiltration by an outlet in a tank (rather than infiltration into
the soil), and StormTac does not include infiltration devices at all.
Seven of the models represent on-site detention tanks explicitly
although P8, RUNQUAL, WBM, and UVQ would only be suitable for assessing the volume effects of the tanks (not flow rates)
due to the fairly coarse timestep use in these models.
The water quality, infiltration, storage, and hydraulic aspects of swales are addressed by MUSIC, while WBM and
SLAMM model only the infiltration function of swales. The
hydraulic aspects of swales can also be modelled using links
in P8, MOUSE, and SWMM, but without infiltration (unless
it is represented as a throttled pipe outflow).
Runon (passing runoff from an impervious area to a pervious area) is allowed for explicitly in four of the models,
although only two of those models are water quality models.
In MUSIC, P8, SLAMM, and RUNQUAL, infiltration through
runon could be modelled approximately using a buffer strip or
infiltration device. MUSIC could also address contaminant
removal approximately through the buffer strip removal equation. In other models, runon could be represented coarsely by
reducing the impervious area.
Three models (MUSIC, PURRS and UVQ) are set up to
model rain tanks. PURRS gives tanks special attention, while
UVQ focuses on water budget aspects rather than flow rate.
Other models could be used to represent rain tanks using
detention tanks with a constant withdrawal rate.
Bio-retention devices or filtration devices are included only
in MUSIC and WBM, and WBM is limited to water quantity
aspects. Other flexible models such as MOUSE and SWMM
could model flow aspects of these devices indirectly and approximately using storage/infiltration devices. Bioretention
devices of the type that rely predominantly on infiltration into
the soil (without underdrains or significant evapotranspiration)
could be modelled as infiltration devices.

402

A.H. Elliott, S.A. Trowsdale / Environmental Modelling & Software 22 (2007) 394e405

None of the models include permeable paving explicitly,

but four of them could be used to represent permeable paving
by adjusting the parameters of a pervious sub-area.
Only WBM addresses green roofs specifically. Some
models could represent green roofs indirectly by changing
the soil and subsoil properties in a given area. This limited
treatment of green roofs reflects their relatively recent introduction into the range of LID devices.
2.8. User interface and model integration
The user interfaces of the models cover a range of sophistication. PURRS uses text input/output files, while P8 and
RUNQUAL add a text-based interface including menus.
SLAMM and UVQ are also essentially based on menus and
text entry, but are based on Microsoft Windows rather than
DOS. WBM also uses text-based input with menus, but is
driven with an internet-based forms/database system. StormTac is based on a Microsoft Excel spreadsheet with graphical
elements. The interfaces for SWMM, MOUSE, and MUSIC
are centred on a spatial editor, with forms-based input of information for user-selected objects representing components of
the catchment, devices, and drainage elements. MOUSE and
SWMM input files can also be edited manually. Such sophisticated and attractive interfaces are likely to promote the use of
the models in mainstream engineering and planning practice,
which could influence the uptake of LID.
Most of the models allow for a range of state and flux variables to be output in text files, along with statistical summaries, while some of the models (MOUSE, MUSIC, SWMM
and WBM) also display graphs of results. In MUSIC,
MOUSE, and SWMM, the simulation results can be accessed
through the spatial editor (by clicking on the relevant object).
Few of the models incorporate or are linked to calibration
or prediction-uncertainty models. MOUSE and PC-SWMM
incorporate automated calibration routines, although this is
for the hydrograph only. Other models allow the user to
view measured hydrographs against predicted hydrographs,
to assist with manual calibration of the flow components.
Those models with text-based input files could be optimised
with third-party software such as PEST (Watermark Numerical Computing).
3. Trends in model development and gaps
in model capabilities
There is a trend towards introducing LID devices into conventional stormwater drainage models, either by modifying
conventional models, building new models based on conventional modelling approaches, or documenting how methods
to model LID devices indirectly. For example, infiltration
source control devices have been recently incorporated into
WinDes (http://www.microdrainage.co.uk), and there are plans
to incorporate more types of distributed stormwater controls
into SWMM5 (personal communication, Lewis Rossman,
CDM). This trend is expected to continue given the institutional investment in existing conventional models.

We also expect that more types of LID devices will be

incorporated into stormwater models as information on their
performance becomes available and as new types of device
are developed. Representation of LID devices is likely to be
refined as more performance monitoring data becomes available and as the understanding of processes increases.
There is a trend to make existing models easier to use. For
example, the USEPA version of SWMM has recently been
upgraded to include a spatial editor. Database-oriented
management of model inputs and metadata, and use of GIS
(either incorporated within the main model software or linked
to the models) is growing. For example, DHI are developing
a modelling framework, MIKE-URBAN, which will use GIS
components and will drive models such as MOUSE and
SWMM. Similarly, there are plans to move MUSIC into an
integrated modelling environment with a GIS basis (personal
communication, Tim Fletcher, Monash University). As the
models become easier to use, it can be expected that their popularity will increase which may in turn promote the use of
LID. As LID components are included explicitly in such
models, LID will be considered more frequently as a mainstream design component for urban stormwater systems.
There is a trend towards the use of continuous long-term
simulation. This is in part because computers now have the
speed and storage capacity to allow long-term simulation
with reasonable run-times. It is also in part due to a realisation
of the importance of antecedent moisture conditions for small
storms, which is accommodated by continuous simulation. Indeed, most of the models we reviewed include continuous
modelling capability. We expect, however, that a need for simpler annual average models and device sizing guidelines will
re-emerge as a convenient way to encapsulate the knowledge
and results from more complex detailed models.
Management and ecosystem components are likely to be
added to core hydrology and water quality components. For
example, life-cycle costing and ecosystem effects models are
being incorporated into MUSIC (personal communication,
Tim Fletcher, Monash University).
Despite the considerable number of stormwater models that
can be applied to LID, and a commonality in process representation and model interfaces suggesting a maturing of the software in this area, several significant gaps in the capabilities of
the models remain. In our opinion, these limitations preclude
comprehensive predictions of the effects of LID on hydrology
and water quality and the resulting ecosystem effects. While
all models involve a degree of approximation and specialisation, it is important for the stormwater modeller to understand
the limitations of the models. Highlighting these limitations
also points to opportunities for further research and
development.
The models we reviewed do not address some key water
quality parameters of interest. None of the models include
temperature, despite this being an important stressor in urban
streams (Burton and Pitt, 2001), although some specialist
models for urban stream temperature have been developed
recently (e.g., ul Haq and James, 2002). Only one of the
reviewed models (MOUSE) addresses dissolved oxygen

A.H. Elliott, S.A. Trowsdale / Environmental Modelling & Software 22 (2007) 394e405

depletion, and even then the emphasis is on wastewater discharges rather than stormwater. Most of the models have limited or no ability to predict pathogenic micro-organisms or
bacterial indicators, yet these are of major concern even for
separated stormwater systems. None of the models are well
suited for the prediction of sediment loads during the earthworks phase of development.
The models we reviewed are not well suited for site or subdivision fingerprinting to maximise the use of absorbent soils
and natural features and to minimise imperviousness, or for
the selection of construction materials on a site and other
emissions controls to reduce the production of contaminants,
yet this is surely where low-impact development approaches
should start. At present such source controls must be modelled
by adjusting soil properties or contaminant load parameters.
The models are not integrated with ecosystem effects
models, limiting their ability to predict the benefits of LID
on the stream ecosystem, which is a key purpose of LID.
None of the models are linked to habitat models such as erosion or baseflow habitat models, none of them predict contaminant accumulation in streams or estuaries (effects are judged
from water column concentrations or load reduction percentages), and none are linked to bioaccumulation models for toxic
contaminants.
There is scope for improving the representation of contaminant transport and removal processes. For example, the effectiveness of various filter media on contaminant removal could
be included. There is no explicit modelling of the effects of
vegetation on settling or sorption of contaminants. The models
incorporate few, if any, chemical or biochemical processes
such as sorption/desorption and complexation, particle interactions, biological uptake reactions, or generation of organic
sediments. None of the models attempt to represent the contribution to sediment load from erosion in streams, yet this can
be a major contributor to sediment loads (Trimble, 1997).
None of the models (except MOUSE) account for storage
and release of contaminants in sediments (especially in
streams) and the associated effects on baseflow concentrations
and timing of storm loads. Incorporation of such process
would certainly increase the number of parameters in the
model, yet without such modelling the representation of transport and treatment processes remains highly empirical.
Most of the models are somewhat limited in relation to prediction of baseflow, reflecting the traditional emphasis on impervious areas and flooding. In MUSIC, SWMM, and
MOUSE, infiltration from devices such as swales or infiltration trenches is not added to the soil or groundwater moisture
stores, which somewhat limits the utility of these models for
assessing baseflow enhancement. All the models exclude
some factors that can affect baseflow, such as leakage from
the water supply network, groundwater interception by stormwater and wastewater drains, the effect of vegetation type on
evapotranspiration or interception, and regional groundwater
flows. Hence the predictions of the effects of urbanisation on
baseflow may be unreliable. Moreover, predictions of changes
in baseflow have rarely been tested. This is relevant to LID, as
maintenance of pre-development baseflow is often a key goal.

403

The models do not incorporate more sophisticated hydrological processes for prediction of storm flow from pervious
areas, such as variable source areas or macropore flow. This
potentially limits the reliability of the models for prediction
of storm runoff from pervious areas, which is of interest as
LID aims to mimic pre-development hydrology. Also, there
is a heavy reliance on calibration of conceptual parameters,
so that for applications at the development scale or smaller
where data collection for calibration would be too expensive,
the modeller must resort to experience or regional parameterisation to obtain parameter values.
Even if suitable data is available for calibration, most of the
models do not incorporate calibration techniques and methodologies, except occasionally for flow components. None of the
models are set up for automated calibration of water quality
parameters. When calibration is incorporated, the techniques
do not allow for calculation of parameter or prediction uncertainty (beyond simple sensitivity analysis).
None of the models integrate the hydrologic and water
quality predictions with costing modules, environmental risk
analysis or receiving-water models, except that StormTac incorporates a simple lake concentration model and MUSIC
compares the predicted frequency distribution of concentrations with concentration criteria.
None of the models are integrated with drawing software
for the preparation of construction drawings or site layout.
We see a role for models that allow visualisation of lowimpact development measures such as rain gardens or narrow
roads into the site layout and landscaping.
There are some unresolved questions about how much spatial detail is needed to represent on-site LID at a catchment
scale. For example, the degree to which models of on-site
devices can be scaled up to the catchment scale using lumped
representations of the devices has not been demonstrated: the
alternative is to represent each device separately. There is
a need for systematic work on the suitability of, and methods
for, such up-scaling. It may be that such distributed systems
are best modelled with a detailed representation of the system
but also by re-using model components to save on setup time
and computational effort (as in the UrbanCylce model being
developed by Kuczera and others at Newcastle University,
Australia).
There are very few documented tests of the ability of stormwater models to predict the actual effect of LID at a subdivision or catchment scale, mainly because there are difficulties
in setting up a suitable study site, especially one including
a spatial control.
Several approaches can be taken to address these limitations. Research into flow and contaminant generation and
transport processes and development of mathematical representation of these processes will improve the fundamental or
empirical representation of processes in the modes. Testing
of the performance of existing or new devices will also lead
to improvements in their representation within models. Some
gaps will require the development of new models (both sophisticated research-level models and simple models), but many
limitations can be overcome through refinement of existing

404

A.H. Elliott, S.A. Trowsdale / Environmental Modelling & Software 22 (2007) 394e405

models. Further application and testing of existing models will

also help address some of the research gaps, and this is likely
to occur as modelling of LID devices becomes more
widespread.
4. Summary and conclusions
A wide range of existing models, including conventional
stormwater models, is available for predicting at least some
of the water quality and flow effects of LID. None of the ten
models we reviewed are intended for the full spectrum of
uses that could be demanded of a model in relation to LID.
The models most commonly address the middle ground of
planning and preliminary design levels of use.
The models use a wide range of temporal resolution, from
average annual to sub-hourly. All but one of the models are capable of long-term simulation. Half of the models use lumped
catchments, while the remaining models are quasi-distributed.
The models use similar methods for generating runoff from
impervious areas, and a range of conventional methods is used
for generating runoff from pervious surfaces. Half of the
models include a groundwater component, which is relevant
for assessing effects of LID on baseflow, but the representation
of groundwater is simplified and generally untested.
Several models are limited in their ability to predict the
flow rates from small catchments incorporating LID due to
the large timestep or limited flow-routing capabilities. Such
models may still be suitable for predicting water budgets or
contaminant modelling. All but one of the models route flows
through devices.
Most of the contaminant models include sediment and
nutrients specifically, and half of them also include heavy
metals explicitly. In many cases, contaminants that are not
included explicitly in the model can be represented either
by using generic transport and treatment options, by modelling another contaminant with a similar behaviour, or by
specifying the association of the contaminant with sediment.
The models use a limited range of methods for contaminant
generation, all of which rely heavily on empirical parameters.
The models either have no contaminant routing in the drainage network, or use simple representation of contaminant
routing and decay.
The models differ in the types of LID device that are included explicitly. All the models can represent the effects of reducing imperviousness or improving soil infiltration
properties, all but one can model ponds, all the contaminant
models can represent reduction of contaminant generation,
most of the models can be used for infiltration devices, and
the majority of the models can represent on-site detention
tanks. Half of the models can represent some of the functions
of swales, MUSIC being the most comprehensive in this regard. Runon is included explicitly in four models, rain tanks
are included in three models, and only two models include
rain gardens, bio-retention devices or filtration devices. Only
one model includes green roofs explicitly, and none of them include permeable pavements. In many cases, a device which is
not represented explicitly in the model can still be modelled

indirectly by altering the parameters of other devices or combining other devices. Conventional stormwater drainage
models can be used to approximate the hydraulic aspects of
several types of device. We expect to see more LID devices
becoming incorporated into conventional stormwater models,
gradual refinement of the algorithms for representing LID devices, and extension of the range of devices as new types of
device are developed and better data are collected.
There is a trend towards incorporation of spatial editors,
GIS, and other graphical interface features into stormwater
modelling systems. Such features will likely encourage the
use of models that incorporate LID, which in turn could
encourage the uptake of LID.
There is considerable scope for improving the capabilities of
the models including: improvement of runoff generation and
groundwater components; extending the range of contaminants;
incorporation of more contaminant biochemical and physical
processes; more integration with receiving-water and ecosystem effects models; incorporation of more non-structural stormwater control measures; more linkage to calibration techniques;
testing of model predictions against field data; and investigation
of methods for and suitability of spatial and temporal aggregation methods. Such gaps and deficiencies are likely to be addressed in future model development, as the use of LID and
associated modelling becomes more commonplace.
References
Beecham, S., 2002. Water sensitive urban design and the role of computer
modelling. In: International Conference on Urban Hydrology for the
21st Century. WMO/UNESCO, Kuala Lumpur, Malaysia, pp 21e45.
Burton Jr., G.A., Pitt, R., 2001. Stormwater Effects Handbook: A Tool Box for
Watershed Managers, Scientists and Engineers. CRC/Lewis Publishers,
Boca Raton, Florida.
Chiew, F., McMahon, T.A., 1999. Modelling runoff and diffuse pollution loads
in urban areas. Water Science and Technology 39, 241e248.
CIRIA, 2000. Sustainable urban drainage systems. In: Design manual for Scotland and Northern Ireland. Construction Industry Research and Information Association, London, England.
Coombes, P.J., 2002. Rainwater tanks revisited: new opportunities for urban
water cycle management. Ph.D. thesis, University of Newcastle, Australia.
Department of Environmental Resources, 1999. Low-impact development
hydrologic analysis. Prince Georges County, Maryland. Department of
Environmental Resources, Programs and Planning Division, Maryland,
USA.
DHI, 2002a. MOUSE surface runoff models reference manual. DHI Software,
Horsolm, Denmark.
DHI, 2002b. MOUSE RDII reference manual. DHI Software, Horsolm,
Denmark.
DHI, 2002c. MOUSE pipe flow reference manual. DHI Software, Horsolm,
Denmark.
DHI, 2002d. MOUSE TRAP Version 2003 user guide. DHI Software,
Horsolm, Denmark.
Guther, R.T., Schechenberger, R.B., Blackport, W.R., 1996. Use of continuous
simulation for evaluation of stormwater management practices to maintain
base flow and control erosion. In: James, W. (Ed.), Advances in Modeling
the Management of Stormwater Impacts, vol. 5. Computational Hydraulics
International, Guelph, Canada, pp. 77e100.
Hall, M.J., 1984. Urban Hydrology. Elsevier Applied Science Publications,
Northern Ireland.
Hammer, T.R., 1972. Stream channel enlargement due to urbanisation. Water
Resources Research 8, 1530e1540.

A.H. Elliott, S.A. Trowsdale / Environmental Modelling & Software 22 (2007) 394e405
Huber, W., Dickinson, R.E., 1988. Storm Water Management Model, Version
4: Users Manual. U.S. EPA, Athens, Georgia. EPA/600/3e88/001a.
Kandasamy, J.K., OLoughlin, G.G., 1995. On-site detention in the Sydney
CBD. In: Proceedings of the Second International Symposium on Urban
Stormwater Management 1995, Melbourne, Australia. Institution of
Engineers, Australia, pp. 343e348.
Kettle, D., Diyagama, T., Shaw, N., Heijs, J., Wilson, G., 2004. Modelling of
low-impact initiatives. In: New Zealand Water and Wastes Association
Stormwater 2004 Conference, Rotorua, New Zealand. New Zealand Water
and Wastes Association, Wellington.
Larm, T., 2000. Watershed-based design of stormwater treatment facilities:
model development and application. Ph. D. thesis. Royal Institute of
Technology.
Larm, T., 2003. An operative watershed management model for estimating
actual and acceptable pollutant loads on receiving waters and for the
design of the corresponding required treatment facilities. In: 1st International Conference on Urban Drainage and Highway Runoff in Cold Climates, Riksgransen, Sweden, Lulea University of Technology and
Statues of IAHR/IWA Joint Committee on Urban Drainage.
Leopold, L.B., 1968. Hydrology for urban land planning e a guidebook on the
hydrologic effects of urban land use. In: Geological Survey Circular 554.
U.S. Geological Survey, Washington, DC.
McAlister, T., Mitchell, G., Fletcher, T., Phillips, B., 2003. Modelling urban
stormwater management systems. Chapter 13 in. In: Wong, T.H.F. (Ed.).
Australian Runoff Quality, Draft. Engineers Australia, Sydney, Australia.
Merritt, W.S., Letcher, R.A., Jakeman, A.J., 2003. A review of erosion and
sediment transport models. Environmental Modelling and Software 18,
761e799.
Mitchell, V.G., Diaper, C., Gray, S.R., Rahilly, M., 2003. UVQ: Modelling the
movement of water and contaminants through the total urban water cycle.
In: 28th International Hydrology and Water Resources Symposium,
Wollongong, NSW. Institution of Engineers, Australia.
Mitchell, V.G., Diaper, C., 2006. Simulating the urban water and contaminant
cycle. Environmental Modelling and Software 21, 129e134.

405

MUSIC Development Team, 2003. MUSIC user guide. Version 2.0. Cooperative Research Centre for Catchment Hydrology, Melbourne, Australia.
NZWERF, 2004. On-Site Stormwater Management Manual. New Zealand
Water Environment Research Foundation, Wellington, New Zealand.
Paul, M.J., Meyer, J.L., 2001. Streams in the Urban Landscape. Annual
Review of Ecology and Systematics 32, 333e365.
Palmstrom, N., Walker, W.W., Jr., 1990. P8 urban catchment model: Users
guide. Program documentation, and evaluation of existing models, design
concepts, and Hunt-Potowomut data inventory. The Narragansett Bay
Project Report No. NBP-90-50.
Pitt, R., 1998. Unique features of the Source Loading and Management Model
(SLAMM). In: James, W. (Ed.), Modeling the Management of Stormwater
Impacts, vol. 6. Computational Hydraulic International, Guelph, Ontario,
pp. 13e35.
Rossman, L.A., 2004. Storm Water Management Model. Version 5.0. National
Risk Management Research Laboratory, United States Environmental Protection Agency, Cincinnati, Ohio.
Schueler, T., 1994. The importance of imperviousness. Watershed Protection
Techniques 1, 100e111.
Singh, V.P., 1995. Watershed modelling. In: Singh, V.P. (Ed.), Computer
Models of Watershed Hydrology. Water Resources Publications, Highlands
Ranch, Colorado, pp. 1e22.
Trimble, S.W., 1997. Contribution of stream channel erosion to sediment yield
from an urbanizing watershed. Science 278, 1422e1444.
ul Haq, R., James, W., 2002. Thermal enrichment of stream temperature by
urban storm water. In: Strecker, E.W., Huber, W.C. (Eds.), Global Solutions for Urban Drainage: 9IUC. American Society of Civil Engineers.
Electronic book.
Wong, T.H.F., Fletcher, T.D., Duncan, H.P., Coleman, J.R., Jenkins, G.A.,
2002. A model for urban stormwater improvement conceptualisation.
In: International Environmental Modelling and Software Society Conference. Lugano, Switzerland.
Zoppou, C., 2001. Review of urban storm water models. Environmental
Modelling and Software 16, 195e231.