Vous êtes sur la page 1sur 6

The Institution of Engineers, Australia

28
th
International Hydrology and Water Resources Symposium
10 - 14 November 2003
Wollongong, NSW
Hydrological Vs Hydraulic Routing
Possibilities with Two-Dimensional Hydraulic Modelling
H U Rehman, Senior Engineer, Lawson & Treloar Pty Ltd
M W Zollinger, Engineer, Lawson & Treloar Pty Ltd
G B Collings, Engineer, Lawson & Treloar Pty Ltd
Abstract: Current practices in catchment-wide modelling utilise hydrological techniques for flow
routing that are generally based on the linear/non-linear reservoir or Muskingum type routing methods.
However, these methods are limited in their approach as various catchment parameters are lumped
together in a simplified manner. Some hydrological routing methods address this issue through a
semi-distributed approach where the catchment is discretised into independent lumped sub-models
and the flow from each model is combined through simple routing methods. However, there are limits
to catchment discretisation. Some of the limitations of hydrological modelling can be addressed using
the hydraulic routing of flow through the catchment. This approach has not found its place in
catchment-wide modelling so far due to excessive terrain data requirements and computer intensive
solutions. However with improvements in data acquisition techniques such as aerial surveys coupled
with enhanced computing power, the idea of hydraulic routing is becoming feasible.
Kinematic wave modelling has previously been proposed to simulate surface runoff. This type of
modelling is limited to hydraulically steep parts of the catchment and is not suitable where backwater
effects are pronounced. Dynamic wave modelling addresses these limitations. As the runoff process in
the catchment is a spatially distributed phenomenon, a fully dynamic two-dimensional hydraulic model
may provide a more realistic flow routing.
This paper provides an overview of hydrologic and hydraulic routing techniques, their relative merits
and various tools available for catchment modelling. A case is developed for the use of a commercially
available two-dimensional hydraulic model in small to medium size catchments where backwater
effects are important. This approach also simplifies the analysis of detailed flood behaviour (flood
levels, velocity, discharge) in a catchment by making the traditional hydrological modelling redundant.
Keywords: Catchment modelling, hydrological routing, hydraulic routing, two-dimensional
hydraulic model
1. INTRODUCTION
Hydrological modelling encompasses a broad
range of physical processes associated with the
land phase of the hydrological cycle. The flow
generated at the catchment outlet may pass
through various phases and contain
contribution from surface run-off, sub-surface
flow and exchange from groundwater.
Numerous mathematical models have been
proposed to define the surface and sub-surface
movement of water.
Whilst the sub-surface flow including the
infiltration processes are extremely complex
and consequently not very well defined (Beven
[2001]), the surface or overland flow is
amenable to accurate description through the
laws of conservation of mass and momentum.
For small to medium size urbanised catchments
with relatively steep slopes, the contribution of
sub-surface flow to an event-based flood
hydrograph may be small. The Hortonian
overland flow would be dominant and constitute
almost the entire flood hydrograph.
Traditional rainfall-runoff modelling employs
hydrological techniques for overland flow
description based on the storage equation, with
linear or non-linear storage function. Moreover,
these techniques use lumped catchment
parameters, where catchment characteristics
are lumped together in a simplified manner for
representation in the model.
A terrain-based model can provide a more
accurate representation of overland flow. With
advancements in terrain data acquisition and
enhanced computing powers, commercially
available two-dimensional hydraulic models
may be used to model the overland flow in the
catchment. The benefits are two-fold, the
The Institution of Engineers, Australia
28
th
International Hydrology and Water Resources Symposium
10 - 14 November 2003
Wollongong, NSW
hydraulic routing of flows is physically more
accurate and the nature of two-dimensional
modelling allows the development of a
distributed catchment model where various
landuses in the catchment are schematised
more accurately.
1.1 Investigation of Overland Flow
All hydrological models have two basic
components. The first deals with the
conversion of rainfall into run-off and the
second with routing of that run-off to the
catchment outlet. Different hydrological models
adopt varying degrees of complexity for the two
components, but in general the run-off
generation component is represented by more
complex processes that are non-linear in
nature.
This paper deals with the run-off routing
processes in the catchment. Hydrological and
hydraulic modelling has been carried out to
compare the two routing methods.
2. HYDROLOGICAL MODELLING
A hydrological model is a mathematical
representation of the physical processes of the
catchment. The models can be lumped models,
which use lumped catchment characteristics, or
they can be distributed to account for the spatial
variability of the catchment characteristics.
Further classification can be made as to
whether the model is deterministic or stochastic
in nature.
The catchment response to a rainfall event is
complex and consists of a number of processes
which essentially result in the loss of a part of
the rainfall and produces a run-off hydrograph
at the catchment outlet. The processes
accounting for rainfall losses are complex and
various mathematical representations of these
processes are available.
This paper provides a comparison of
hydrological and hydraulic routing techniques.
A brief description of various routing techniques
applied in hydrological models is described
below.
2.1 Lumped Routing
The lumped flow routing procedures of a
hydrological system are based on the continuity
equation (reference 2)
dS/dt = I(t) - Q(t) 1
which can be stated in words as change in
storage S of a hydrological system with time is
equal to the difference in the inflow I and
outflow Q from the system. There are two
unknowns in the above equation, S and Q. To
solve the above equation for a known inflow, a
storage function is used. Depending on the
nature of the system being analysed, the
storage function takes various forms. For
example, in reservoir routing using the level
pool method, the storage function is a non-
linear function of outflow only.
The Muskingum method for flow routing
considers the storage function to be linearly
related to inflow and outflow whereas in a linear
reservoir model, the storage is a linear function
of outflow.
A number of hydrological models use the theory
of linear reservoirs for flow routing. The storage
function is defined as:
S=kQ 2
ie storage S is related to outflow Q by a storage
constant k. Nash [1957] used this concept to
represent a catchment by a series of identical
linear reservoirs. Dooge [1959] and Eagleson et
al [1969] in their MIT models further refined the
concept of linear reservoir.
2.1.1 Runoff-Routing Model
Laurenson [1964] pioneered a different
approach to runoff generation from a
catchment. His model is based on routing of
rainfall excess through non-linear catchment
storage. This model has been incorporated in
two widely used rainfall-routing modelling
systems in Australia, RORB [2002] and RAFTS
[2002]. Both models are semi-distributed where
the catchment is discretised to independent
lumped sub-models and flow from each sub-
model is combined through simple routing
methods.
The storage relationship is described as
S=kQ
m
3
where
k and m are empirically derived and defined
below.
In RORB k is defined as
k=k1.k2 4
The Institution of Engineers, Australia
28
th
International Hydrology and Water Resources Symposium
10 - 14 November 2003
Wollongong, NSW
where
k1 is a constant representative of the catchment
characteristic as a whole and k2 represents the
characteristics of a particular sub-model
storage and is called the relative delay time.
The exponent m of the storage function in
RORB varies from 0.6 to 0.8.
In RAFTS k is the storage delay time and is a
function of catchment area, slope and degree of
urbanisation in the catchment. The relationship
for k was initially established for urban
catchments but has been successfully used
over a wide range of catchment types including
rural catchments.
The modelling system RAFTS has been
adopted in this paper for modelling excess
rainfall routing.
2.2 Distributed Routing
Freeze and Harlan [1969] provided the concept
of a physically based fully distributed
hydrological model, which has served as a
standard for most of the distributed modelling
systems currently available. The
comprehensive model description includes
three-dimensional surface-subsurface flow,
coupled to a two-dimensional overland flow and
a one-dimensional channel flow in a single
modelling domain.
Examples of fully distributed models include
SHE, ANSWERS, TOPOG, IHDM and
TOPMODEL. Singh [1995] provides details of a
number of distributed hydrological models.
Whilst the available surface-subsurface process
descriptions are complex and have limitations in
their application in a fully distributed model, the
surface flow process descriptions are well
established.
Surface flow or runoff is governed by the
dynamic unsteady flow equations based on the
laws of conservation of mass and momentum.
Most distributed models use a simplified form of
the fully dynamic equations of flow. The most
widely used flow equations are the kinematic
flow equations where the force of gravity is
equated to the force of friction in the momentum
equation.
Kinematic representation for overland flow has
been found to be adequate for sheet flow over
the catchment surface but may or may not be
suitable for routing flow through channels.
2.2.1 Two-Dimensional Hydraulic Models
Commercially available fully dynamic two-
dimensional (2D) hydraulic models provide an
opportunity for routing both surface and channel
flows in a catchment, provided the excess
rainfall can be determined through other
means.
In urban areas the dominant flow processes are
Hortonian in nature and can be described
through an initial loss-continuing loss (IL-CL)
rainfall loss model. The excess rainfall derived
from such a loss model can easily be
incorporated in a 2D hydraulic model for flow
routing.
A number of 2D hydraulic models are currently
available which are predominantly used for
flooding investigations of creeks and associated
floodplains. The inflow hydrographs for these
models are usually obtained separately through
other methods.
2.2.2 SOBEK
SOBEK is a fully dynamic 2D hydraulic model,
which is used for floodplain modelling. The
model is grid based and the solution scheme
utilises the finite difference method. The model
has the ability to incorporate one dimensional
(1D) channels and hydraulic structures within
the 2D domain.
Rather than representing just the floodplain, if
the entire catchment is included in the 2D
domain of the model, rainfall excess can be
directly applied as input to SOBEK. In urban
areas where overland flowpaths are
complicated and major drainage channels have
numerous hydraulic controls in the shape of
bridges and culverts, direct rainfall excess input
to the model can provide a very accurate
representation of flow throughout the
catchment.
3. FLOW ROUTING
Two hypothetical catchments have been
considered for modelling purposes. The
catchments have been modelled using RAFTS
and SOBEK. Flow hydrographs obtained from
the two modelling systems were compared. The
comparison was carried out for two different
catchment shapes; a Fern-type (long and
narrow) and a Fan-type (semi-circular)
catchment. Two different shapes were
considered to determine their impact on the
hydraulic model results. In addition, two
different catchment landuses were considered;
The Institution of Engineers, Australia
28
th
International Hydrology and Water Resources Symposium
10 - 14 November 2003
Wollongong, NSW
urban and rural. The area of each catchment
was 10 hectares and the catchment slope was
kept at 1 in 100. The two catchments are shown
in Figure 1.
Two rainfall events were adopted from
Australian Rainfall and Runoff (AR&R).
Following the recommended procedures of
AR&R, design events of 1 in 100 year and 1 in
2 year were developed for modelling purposes.
A 2-hour duration was adopted for both the
events.
In the RAFTS modelling, the recommended
parameter values were adopted for the urban
and rural landuse. In SOBEK, the catchments
were represented with a uniform roughness
with a Mannings n=0.02 for the urban
catchment and n=0.06 for the rural.
Model runs for both urban and rural catchments
were carried out. Table 1 shows the model run
details for each catchment type.
Table 1: Model Run Details
Fern-type Fan-type
Model
1 in 100
year
1 in 2
year
1 in 100
year
1 in 2
year
RAFTS *

- -
SOBEK

* RAFTS does not distinguish between catchment shapes.
3.1 Modelling Results
The modelling results from the two modelling
systems were processed and various
comparisons were carried out. Figure 2
provides a comparison of flow hydrographs
obtained from urban and rural landuse for the 1
in 100 year event for the Fern-type catchment.
A similar comparison is provided for the Fan-
type catchment in Figure 3.
The above comparisons for the 1 in 2 year
event are provided in Figures 4 and 5.
Another comparison was made to determine
the impact of the shape of the catchment.
Figure 6 provides the comparison for the two
catchment shapes for the 1 in 100 year event.
F e r n - t y p e - 1 % 2 h r
0
1
2
3
4
5
6
0 : 0 0 0 : 3 0 1 : 0 0 1 : 3 0 2 : 0 0 2 : 3 0 3 : 0 0 3 : 3 0
t i m e
f
l
o
w

(
m
3
/
s
)
S O B E K - R u r a l
R A F T S - R u r a l
S O B E K - U r b a n
R A F T S - U r b a n
Figure 2: Comparison of hydrographs for
fern-type catchment- 1% 2hr event
Figure 3: Comparison of hydrographs for
fern-type catchment - 1% 2hr event
Figure 1: Test catchments for modelling
F a n - t y p e - 1 % 2 h r
0
1
2
3
4
5
6
0 : 0 0 0 : 3 0 1 : 0 0 1 : 3 0 2 : 0 0 2 : 3 0 3 : 0 0 3 : 3 0
t i m e
f
l
o
w

(
m
3
/
s
)
S O B E K - R u r a l
R A F T S - R u r a l
S O B E K - U r b a n
R A F T S - U r b a n
Flow Direction
The Institution of Engineers, Australia
28
th
International Hydrology and Water Resources Symposium
10 - 14 November 2003
Wollongong, NSW
3.2 Discussion
The modelling results indicate that the response
from an urban catchment is similar for the two
routing techniques for both of the catchment
shapes.
In the Fan-type catchment the peak flow and
time to peak match more closely than in the
Fern-type catchment.
A reasonable match for the urban catchment is
expected as the RAFTS model parameters are
based on observed behaviour in urban
catchments.
The response for the rural catchment is
significantly different and the comparison is
worse for the Fan-type catchment than the
Fern-type. For rural catchments, in general, the
overland flow processes other than Hortonian
are quite significant. As processes such as
saturation overland flow are not represented in
the IL-CL rainfall loss model, the rainfall excess
may not be accurately determined and
consequently the flow hydrograph at the
catchment outlet may not be accurate. This is
true for both the modelling systems.
In hydrological modelling with an IL-CL rainfall
loss model, the inadequacy of detailed
catchment representation is met by model
calibration.
Similar calibration exercise can be carried out
with the hydraulic modelling, thus making it
suitable for rural catchments.
Various comparisons shown in Figures 2 to 5
indicate a similar recession curve or the falling
limb of the hydrographs. This indicates that the
representation of the catchment storage for the
two modelling techniques is quite similar.
Lumped semi-distributed runoff-routing models
such as RAFTS require knowledge of the flow
direction a-priori. In flat catchments such
information is not readily available. A 2D
F e r n - t y p e - 5 0 % 2 h r
0
0 . 5
1
1 . 5
2
2 . 5
3
0 : 0 0 0 : 3 0 1 : 0 0 1 : 3 0 2 : 0 0 2 : 3 0 3 : 0 0 3 : 3 0
t i m e
f
l
o
w

(
m
3
/
s
)
S O B E K - R u r a l
R A F T S - R u r a l
S O B E K - U r b a n
R A F T S - U r b a n
Fan-type - 50% 2hr
0
0.5
1
1.5
2
2.5
3
0:00 0:30 1:00 1:30 2:00 2:30 3:00 3:30
time
f
l
o
w

(
m
3
/
s
)
SOBEK-Rural
RAFTS-Rural
SOBEK-Urban
RAFTS-Urban
Circular Vs Rect Catchment
1%-2hr
0
1
2
3
4
5
6
0:00 0:30 1:00 1:30 2:00 2:30 3:00 3:30
time
f
l
o
w

(
m
3
/
s
)
Circ-Urban
Rect-Urban
Circ-Rural
Rect-Rural
Figure 4: Comparison of hydrographs for
fern-type catchment 50% 2hr event
Figure 5: Comparison of hydrographs for fan-
type catchment 50% 2hr event
Figure 6: Comparison of hydrographs obtained
from SOBEK for fern and fan-type catchments
The Institution of Engineers, Australia
28
th
International Hydrology and Water Resources Symposium
10 - 14 November 2003
Wollongong, NSW
hydraulic model would provide an accurate
representation of the flow behaviour in such a
catchment.
4.0 CONCLUSIONS
The results from the hydraulic routing of flows
through two catchment types were compared
with a traditional hydrological routing method.
The hydraulic routing was carried out using a
commercially available 2D hydraulic modelling
system SOBEK and hydrological analysis was
carried out using another commercially
available modelling system RATFS. The
conclusions of this study are summarised
below:
1. The fully dynamic hydraulic routing of
catchment flows is suitable where
Hortonian flow is prevalent. A fully dynamic
2D hydraulic model would provide reliable
flow behaviour when used with the IL-CL
rainfall loss model.
2. Catchment response can be complex and
difficult to model even with a physically
based fully distributed model. The reliability
of the lumped hydrological catchment
models is improved by calibrating the
model. A similar exercise can be carried out
for a hydraulic model, which can be
calibrated to the response of a given
catchment, urban or rural.
3. Where distributed modelling can not be
carried out for excess rainfall estimation
and instead IL-CL loss model is used, a
hydraulic model may provide a better
estimate of the overland flow.
4. Storage in the catchment is represented in
a similar fashion in both the modelling
systems presented above, with recession
curves very similar in shape for both rural
and urban catchments.
5. Lumped hydrological modelling requires
knowledge of the flow direction a priori. In
flat catchments, where it is difficult to define
the flow directions a 2D hydraulic model
would provide a better estimate of flow
distribution in the catchment.
6. Effect of the catchment shape on the outlet
hydrograph can only be investigated in a
2D hydraulic model. As expected, higher
peak flow is obtained from a Fan-type
catchment than from a Fern-type.
7. If 2D hydraulic modelling is used for flow
routing, separate hydrological modelling is
not required for detailed flood analysis
where flood level and velocity data is
required.
5. REFERENCES
Beven, K.J., Rainfall-Runoff Modelling The
Primer, John Wiley, New York, 2001
Nash, J.E., The form of the instantaneous unit
hydrograph, IASH Pub. No. 45(3), 114-
121, 1957
Dooge, J.C.I., A general theory of the unit
hydrograph, J. Geophys. Res., 64(2),
241-256, 1959
Eagleson, P.S., and W.O. Maddaus, A
Distributed Linear Representation of
Surface Runoff, Report No. 115,
Hydrodynamics Lab., MIT Department of
Civil Engineering, 1969
Laurenson, E.M., A catchment storage model
for runoff routing, J.Hydrol., 2, 141-163,
1964
RAFTS, Users Manual Version 6.12, XP
Software, Belconnen ACT, 2002
RORB, Users Manual Version 4.2, Department
of Civil Engineering, Monash University,
Clayton, Victoria, 2002
Singh, V.P., Computer Models of Watershed
Hydrology, Water Resources
Publication, Colorado, 1995
Freeze R.A. and R.L. Harlan Blueprint for a
physically-based, digitally simulated
hydrologic response model. J.Hydrol., 9:
237-258, 1969.

Vous aimerez peut-être aussi