modelling losses in flood estimation · flood estimation: a case study in queensland. 28th...
TRANSCRIPT
KODE
MODELLING LOSSES IN FLOOD ESTIMATION
MAHBUB ILAHEE
MODELLING LOSSES IN FLOOD
ESTIMATION
School of Urban Development Queensland University of Technology
By
MAHBUB ILAHEE B.E (CIVIL)
MASTER OF TECHNOLOGY (CIVIL)
A THESIS SUBMITTED TO THE SCHOOL OF URBAN DEVELOPMENT QUEENSLAND UNIVERSITY OF TECHNOLOGY IN PARTIAL FULFILMENT OF THE REQUIRMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY
MARCH 2005
Modelling losses in flood estimation iii
ACKNOWLEDGEMENT
The author wishes to express his appreciation and wholehearted sense of gratitude to
his principal supervisor Professor Rod Troutbeck, postgraduate research co-ordinator
Professor David Thambiratnam and associate supervisor Dr. Ashantha Goonetilleke
for their enthusiastic guidance, valuable suggestions, constructive criticisms, friendly
discussions and persistent supervision which were indispensable for the completion
of this research work
The author also wants to express his heartfelt thanks to Professor Mahen Mahendran,
Dr. Ataur Rahman, Dr. Walter Boughton, Erwin Wienmann and Peter Hill for their
valuable suggestion, advice, constant encouragement and meticulous efforts of this
research work.
Innumerable thanks are extended to Queensland University of Technology (QUT) for
providing a postgraduate research scholarship to conduct this research, Department
of Natural Resources and Mines (DNRM) and Bureau of Meteorology (BOM) for
providing data and material for the research, all the staff of School of Civil
Engineering for the facilities and assistance for the research.
The author gratefully acknowledges the continuing patience and vital sacrifices of his
beloved mother, father, wife and children and all the brothers and sisters whose love
and support were a constant source of encouragement and guidance to him.
Modelling losses in flood estimation iv
KEYWORDS Losses, Rainfall and runoff modelling, Flood estimation, Initial losses, Continuing
losses, joint probability Approach, Design Event Approach, IL–CL model,
Pluviograph station, Stream gauging station, Daily rainfall station, Baseflow,
streamflow, Study catchments, Storm losses, Design floods, Rainfall-runoff, Runoff
routing
Modelling losses in flood estimation v
ABSTRACT Flood estimation is often required in hydrologic design and has important economic
significance. For example, in Australia, the annual spending on infrastructure
requiring flood estimation is of the order of $650 million ARR (I.E. Aust., 1998).
Rainfall-based flood estimation techniques are most commonly adopted in practice.
These require several inputs to convert design rainfalls to design floods. Of all the
inputs, loss is an important one and defined as the amount of precipitation that does
not appear as direct runoff. The concept of loss includes moisture intercepted by
vegetation, infiltration into the soil, retention on the surface, evaporation and loss
through the streambed and banks. As these loss components are dependent on
topography, soils, vegetation and climate, the loss exhibits a high degree of temporal
and spatial variability during the rainfall event.
In design flood estimation, the simplified lumped conceptual loss models were used
because of their simplicity and ability to approximate catchment runoff behaviour. In
Australia, the most commonly adopted conceptual loss model is the initial loss-
continuing loss model. For a specific part of the catchment, the initial loss occurs
prior to the commencement of surface runoff, and can be considered to be composed
of the interception loss, depression storage and infiltration that occur before the soil
surface saturates. ARR (I. E. Aust., 1998) mentioned that the continuing loss is the
average rate of loss throughout the remainder of the storm.
At present, there is inadequate information on design losses in most parts of
Australia and this is one of the greatest weaknesses in Australian flood hydrology.
Currently recommended design losses are not compatible with design rainfall
information in Australian Rainfall and Runoff. Also design losses for observed
storms show a wide variability and it is always difficult to select an appropriate value
of loss from this wide range for a particular application. Despite the wide variability
of loss values, in the widely used Design Event Approach, a single value of initial
and continuing losses is adopted. Because of the non-linearity in the rainfall-runoff
process, this is likely to introduce a high degree of uncertainty and possible bias in
the resulting flood estimates. In contrast, the Joint Probability Approach can consider
probability-distributed losses in flood estimation.
Modelling losses in flood estimation vi
In ARR (I. E. Aust., 1998) it is recommended to use a constant continuing loss value
in rainfall events. In this research it was observed that the continuing loss values in
the rainfall events were not constant, rather than it decays with the duration of the
rainfall event. The derived loss values from the 969 rainfall and streamflow events of
Queensland catchments would provide better flood estimation than the recommended
design loss values in ARR (I. E. Aust., 1998). In this research, both the initial and
continuing losses were computed using IL-CL loss model and a single median loss
value was used to estimate flood using Design Event Approach. Again both the
initial and continuing losses were considered to be random variables and their
probability distribution functions were determined. Hence, the research showed that
the probability distributed loss values can be used for Queensland catchments in near
future for better flood estimate.
The research hypothesis tested was whether the new loss value for Queensland
catchments provides significant improvement in design flood estimation. A total of
48 catchments, 82 pluviograph stations and 24 daily rainfall stations were selected
from all over Queensland to test the research hypothesis. The research improved the
recommended design loss values that will result in more precise design flood
estimates. This will ultimately save millions of dollars in the construction of
hydraulic infrastructures.
Modelling losses in flood estimation vii
LIST OF PUBLICATIONS NATIONAL AND INTERNATIONAL CONFERENCE PAPERS 1. Ilahee, M. and Rahman, A. (2003).Investigation of continuing losses for design
flood estimation: A case study in Queensland. 28th Hydrology and Water Resources
Symposium, the Institution of Engineers, Australia, Wollongong NSW Australia,
10-13 November
2. Ilahee, M. and Rahman, A. (2003). Storm loss studies for the Bremer river
catchment in Australia. International Union of Geodesy and Geophysics (IUGG),
Sapporo, Japan (Only Abstract published), June 30 – July 11.
3. Ilahee, M., Rahman, A. and Boughton, W.C. (2002). New design losses in
Queensland. 27th Hydrology and Water Resources Symposium, the Institution of
Engineers, Australia, Melbourne, 20-23 May.
4. Ilahee, M., Rahman, A. and Boughton, W. C. (2001). Probability distributed initial
losses for flood estimation in Queensland. International Congress on Modelling and
Simulation, The Australian National University Canberra, 10-13 December.
5 Ilahee, M., Rahman, A. and Boughton, W. C. (2001). A new loss function for flood
estimation in Queensland. 6th Physical Infrastructure Centre Postgraduate
Conference, School of Civil Engineering, Queensland University of Technology, 19
June 2001, p. 87-96
6. Ilahee, M., Rahman, A. Assessment of design losses for flood estimation in
Queensland. Research Report, Physical Infrastructure Centre, School of Civil
Engineering, Queensland University of Technology, 2001, pp. 49.
7. Rahman, A., Ilahee, M. and Boughton, W. C. (2001). Improved fixed and
stochastic design losses for Queensland. 13th Queensland Hydrology Symposium, 6-
7 Nov, 2001, Brisbane.
Modelling losses in flood estimation viii
TABLE OF CONTENTS
Acknowledgements iii
Keywords iv
Abstract v
List of Publications vii
Table of Contents viii
List of Tables xiv
List of Figures xvi
List of Symbols xix
Statement of Original Authorship xxii
1. INTRODUCTION 1-1 1.1 Background 1-1
1.2 Scope and Objectives 1-3
1.3 Research hypotheses 1-4
1.4 Justification for the study 1-5
1.5 Outline of the thesis 1-5
2. REVIEW OF DESIGN LOSSES IN FLOOD
ESTIMATION 2-1 2.1 General 2-1
2.2 Methods based on point infiltration equations 2-1 2.2.1 Horton equation 2-1
2.2.2 Green-Ampt model 2-2
2.2.3 Philip equation 2-4
2.2.4 Richards equation 2-5
2.2.5 Applications of the theoretical equations in flood estimation 2-7
2.3 Conceptual loss models 2-7
2.3.1 General 2-7
2.3.2 Spatially lumped loss models 2-8
Modelling losses in flood estimation ix
2.3.3 Constant loss rate 2-8
2.3.4 Initial loss–Continuing loss (IL-CL) model 2-10
2.3.5 Proportional loss rate (PLR) model 2-12
2.3.6 SCS curve number method 2-13
2.4 Loss accounting in rainfall runoff models 2-15 2.4.1 Rainfall Runoff Routing using Burroughs (RORB) 2-15
2.4.2 Watershed Bounded Network Model (WBNM) 2-15
2.4.3 Urban Runoff and Basin System (URBS) 2-16
2.4.4 Australian Water Balance Model (AWBM) 2-17
2.4.5 Comments 2-18
2.5 Loss estimate for real-time flood forecasting 2-18 2.5.1 Loss models adopted in real-time flood forecasting 2-19
2.5.2 Initial loss-continuing loss and initial
loss-proportional loss models 2-20
2.5.3 Constant loss rate 2-20
2.5.4 Antecedent precipitation index (API) 2-20
2.5.5 Point infiltration equations 2-21
2.5.6 Comments 2-21
2.6 Currently recommended design loss values
in Australia 2-22 2.6.1 Past studies on losses 2-22
2.6.2 Losses recommended in ARR (I. E. Aust., 1998) 2-25
2.7 Summary 2-28
3. FLOOD ESTIMATION METHODS 3-1 3.1 General 3-1
3.2 Design Event Approach 3-1
3.3 Joint Probability Approach 3-4
3.4 Continuous Simulation 3-8
3.5 Selection of flood estimation method 3-9
3.6 Summary 3-9
Modelling losses in flood estimation x
4. STUDY CATCHMENTS 4-1 4.1 General 4-1
4.2 Criteria to select study catchments 4-1
4.2.1 Catchment area 4-1
4.2.2 Regulation 4-2
4.2.3 Record length 4-2
4.2.4 Catchment boundary 4-3
4.3 Mapping catchment boundaries and the selected
pluviograph, daily rainfall and streamflow stations 4-4
4.4 Selected catchments 4-6
4.5 Summary 4-11
5. DATA ANALYSIS 5-1 5.1 General 5-1
5.2 List of stream gauge, pluviograph and daily
rainfall stations 5-1
5.3 Streamflow data 5-4
5.4 Pluviograph data 5-6
5.5 Daily Rainfall data 5-8
5.6 Formatting stream flow, rainfall and
pluviograph data 5-10
5.7 Process to estimate the weighting average rainfall of a
catchment for more than one pluviograph and daily
rainfall stations 5-11
5.8 Selection of rainfall and streamflow events 5-14
5.9 Summary 5-16
6. METHODOLOGY FOR LOSS COMPUTATION 6-1 6.1 General 6-1
Modelling losses in flood estimation xi
6.2 Runoff generation mechanism 6-2
6.3 Method of initial loss computation 6-2
6.4 Method of continuing loss computation 6-4
6.5 Method of proportional loss computation 6-7
6.6 Method of volumetric runoff coefficient computation 6-7
6.7 Losses versus catchment size 6-8
6.8 FORTRAN program developed 6-9
6.9 Surface runoff threshold value 6-10
6.10 Reasons for the variability of calculated loss values 6-10
6.11 Summary 6-11
7. METHOD OF BASEFLOW SEPARATION 7-1 7.1 General 7-1
7.2 Review of baseflow separation process 7-1
7.3 Results 7-7
7.4 summary 7-14
8. RESULTS AND DISCUSSIONS 8-1 8.1 General 8-1
8.2 Fixed losses 8-1 8.2.1 Comparison was made between derived computed losses
and ARR (I. E. Aust., 1998) recommended losses 8-2
8.2.2 Interactions of losses with other variables 8-6
8.2.2.1 Different methods of computing continuing losses
and their comparison 8-6
8.2.2.2 Comparison of derived loss values with different
surface runoff threshold values 8-10
8.2.2.3 Variation of continuing losses with duration 8-14
8.3 Stochastic losses 8-18 8.3.1 Initial losses 8-18
Modelling losses in flood estimation xii
8.3.2 Continuing losses 8-23
8.4 Regionalisation of derived loss values 8-24 8.4.1 General 8-24
8.4.2 Basis of Queensland recommended design
losses in ARR (I. E. Aust., 1998) 8-25
8.4.2.1 Recommended loss values in ARR (I. E. Aust., 1998) 8-25
8.4.2.2 Recommended median initial loss for eastern
Queensland 8-25
8.4.2.3 Recommended median continuing loss for eastern
Queensland 8-25
8.4.2.4 Recommended initial loss for eastern Queensland 8-26
8.4.3 Comments 8-26
8.5 Validation of results 8-27 8.5.1 Calibration of runoff routing model 8-27
8.5.2 Comparison of flood frequency curves with the observed
losses and the derived losses from this research 8-28
9. IMPLICATIONS AND IMPACTS OF THE
RESEARCH 9-1 9.1 General 9-1
9.2 Implications on model development 9-1
9.3 Socio – economic implication 9-3
10. CONCLUSIONS AND RECOMMENDATIONS 10-1 10.1 Introduction 10-1
10.2 Overview of this study 10-2
10.3 Conclusions 10-7
10.4 Suggestions for future research 10-10
REFERENCES R-1
APPENDIX A A-1
Modelling losses in flood estimation xiii
APPENDIX B B-1
APPENDIX C C-1
APPENDIX D D-1
APPENDIX E E-1
APPENDIX F F-1
APPENDIX G G-1
APPENDIX H H-1
Modelling losses in flood estimation xiv
LIST OF TABLES
Table 2.1 Summary of previous studies on losses 2-23
Table 2.2 Currently recommended Design Losses
(From ARR, 1987, 1998) 2-26
Table 4.1 Stream gauge station number, their location, catchment
area and streamflow record length of the study catchments 4-9
Table 5.1 List of the selected stream gauge, pluviograph and daily
rainfall stations 5-2
Table 7.1 Effects of changing α value on CL, PL (proportional loss)
and Volumetric runoff co-efficient 7-10
Table 7.2 Effects of changing value on CL, PL (proportional loss)
and Volumetric runoff coefficient 7-14
Table 8.1 Descriptive statistics of the computed IL and CL values
(N = number of events) 8-3
Table 8.2 Comparison of computed median initial and median
continuing loss values with ARR recommended
median loss values for eastern Queensland catchments 8-4
Table 8.3 Comparison of initial loss values from Queensland and
Victorian Catchments 8-5
Table 8.4 Comparison of computed median initial and median continuing
loss values with ARR (I. E.. Aust., 1998) recommended median
loss values for western Queensland catchments 8-6
Table 8.5 Descriptive statistics of the computed CL and CL1 values
(N = number of events) 8-9
Table 8.6 Comparison between computed CL and CL1 values for all the
selected catchments 8-10
Table 8.7 Comparison of computed IL, CL and CL1 values
with different surface runoff threshold values
(N = number of events) 8-11
Table 8.8 Comparison of loss values with different
threshold values 8-13
Modelling losses in flood estimation xv
Table 8.9 Loss statistics for 15 Queensland catchments
(N = number of events, SD = standard deviation) 8-20
Table 8.10 Comparison at-site observed and generated ILs
data for the 15 catchments 8-23
Modelling losses in flood estimation xvi
LIST OF FIGURES
Figure 2.1 Constant loss rate 2-9
Figure 2.2 Initial loss - continuing loss (IL-CL) 2-11
Figure 2.3 Proportional loss 2-12
Figure 2.4 Schematic diagram of AWBM model (Boughton 1993) 2-18
Figure 3.1 Schematic diagram of the Design Event Approach 3-3
Figure 3.2 Schematic diagram of the Joint Probability Approach
(Monte Carlo simulation technique, probability-distributed
components shaded) 3-7
Figure 4.1 Example of a single catchment with a stream gauging
station 119006A, pluviograph and daily rainfall stations 4-5
Figure 4.2 Distribution of the drainage divisions and the basins
in Queensland 4-7
Figure 4.3 Locations of the 48 study catchments in Queensland 4-8
Figure 4.4 Distribution of catchment areas of the study catchments
in Queensland 4-11
Figure 5.1 Histogram showing the distribution of streamflow
record lengths 5-4
Figure 5.2 Cumulative plot showing the range and median value of
streamflow data record lengths 5-5
Figure 5.3 Histogram showing the distribution of pluviograph record
lengths 5-7
Figure 5.4 Cumulative plot showing the range and median value of
Pluviograph data record lengths 5-8
Figure 5.5 Histogram showing the distribution of daily rainfall record
Lengths 5-9
Figure 5.6 Cumulative plot showing the range and the median value of
daily rainfall record lengths 5-10
Figure 5.7 Example of the grids ranging from 40×40 to 80×80 depending
on the catchment size 5-12
Figure 5.8 Rainfall events: complete storms and storm-cores 5-16
Figure 6.1 Initial loss – Continuing loss (IL-CL) model 6-3
Figure 7.1 Comparison of baseflow separation by two different models 7-5
Modelling losses in flood estimation xvii
Figure 7.2 Example of comparison of baseflow recharge models in large
and small runoff events 7-6
Figure 7.3 Separation of streamflow components in a semi-log graph for
Event 1, when α=0.003, 0 .004, 0.005, 0.008 (Bremer River) 7-9
Figure 7.4 The baseflow separation with α = 0.004 for 3 events in the
Bremer River catchment 7-11
Figure 7.5 Baseflow separation of four events for Tenhill Creek
catchment. 7-12
Figure 7.6 Baseflow separation for the four events with α =0.0055
(Tenhill Creek) 7-13
Figure 8.1 Variation of continuing loss values with duration in all 48
selected Queensland catchments 8-15
Figure 8.2 Variation of continuing loss values with duration in 11
eastern Queensland catchments 8-16
Figure 8.3 Variation of continuing loss values with duration in 5
western Queensland catchments 8-16
Figure 8.4 Variation of continuing loss values with duration in 12
northern Queensland catchments 8-17
Figure 8.5 Locations of the 15 study catchments in Queensland 8-19
Figure 8.6 Histogram showing distribution of ILs
for catchment 143110 8-21
Figure 8.7 Histogram showing distribution of ILc for
catchment 143110 8-22
Figure 8.8 Distribution of observed CL values 8-24
Figure 8.9 Fitting the single non-linear storage model for
the Black River catchment (event 27/12/97) 8-28
Figure 8.10 Comparison of derived flood frequency curve with the
results of 48 selected catchments and ARR
for Burnett river 8-30
Figure 8.11 Comparison of derived flood frequency curve with
the results of 48 selected catchments and ARR
for Bluewater creek 8-31
Modelling losses in flood estimation xviii
Figure 8.12 Comparison of derived flood frequency curve with
the results of 48 selected catchments and
ARR (I. E. Aust., 1998) for Black river. 8-32
Modelling losses in flood estimation xix
LIST OF SYMBOLS
A = Constant (mm/h)
a = Filter parameter ( or factor).
α = The fraction of the surface runoff
a1 = Constant
a2 = Constant
a3 = Constant
B = Baseflow
BF = Baseflow (mm)
Bi = Rate of baseflow at any time
CL = Continuing loss
C1 = Rainfall (mm/h)
C2 = Rainfall (mm/h)
CL1 = Continuing loss occurred within the time t1
CWAR = Catchment weighting average rainfall
D = Rainfall duration
D1 = Daily rainfall station site1
Dc = Storm-core duration
fp = Infiltration capacity (mm/h)
fo = Initial infiltration capacity (mm/h)
fc = Final infiltration capacity (mm/h)
kf = Fitted quick response streamflow at the kth sampling instant:
=F Cumulative infiltration volume from beginning of the event (mm)
F1 = Areal reduction factor
F2 = Areal reduction factor
H0 = Null hypothesis
H1 = Alternative hypothesis
IL = Initial loss
ILs = Complete storm initial loss
ILc = Storm-core initial loss
Modelling losses in flood estimation xx
2ID = 2 year ARI intensity for the selected storm duration D 2Id = 2 year ARI intensity for the sub-storm duration d
I = Rainfall intensity
k = Exponential decay constant
k = Runoff model calibration
=sK Saturated hydraulic conductivity (mm/h)
k = Hydraulic conductivity of soil
k = Storage delay coefficient (hour)
k = Constant
LL = Lower limit
m = Non-linearity parameter
m = Runoff model calibration
=M Initial soil moisture deficit (vol/vol)
N = Number of events
P = Catchment response model and parameters
P1 = Pluviograph station site 1
PL = Proportional loss
Q = Streamflow discharge (m3/s)
QF = Quickflow, assumed to be resulted from the rainfall event, expressed in mm
RFID = Average rainfall intensity during the entire storm duration
RFId = Average rainfall intensity during a sub-storm duration
R = Total rainfall of the event expressed in average depth of rainfall in mm
S = Sorptivity of soil (mm/h1/2)
S1= Storage capacity
S2 = Storage capacity
S3 = Storage capacity
S = Spatial pattern
S1 = Stream gauging station
SD = Standard deviation
SFT = Total streamflow (mm)
S = Storage (m3)
t = Time
Modelling losses in flood estimation xxi
t2 = duration of infiltration (h)
t = Time elapsed between the start of the surface runoff and end of the rainfall event
t1 = Time elapsed between the start of the surface runoff and end of the surface
runoff (hour)
TP = Rainfall temporal pattern
Ui = Incremental value
UL = Upper limit
ky = Total streamflow
z = Space coordinate in vertical direction
=ψ Capillary suction at wetting front ( mm of water)
=θ Volumetric water content
=ψ Total metric potential
Modelling losses in flood estimation xxii
The work contained in this thesis has not been previously submitted for a degree or
diploma in any other higher education institution. To the best of my knowledge and
belief, the thesis contains no material previously published or written by another
person except where due reference is made.
Signed: ________________________________________________________ Date: _________________________________________________________
Modelling losses in flood estimation 1-1
CHAPTER 1
INTRODUCTION
1.1 Background
Flood estimation is often required in hydrologic design and has important economic
significance. For example, in Australia, the annual spending on infrastructure
requiring flood estimation is of the order of $650 million (I. E. Aust., 1998). Design
flood estimation methods can be broadly classified into two groups: streamflow
based methods and rainfall based methods. Rainfall based flood estimation
techniques are most commonly adopted in practice. These require several parameters
to convert design rainfalls to design floods. Among the parameters of flood
estimation, loss is an important factor and is defined as the amount of precipitation
that does not appear as direct runoff. Loss includes all the factors involved in
reducing the runoff during a flood event. The concept of loss includes moisture
intercepted by vegetation (interception loss), infiltration into the soil (infiltration),
retention on the surface (depression storage), evaporation and loss through the
streambed and banks. As these components are dependent on topography, soils,
vegetation and climate, the losses exhibit a high degree of temporal and spatial
variability during a rainfall event. Most loss models do not directly account for
individual components such as interception, depression storage and transmission
losses directly.
In design flood estimation, simplified lumped conceptual loss models are commonly
used because of their simplicity and ability to approximate catchment runoff
behaviour. Secondly, the detailed parameters needed for calculating individual loss
components are generally not available. This is particularly true for design loss
which is probabilistic in nature and for which complicated theoretical models may
not be required. In Australia, the most commonly adopted conceptual loss model is
the initial loss-continuing loss model (I. E. Aust., 1998; Hill et al., 1996a,b; Rahman
et al., 2000). The initial loss occurs prior to the commencement of surface runoff,
and can be considered to be composed of the interception loss, depression storage
Modelling losses in flood estimation 1-2
and infiltration that occur before the soil surface is saturated. In design rainfall
events, the continuing loss is computed as the average rate of loss that occurs up to
the end of the rainfall event, after the initial loss is satisfied.
Losses are important inputs to rainfall runoff models. However, the paucity of
information on design losses constitutes one of the greatest weaknesses in Australian
flood design (Pilgrim and Robinson, 1988). The design loss values, which are
available for the initial loss and recommended in Australian Rainfall and Runoff (I.
E. Aust., 1998), exhibit a wide range, which makes it difficult to select an
appropriate value for a particular design application.
Many studies (such as Hill and Mein, 1996; Waugh, 1991) have found that the use of
the design losses recommended in ARR (I. E. Aust., 1987) with the design rainfalls
results in overestimation of design peak flows, when compared with a frequency
analysis of recorded peak flows. This indicates that, for many catchments, the design
losses recommended in ARR (I. E. Aust., 1987; I. E. Aust., 1998) are too low.
There are two inadequacies in the current recommended loss values in ARR (I. E.
Aust, 1998), most of which were derived from analysis of large flood events. The
selection of high runoff events for loss derivation is biased towards wet antecedent
conditions. That is, loss tends to be low. The second inadequacy is that storm losses
do not account for the nature of design rainfall information, which is the design
intensity-frequency-duration and design temporal patterns of rainfall. This design
rainfall information has been derived from a rainfall burst within longer storms, but
losses are derived from complete storms. Since antecedent rainfall will pre-wet the
catchment, losses derived for storms will tend to be too high for application to bursts.
It is recognised in ARR (I. E. Aust., 1998) that these two inadequacies have opposite
effects, and it is assumed by users of the current design loss values that they
compensate for each other.
Modelling losses in flood estimation 1-3
1.2 Scope and Objectives
The scope and objectives of the research work are described here.
The scope of the proposed research was:
• This research was restricted to the estimation of losses for rural catchments and
therefore urbanised catchments were not considered.
• The selected catchments size was small to medium, therefore large catchments
were not considered.
• The streamflow needed to represent ‘natural’ flow, with no large diversions or
flow regulated by reservoirs.
• All the catchments which were used in this research work were selected from
Queensland only.
• Baseline data such as rainfall, streamflow, catchment boundaries and catchment
areas were obtained from relevant Commonwealth and State government
agencies.
• Commercially available software was used for processing data.
The objectives of the proposed research were:
• To identify the impact of losses, its distribution and its functional form for design
flood estimates.
• To develop a loss function to be used in design flood estimation. The new loss
function would account for the probability-distributed nature of both initial and
continuing losses and their variation with rainfall duration.
• To regionalise the derived loss values for Queensland so that the loss distribution
can be derived for ungauged catchments in Queensland.
• To recommend new design loss values for Queensland, this will ultimately save
millions of dollars in the construction of hydraulic infrastructures.
Modelling losses in flood estimation 1-4
1.3 Research hypotheses It was hypothesised that a new loss function that considers continuing loss as a
probability-distributed variable and variation of the continuing loss with rainfall
duration would result in significant reduction of uncertainty and bias associated with
design flood estimates. The nature of the variation of continuing loss with rainfall
duration would be identified and incorporated in the new loss function. Thus,
continuing loss would be a function of rainfall duration as given by the following
equation:
)(DfCL = (4.1)
The hypotheses to be tested were:
(H0 = Null hypothesis, H1 = Alternative hypothesis)
(a) H0: The new loss value provides more accurate design flood estimates
over the method recommended in ARR (I. E. Aust., 1998).
H1: The new loss value does not offer any improvement over the
existing method.
(b) H0: The underlying probability distribution for initial loss is 4–parameter
Beta distribution.
H1: The underlying probability distribution for Initial loss is not 4-parameter
Beta distribution.
(c) H0: The initial loss of whole of Queensland can be described by a single
probability distribution.
H1: The initial loss of whole of Queensland can not be described by a
single probability distribution.
(d) H0: The continuing loss of whole of Queensland can be described by a
single probability distribution.
H1: The continuing loss of whole of Queensland cannot be described by
a single probability distribution.
Modelling losses in flood estimation 1-5
1.4 Justification for the study
Loss is a vital input to rainfall runoff models, but there is inadequate information
and design data on losses in Queensland catchments. There are two inadequacies
identified in the currently recommended design loss values as discussed in Section
1.1. These inadequacies result in over estimation of design floods. Ultimately the
overestimation of design floods leads to the construction of over design and
expensive hydraulic infrastructures. It is therefore important to estimate suitable
design losses, which would ultimately save millions of dollars in the construction of
hydraulic infrastructures. Another reason for this research work is, a single
representative design loss value (mean or median value) is used from a wide range of
recommended losses with the Design Event Approach, as recommended by ARR (I.
E. Aust., 1998). This is likely to introduce a high degree of uncertainty and possible
bias in the resulting flood. It is therefore important to formulate an improved loss
function that can lead to more accurate design flood estimates.
1.5 Outline of the thesis
The thesis contains ten chapters. Chapter 2 discussed the method of infiltration, loss
models, recommended losses for Queensland and the review of design losses in flood
estimation. Different flood estimation methods are covered in Chapter 3. Chapter 4
presented the criteria to select the study catchments. Chapter 5 discussed how
pluviograph, daily rainfall and streamflow data were collected, formatted and
analysed for this research. The adopted methodology to compute the initial loss and
continuing loss of a rainfall streamflow event is detailed in Chapter 6. Chapter 7
describes how baseflow separation was executed from the streamflow, which was
necessary to compute continuing losses. Chapter 8 provides the results of all the
investigation performed during the analysis and compares the derived design loss
values of this research with the recommended design losses for Queensland. The
implications and impacts of this research work are documented in Chapter 9. The
conclusions of this research and the recommendations for future research and
development are outlined in Chapter 10.
Modelling losses in flood estimation
2-1
CHAPTER 2
REVIEW OF DESIGN LOSSES IN FLOOD
ESTIMATION
2.1 General
Rainfall-based flood estimation techniques are most commonly adopted in practice.
These require several parameters such as rainfall duration, intensity, temporal pattern
and loss. Selection of an appropriate loss value is very important. A low loss value
results in over estimation of design floods. A high loss value results in under
estimation of design floods. Of all the parameters, loss is the most important factor
and is defined as the amount of precipitation that does not appear as direct runoff.
Loss includes all the factors involved in reducing runoff during a flood event.
Losses during rainfall events can be represented by several methods. There are
mathematical equations based on point infiltration to estimate losses. However, in
practice losses are frequently represented by conceptual models. This chapter
describes the point infiltration equations used in loss estimation, followed by a
discussion on conceptual loss models. It also covers how losses are accounted for in
rainfall runoff models. It also discusses how loss is estimated for real-time flood
forecasting and the most commonly adopted loss models in real-time flood
forecasting. Finally it describes the currently recommended design losses for various
regions in Australia.
2.2 Methods based on point infiltration equation 2.2.1 Horton equation
Horton (1940) suggested that the infiltration capacity is a decay process as the soil
voids become exhausted. He described the infiltration rate at any time from the start
of an adequate supply of rainfall by the equation 2.1:
Modelling losses in flood estimation
2-2
pf = cf + ktc eff −− )( 0 (2.1)
where
fp = Infiltration capacity (mm/h)
fo = Initial infiltration capacity (mm/h)
fc = Final infiltration capacity (mm/h)
t = Time
k = Exponential decay constant.
fo is the initial infiltration rate at time t = 0 depending on soil moisture content and
soil type. fc and k are dependent on soil type and vegetation. The limitation of this
equation is that it is only applicable for shallow ponded conditions. Parameters are
difficult to predict because they have no physical significance (Mein and Larson,
1971). This infiltration equation is generally suitable for small catchments but not for
larger catchments. This is because as fc and k are dependent on soil type and
vegetation, larger catchments tend to be heterogeneous and the soil type and
vegetation do not remain the same throughout the catchment.
2.2.2 Green–Ampt model
A number of theoretically based infiltration models have evolved over the years.
Many of these models are based on the work performed by Green-Ampt (1911). The
Green-Ampt (G-A) method of infiltration estimation is based on Darcy’s law. This
method is based on the following assumptions:
(a) The soil surface is covered by a pool of water whose depth can be neglected.
(b) There is a definable wetting front in the soil, which separates a uniformly wetted
infiltrated zone from a totally dry uninfiltrated zone.
(c) Once the soil is wet, the water content in the soil does not change.
(d) A negative pressure is present at the wetting front.
Modelling losses in flood estimation
2-3
The G-A infiltration equation is described as:
)1( FM
sp kf ψ+= (2.2)
where
=pf Infiltration capacity (mm/h)
=sk Saturated hydraulic conductivity (mm/h)
=M Initial soil moisture deficit (vol/vol)
=ψ Capillary suction at wetting front ( mm of water)
=F Cumulative infiltration volume from beginning of the event (mm).
Mein and Larson (1971) showed that the Green-Ampt (G-A) model could be adapted
for a constant intensity rainfall at the surface (rather than ponded conditions). Chu
(1978) extended the application to an unsteady rain condition by shifting the time
scale to account for the effect of cumulative infiltration before ponding time. Mein
(1980) showed that G-A model can also be used successfully with variable rainfall by
shifting the time scale to account for the effect of cumulative infiltration before
ponding time. Brakensiek (1970) extended the model to allow for multiple soil
layers. Bouwer (1969) and Onstad et al. (1973) suggested modifications in capillary
suctions. Skaggs and Khalil (1982) outlined estimation procedures for the parameters
in the G-A equation. Todd (1980) had done the verification of the assumptions in the
derivation of G-A equation. Moore et al. (1981) mentioned that ks should be modified
for the effect of air entrapment in the field and surface conditions. Van Mullem
(1991) applied the G-A model to 12 catchments and it was found that the G-A model
did not predict peak flows well. Rajendran and Mein (1986) accounted for spatial
variability of infiltration rates in a catchment by using a scaling factor applied to
saturated hydraulic conductivity. Although the approach used by Rajendran and
Mein (1986) takes into account the spatial variability of infiltration, it does not
consider the effect of rainfall excess flowing and rainfall infiltration from one
subarea to another subarea. Also Smith and Hebbert (1979) showed that relative
positions of ponding regions are important in determining the catchment infiltration
loss.
Modelling losses in flood estimation
2-4
Since sK is a multiplier of other terms in the G-A equation (Equation 2.2), the
prediction accuracy of the equation is most sensitive to the value of the sK
parameter. It is important to note that sK value varies with the soil texture of the
catchment. Nandakumar et al. (1994) noted that for sand, sK is 23.56cm/h and for
clay sK is 0.06 cm/h. It is difficult to predict a suitable sK value for the whole
catchment, as catchments are heterogeneous in nature. For loamy sand sK is 5.98cm/h
and for sandy loam sK is 2.18cm/h. Hence there is no average value which can be
applicable for the heterogeneous catchment. To find a suitable sK for the whole
catchment is one of the limitation for this G-A infiltration equation. The other
limitation of the equation is that soil infiltration cannot be uniform throughout the
catchment. For example where there is a crack, hole or tree root in the catchment, the
infiltration rate in one part of the catchment will be different than another part of the
catchment. Mein and Larson (1971) noted that the variables in this G-A infiltration
equation are all predictable. Bouwer (1966) assumed ψ to be the water entry value,
which was reported to be approximately half the air entry value. Philip (1957a)
considered ψ to be the height of capillary rise in the soil. Mein and Larson (1971)
noted that the variables in G-A infiltration equation have physical significance,
although determining ψ causes some difficulties.
2.2.3 Philip equation
Infiltration capacity depends on many factors such as soil type, its moisture content,
organic matter in it, vegetative cover and season. Of the soil characteristics affecting
infiltration, noncapillary porosity is perhaps the most important. Porosity determines
storage capacity and also affects resistance to flow. Thus, infiltration tends to
increase with porosity. For a homogeneous soil with a uniform initial moisture
content and ponded conditions at the surface, Philip (1957) developed a simple
formula for infiltration rate related to time:
22 AttSF += (2.3)
Modelling losses in flood estimation
2-5
where
F = Cumulative infiltration (mm)
t2 = Duration of infiltration (h)
S = Sorptivity of soil (mm/h1/2)
A = Constant (mm/h)
The monash rainfall-runoff model (Porter and McMahon, 1971) is based on the
Philip equation. Sharma and Seely (1979) used the Monte-Carlo Simulation method
to evaluate the effect of spatial variability of parameters S and A and they concluded
that an average infiltration curve based on average parameters would underpredict
rainfall excess. Maller and Sharma (1980) studied the effect of spatial variability of
the parameters S and A on ponding time using analytical methods. For log-normally
distributed parameters, the predicted ponding time was also found to have the log-
normal distribution. As the log-normal distribution is highly skewed and with a long
tailed there is a high probability that ponding of some areas of the catchment will
occur quite late. As a result, this will affect the catchment rainfall excess dependent
on location of such areas (Smith and Hebbert, 1979).
Unfortunately, computing the parameters of the Philip equation is difficult. Also the
parameter values are more commonly obtained by curve fitting (Mein and Larson,
1971). A further constraint is the assumption of ponded conditions at the surface,
thus an event with a high infiltration rate cannot be represented.
Nandakumar et al. (1994) noted that parameter values of the Philip equation are
related to the soil texture as they have physical meaning. They noted that several
varying interpretations of the parameter A values are found in Singh (1989), such as
the value of parameter A ranges from 1/3 sK to sK . The sK is the saturated hydraulic
conductivity.
2.2.4 Richards equation
Richards equation is a physically based mathematical model of flow in porous media.
It is derived by combining Darcy’s law and the continuity equation. This equation is
Modelling losses in flood estimation
2-6
used for the derivation of vertical unsaturated soil moisture flow. (e.g., Childs,
1969)
The one dimensional form of the Richards equation (e.g., Childs, 1969) is stated as
given below:
( ) ⎥⎦
⎤⎢⎣
⎡⎟⎠⎞
⎜⎝⎛ −∂∂
∂∂
=∂∂ 1
zk
ztψψθ (2.4)
where
=θ Volumetric water content
t = Time
z = Space coordinate in vertical direction
k = Hydraulic conductivity of soil
=ψ Total metric potential
Numerical solutions of Richards equation have been combined into many physically
based computer packages, which simulate individual hydrological processes. This
equation provides a commonly accepted basis for detailed studies of soil water
movement, but its solution is computationally complex. Similarly, simulating field
conditions for infiltration is highly complicated because of initial and boundary
conditions are not constant and soil characteristics vary with time and space.
Recently, more practical methods have been developed (Kool et al.,1985) which
have reduced computational limitations. A further development has been undertaken
by Campbell (1985) and Ross (1990a) on Richards equation. TOPOG is a
hydrological model developed by CSIRO Division of Water Resources (CSIRO,
1992). TOPOG is modeled by a mixed form of the Richards equation. The computer
package SWIM provides a numerical solution of the Richard’s equation for cases
involving infiltration, redistribution and evapotranspiration of soil water (Ross,
1990b). This equation is not suitable for general application as the numerical solution
is rather complex, and the required soil data are difficult to obtain (Mein and Larson,
1971).
Modelling losses in flood estimation
2-7
2.2.5 Applications of theoretical equations in flood estimation
The theoretical equations mentioned above are suitable for small catchments and
each theoretical equation has its own limitations. Horton’s equation is applicable for
shallow ponded conditions. In the Green-Ampt equation, assumption is made for
ponded conditions, a constant matric potential at the wetting front and uniform
moisture content which are hardly satisfied in the real situations. Philip’s equation is
based on homogeneous soil with a uniform initial moisture content and ponded
conditions at the surface. Again, these assumptions are hardly met in real conditions.
In Richards equation, input data (soil moisture characteristics) are not commonly
available and the solution methods are computationally demanding.
A pilot study was undertaken by the CRC for Catchment Hydrology on nine
catchments to evaluate if the application of a ‘theoretically correct’ loss model based
upon a point infiltration equation (Green-Ampt) provided superior results to the
simplified models at the catchment scale. Although the Green-Ampt equation was
able to be successfully applied to each catchment, the results were not on average
superior to those produced using the simplified lumped conceptual loss models (Hill
et al., 1998). Hence simplified lumped conceptual loss model was used in this
research work to estimate loss values.
2.3 Conceptual loss models 2.3.1 General
In hydrology, design flood estimation is a common problem. Estimates of design
flood can be made using frequency analysis of the streamflow data if adequate data is
available. Design floods are typically estimated from design storms either using the
Rational Method for peak flows or by using routing models in association with loss
models for flow hydrograph estimation. No physical process involved in runoff
generation is considered in the Rational Method for peak flow estimation. Design
peak runoff rate of a catchment is related to a storm of specified intensity using
runoff coefficients.
Modelling losses in flood estimation
2-8
A loss model is applied in rainfall routing models to convert gross rainfall to rainfall
excess. Rainfall excess is routed using an appropriate routing model to obtain the
flood hydrograph. Design flood hydrographs are needed where storage is significant
or where the duration of flooding is required. The purpose of design loss is to
achieve a flood with a given Annual Exceedance Probability (AEP) from a design
rainfall with the same AEP. Since actual losses vary from event to event, design loss
can be viewed as probabilistic or statistical estimates of the most likely value
(Nandakumar et al., 1994).
2.3.2 Spatially lumped loss models
The term “storm losses” include interception, depression storage and infiltration
losses, which have been conceptualised in simple forms. These types of
conceptualised models do not consider the spatial variability or the actual temporal
pattern of storm losses. Consideration of the model parameters are considered using
the total catchment response i.e. runoff. However, spatially lumped loss models are
widely used because of their simplicity and ability to approximate catchment runoff
behaviour.
Some of the most frequently used methods in ARR (I. E. Aust., 1998) for spatially
lumped losses include:
(i) Constant loss rate
(ii) Initial loss-continuing loss
(iii) Proportional loss rate
(iv) SCS curve number procedure.
2.3.3 Constant loss rate
Constant loss rate, where the rainfall excess is the residual left after a selected
constant rate of infiltration capacity is satisfied. In this method, a constant loss rate is
subtracted from the design storm as shown in Figure 2.1. ARR (I. E. Aust., 1998)
mentioned that for design case, use of constant loss rate would be more appropriate
for large storms from which significant runoff is likely to occur. Flavall and Belstead
Modelling losses in flood estimation
2-9
(1986) applied this model to an area in the Kimberly region of Western Australia
with large extents of bare rock and shallow sand cover and the constant loss rate
varied from 2.0 to 4.8mm/h for a range of average recurrence intervals (ARI). Flavall
and Belstead (1986) mentioned that they applied IL-CL loss model for most of the
catchments outside the south west region of Western Australia, but constant loss rate
was applied for the Kimberly region of Western Australia because of low losses and
high rates of runoff.
The constant loss rate is not suitable for storms which produce low rates of runoff.
Also the constant loss rate method does not allow initial loss during the rainfall
event. Hence in the constant loss rate method the matching of initial rise of the
observed hydrograph is difficult.
Nandakumar et al. (1994) noted that this model does not closely conceptualise the
actual infiltration process for many Australian catchments, it can be used in
catchments that produce high runoff from storms. In Queensland most of the
catchments showed high loss rates and can not produce high runoff from storms
unless the storm is very large. Hence the constant loss rate model is not suitable for
many of the Queensland catchments loss estimations.
Figure 2.1 Constant loss rate
Modelling losses in flood estimation
2-10
2.3.4 Initial loss–Continuing loss (IL-CL) model
This model (Figure 2.2) has been widely used in Australia (I. E. Aust., 1998), due
both to its simplicity and to its generally better approximation of the temporal pattern
of storm losses than the constant loss rate approach. Initial loss is the loss that occurs
in the early part of the storm, prior to the commencement of surface runoff. Thus, it
can be considered to be composed of the interception loss, depression storage and
infiltration that occurs before the soil surface is saturated. Continuing loss is the
average rate of loss during the remaining period of the storm.
Cordery (1970a) found that initial loss (IL) is highly correlated with water content
present in the catchment prior to rainfall i.e. antecedent precipitation index (API),
and suggested that IL is an important factor in low rainfall areas, but can be
neglected in areas where mean annual rainfall is greater than 1270mm. Laurenson
and Pilgrim (1963) presented 150 continuing loss rate values for 24 catchments in
South Eastern Australia. The values varied markedly from catchment to catchment,
and were found to be significantly influenced by antecedent wetness and season. The
median values of continuing loss rate for the 24 catchments were between 0 to
5mm/h, with half of them equal to or less than 2.5mm/h.
Modelling losses in flood estimation
2-11
Figure 2.2 Initial loss-continuing loss (IL-CL)
Flavell and Belstead (1986) used the IL-CL model to obtain design losses for
catchments in Western Australia. They found that IL decreased with ARI in the
region where flooding tends to increase towards the end of winter as catchments
become wetter with the rarer events being associated with high antecedent wetness.
In the other regions, IL increased with ARI up to 10 years, and then decreased. A
minimum IL of 20mm was obtained with an ARI of 2 years for loamy soil areas, and
a maximum IL of 98mm was derived with ARI of 10 years for sandy soil catchments.
This indicates the broadness of the range of values of this parameter. For the same
data, CL varied from 3mm/h to 5mm/h.
Using the design rainfall given in ARR (I. E. Aust., 1987), Walsh et al. (1991)
estimated design IL values for 22 catchments in NSW. An average CL rate of 2.7
mm/h was used to obtain the IL for a range of ARIs. The derived IL values varied
from 15 to 50mm and were dependent on the degree of nonlinearity used in the flood
hydrograph model (RORB). The derived IL values also showed some degree of
dependency on design rainfall pattern.
Modelling losses in flood estimation
2-12
Cordery and Pilgrim (1983) found no relationships between catchment characteristics
and median continuing loss rate. They concluded that the difficulty of obtaining
relationships between loss model parameters and catchment characteristics was due
to: (i) the probabilistic nature of the design losses, (ii) the inadequacy of the loss
models for conceptualising real catchment processes both, temporally and spatially.
2.3.5 Proportional loss rate (PLR) model
For this model, loss is assumed to be a fixed proportion of storm rainfall (Fig 2.3). In
Australia, the approach is generally used in conjunction with an initial loss for rural
catchments. The model performs well on catchments where runoff is generated from
particular areas of the catchment, not from all over the catchment. Since loss is
proportional to the rainfall after surface runoff has commenced, the surface runoff is
also proportional to the rainfall (I. E. Aust., 1998).
02
468
101214
1618
1 2 3 4 5 6 7 8 9
Time (hour)
Rai
nfal
l and
loss
(mm
)
Loss
Rainfall excess
Figure 2.3 Proportional loss rate
Harvey (1982) found the PLR model, with an initial loss model, performed well in
South Western Australia. Flavell and Belstead (1986) used the PLR model to derive
design losses in forested catchments in Western Australia. PLR was related to the
Modelling losses in flood estimation
2-13
steamlength, the percentage of forest cleared and the mean annual rainfall. The
relationship for a given ARI was given in the form:
PLR = k 10 a 1 c pa 2 La 3 (2.5)
where
k, a1 , a 2
and a3 are constants, depending on the type of forest and soil.
L = Steam length (km),
C = The percentage of forest cleared
P = Mean annual rainfall (mm)
The statistical significance level of a1 , a 2
and a3was derived and it was found that
most of the fitted regression equations were significant at 5.0% level or a lower level
for the catchments considered.
Dyer et al. (1994) compared the performance on the initial loss–proportional loss and
initial loss–continuing loss models on 24 catchments and found that the initial loss-
proportional loss gave a better result between observed and calculated hydrographs
using the RORB model.
Although proportional loss model closely simulates the source area concept, it has
not been recommended in ARR (I.E. Aust, 1998), except for some regions of
Western Australia. There is virtually no guidance in ARR (I.E. Aust, 1998), as to
suitable values of proportional losses for design. It can be concluded that this model
needs further investigation for application in other parts of Australia, especially in
forested catchments where source area runoff generation is dominant.
2.3.6 SCS curve number method
The U.S. Soil Conservation Services (SCS) curve number procedure is widely used
for estimating stream flow volumes for small to medium sized watersheds in the
United States (U.S. Soil Conservation Services, 1985). It was originally developed to
Modelling losses in flood estimation
2-14
estimate runoff volume and peak discharge for the design of soil conservation works
and flood control projects, but later extended to estimate the complete hydrograph.
The procedure, which is empirical, aims to provide a consistent basis for estimating
the amount of runoff under varying land use and soil types. The primary input
parameter is a runoff curve number (CN) defined in terms of soil type, antecedent
moisture conditions, land use treatment and hydrological condition of the catchment.
The accuracy of the method for simulating the runoff volume is largely determined
by the selection of the appropriate curve number, and accurate estimation of
watershed antecedent moisture conditions (Boughton, 1989).
The CN value is determined from the watershed characteristics, using tables
contained in U.S. Soil Conservation Service (1985) or other texts. From this, the
depth of runoff from a given rainfall is determined. The CN method has been
extended to calculate rates of runoff. For this, the peak discharge is estimated using a
triangular approximation to the hydrograph, corresponding to a uniform rainfall
excess. For a complex storm, rainfall excess over different time intervals may be
computed separately and the resulting triangular hydrographs superimposed to obtain
the peak discharge.
The SCS curve number procedure has been adapted for application to a wide range of
catchments, such as urban and semi-arid areas. It has also been incorporated in a
number of computer programs to estimate runoff hydrographs for rural and urban
catchments.
Titmarsh et al. (1989) carried out extensive testing of the SCS method using data
from 140 catchments in two regions of Eastern Australia. They also found that the
probabilistic CN values were fairly weakly related to catchment characteristics, such
as percentage of area of the catchment cultivated, but with a much smaller range of
values than indicated by the US SCS recommendations. The relationship between CN
values estimated from catchment characteristics by the US procedure, and the
probabilistic values derived from Australian data, was found to be very poor.
Modelling losses in flood estimation
2-15
2.4 Loss accounting in rainfall runoff models
2.4.1 Rainfall Runoff Routing using Burroughs (RORB)
The Rainfall Runoff Routing using Burroughs (RORB) model (Laurenson and Mein,
1997) provides for estimating the outflow hydrograph from a catchment by
modelling the drainage network as a series of conceptual nonlinear storages. It allows
for the subdivision of the catchment based on catchment divides and drainage lines.
Each division is independent of rainfall inputs, allowing for spatial and temporal
variation of rainfall. As outlined by Mein et al. (1974), nonlinear conceptual storages
are used to route the rainfall excess along the reaches of the drainage network to the
catchment outlet.
ARR (I. E. Aust., 1998) recommends the RORB model for estimating outflow
hydrographs and uses IL-CL loss model. The IL is taken from the results of the
calibration of the model i.e. a representative value (either mean or median) from the
available values. Alternatively, the IL values recommended in ARR (I. E. Aust.,
1998) are used. The problem in selecting an appropriate value of IL or CL is that the
observed or recommended values in ARR (I. E. Aust., 1998) cover a wide range and
it is not easy to select the most appropriate value for a particular application. The use
of arbitrary value of loss is a major source of uncertainty in flood estimation by
RORB and similar models.
2.4.2 Watershed Bounded Network Model (WBNM)
Watershed Bounded Network Model (WBNM) is an event based nonlinear runoff
routing model for calculating a flood hydrograph from a rainfall hyetograph (Boyd et
al., 2001). The model structure is based on the geomorphology of the catchment and
its stream network, and model nonlinearity and lag parameter values are based on
measured catchment lag times. WBNM permits the user to specify a design rainfall
spectrum by accepting a range of standard ARIs and durations for analysis in a
particular run. The model also includes the storm temporal patterns for all Australian
rainfall zones, for ARIs from 1 to 100 years and burst durations from 10 minutes to
Modelling losses in flood estimation
2-16
72 hours. Once the rainfall hyetograph at each gauge has been read in or constructed,
the model calculates the hyetograph for each subcatchment using the temporal
pattern of the nearest gauge, and an average intensity at the gauge calculated from
surrounding gauges weighted according to the inverse of the squared distance from
the subcatchment.
In relation to loss values, WBNM has an advantage over other loss models because it
can handle five alternative loss models as listed below:
(i) Initial loss-continuing loss rate
(ii) Initial loss-loss rate varying in steps
(iii) Initial loss-runoff proportion
(iv) Horton continually time varying loss rate
(v) Green-Ampt time varying loss, with recovery after period of low rainfall
Once a rainfall loss model is selected, WBNM calculates the excess rainfall
hyetograph for each subcatchment. WBNM is not based on probability distributed
loss values but on a single loss value similar to RORB, though WBNM can
accommodate different loss values for different areas.
2.4.3 Urban Runoff and Basin System (URBS)
The Urban Runoff and Basin System (URBS) runoff routing model is based on a
network of subcatchments whose centroidal inflows are routed along a prescribed
routing path to generate runoff (Carroll, 1994). The catchment variables used by
URBS are, stream length, catchment area, channel slope, catchment slope, fraction
urbanised, fraction forested and channel roughness. This model requires that the
stream length is specified to define the extent of catchment routing, and that
catchment area is specified to determine excess rainfall. All other variables are
optionally included in the modelling process at the discretion of the user. There are
two runoff routing models; the basic model (RORB based) and the split model. The
split model separates the channel and catchment routing processes.
The URBS model uses four types of loss models; the first three are simplistic models
whose parameters can vary for each sub-catchment whilst the fourth model is based
Modelling losses in flood estimation
2-17
on losses calculated by a suitable water balance model such as the AWBM model
(Boughton, 1993). The first three loss models are;
(i) Initial loss - continuing Loss model (IL/CL)
(ii) Proportional Loss model (IL/PL) and
(iii) Manly-Philips Loss model.
2.4.4 Australian Water Balance Model (AWBM)
Boughton and Carroll (1993) have combined the AWBM (Boughton, 1993) with the
URBS runoff routing model (Carroll, 1992). AWBM is a partial area saturation
overland flow model, which allows modelling in hourly/daily steps for spatially and
temporally variable source areas of surface runoff. The partial area concept is
modelled by adopting different soil storage capacities over a catchment area, with
computation based on a conceptual soil moisture accounting procedure.
As shown in Figure 2.4, this model uses three capacities which allow for different
source areas of surface runoff. Surface runoff occurs when one or more of the stores
are filled and overflow occurs. A fixed proportion of the surface runoff is diverted
for recharge of baseflow storage with baseflow discharge at time step t being
proportional to that at time step t-1. These three storages S1, S2, S3 and baseflow
recharge are considered as losses in flood estimation. The rainfall excess from partial
areas of the catchment calculated from the calibrated AWBM model is routed by the
URBS model to calculate the flood hydrograph. Nandakumar et al. (1994) illustrated
the AWBM model as follows:
Modelling losses in flood estimation
2-18
Surface runoff =(1-BFI)*Excess
Baseflow recharge =BFI*Excess
Baseflow = (1-K)*BS
S1S2
S3
Excess
Rainfall Evaporation
BS
Figure 2.4 Schematic diagram of AWBM model (after Boughton 1993)
2.4.5 Comments
The AWBM rainfall runoff model and other runoff routing models discussed here use
a representative (single not probability distributed) value of losses in design flood
estimation. Most of these models use ARR (I. E. Aust., 1998) recommended loss
values, which has a wide range and it is generally difficult to select a single
representative value that will preserve the AEP of input rainfall depth in the final
flood output. In this research the loss values were estimated using actual rainfall and
streamflow data. No runoff routing model was used, as the surface runoff hydrograph
was derived from the streamflow data.
2.5 Loss estimate for real-time flood forecasting
The procedures followed in real-time forecasting are similar to the procedures used
in the design storm method of flood estimation. In real-time forecasting, the actual
(or forecasted) rainfall hyetograph over the catchment is combined with a model of
losses to determine the rainfall excess for input to a runoff routing model. The
Modelling losses in flood estimation
2-19
estimation of hydrographs in real-time may include a feedback component for the
correction of forecasts according to the discrepancies observed in earlier forecasts.
There are different types of models that can be used in real-time flood forecasting,
such as linear runoff routing models, non-linear runoff routing models and
conceptual models. Conceptual models make use of explicit soil moisture accounting
through allocation of water into various stores. However their use is limited in
practice due to complexity and computational effort, along with the problem of
estimating values for the large number of parameters involved.
There are some simple techniques which can be used for updating in real-time, such
as adjustment of volumetric runoff coefficient (Simpson et al., 1980). Adjustment of
mean areal precipitation (Peck et al., 1980) has also been proposed as an updating
techniques for hydrological forecasts. Some models adopt techniques to optimise
model parameters in successive forecasting time steps. Recursive techniques, such as
application of the Kalman filter to non-linear conceptual catchment models for
parameter estimation and model forecast updating, have gained prominence in the
last decade (Nandakumar, et. al., 1994).
One of the most common indices of soil moisture condition of a catchment is the
Antecedent Precipitation Index (API). Antecedent Mean Daily Flow shows a good
index of initial soil moisture conditions in humid areas, where streamflow is
continuous and baseflow discharge is estimated at the beginning of the storm. Loy et.
al. (1996) investigated regional loss model relationships for catchments in south-east
Queensland. Three types of loss model relationships were examined for estimating
initial loss. Antecedent Precipitation Index (API) and Antecedent Mean Daily Flow
(AMDF) were found to be best in overall performance for initial loss estimates.
2.5.1 Loss models adopted in real-time flood forecasting
The most commonly adopted loss models in real-time flood forecasting are the initial
loss-continuing loss, initial loss-proportional loss rate models, constant loss rate,
antecedent precipitation index and point infiltration models (Nandakumar, et. al.,
1994).
Modelling losses in flood estimation
2-20
2.5.2 Initial loss-continuing loss and initial loss-proportional loss models
Spatially lumped loss models for initial loss, combined either with continuing loss or
proportional loss rate, have achieved wide adoption in real-time flood forecasting,
mainly due to the simplicity and ease of application. Rainfall runoff models such as
RORB (Crapper, 1989; Avery, 1989), URBS (Malone, 1999), AWBM (Srikanthan et
al., 1997) which incorporate the initial loss-continuing loss and initial loss-
proportional loss models are used to forecast runoff hydrographs in real time.
2.5.3 Constant loss rate
Chander and Shanker (1984) adopted a simple unitgraph procedure for real-time
flood forecasting with on-line estimation of the constant loss rate and rainfall excess.
The first estimate of constant loss rate is calculated to match the computed flow at
the end of the first time step with the observed flow. In each subsequent step, the
constant loss rate is updated to match calculated and observed hydrographs with the
current best estimate of constant loss rate at any time step being used in determining
rainfall excess for forecasting the next step. Corradini et al. (1986) developed a semi
distributed adaptive model for real-time flood forecasting, based on Clark’s (1945)
procedure, using calibrated parameters for the catchment.
2.5.4 Antecedent precipitation index (API)
In Australia, the Bureau of Meteorology uses the API concept, in which a correlated
relationship of initial loss against an antecedent precipitation index is used in flood
forecasting procedures (I. E. Aust., 1998). The loss models incorporated in the
Melbourne’s flood warning system include fitted regression relationship between
API and mean catchment loss rate for various rainfall durations (Gieseman, 1986). In
applications, the above loss models are applied to routing models such as unit
hydrographs, or non-linear runoff routing models to derive the forecasted
hydrographs.
Modelling losses in flood estimation
2-21
2.5.5 Point infiltration equations
The Clark routing procedure loss model has several applications in Australia in real-
time flood forecasting (Corradini and Melone, 1986). The effective rainfall was
based on a time varying point rainfall infiltration model, representing pre-ponding
and post-ponding stages through the model parameters of rainfall intensity, saturated
hydraulic conductivity (ks), and sorptivity parameter (S). For infiltration rates less
than soil infiltration capacity, a temporal redistribution of rainfall was used.
Variability of infiltration in space was incorporated through spatial variability in the
rainfall pattern and using spatially “equivalent” values of Ks and S. On-line
correction of flow forecast was then carried out by updating the S parameter and a
runoff scaling factor.
A number of more simpler soil moisture accounting conceptual models, with smaller
numbers of model parameters, have been proposed. In these models, infiltration is
modelled according to the Horton equation. Forecast updating, using an auto-
regressive moving average process (ARMA model) is a fairly common technique with
this type of model. Examples of these types of conceptual models (Marivoet and
Vandewiele, 1980; Tucci and Clarke, 1980).
2.5.6 Comments
Loss models adopted for real-time flood forecasting have usually been simple
lumped models, such as initial loss combined with continuing or proportional loss
rates, constant loss rate and API relationships. In most cases, temporal variation in
losses is not considered, but alternatively some other system such as changing the
values of the parameters is used to match the observed hydrograph. The more
“advanced” conceptual models, which have gained prominence in flood forecasting,
include loss models which are capable of modelling temporal variations. However
the difficulties in parameter optimisation and in updating have continued to limit
their use.
Modelling losses in flood estimation
2-22
2.6 Currently recommended design loss values in Australia
2.6.1 Past studies on losses
Design Event Approach, Joint probability Approach and Continuing Simulations are
the common flood estimation methods. These methods of flood estimation are
discussed in Chapter 3. A selected list of major studies in relation to storm losses
such as theoretical losses, conceptual losses and losses used in flood estimation
methods are summarised in Table 2.1 under following topics:
(i) Theoretical losses
(ii) Conceptual losses
(iv) Losses used in Design Event Approach
(v) Losses used in Joint Probability Approach and
(vi) Losses used in Continuous Simulation.
Modelling losses in flood estimation
2-23
Table 2.1 Summary of previous studies on losses
Types of losses/Modelling Nature of losses Researchers
Theoretical losses These mainly account for
infiltration of water into the
ground at a point.
Horton(1940), Green and Ampt (1911), Philip (1957), Richard (1931),
Bauer(1974), Akan (1992), Walsh and Pilgrim (1993), Mein and
Larson (1971), Chu (1978), Van Mullem (1991), Aston and Dunin.
(1979), Chapman (1968), Sharma and Seely (1979), Maller and
Sharma (1980), Kool et al.(1985), Campbell (1985), Ross (1990a,b).
Conceptual losses These are probabilistic in nature
and have the ability to
approximate catchment runoff
behaviours.
Flavell and Belstead (1986), Laurenson and Pilgrim (1963), I. E. Aust.
(1987), Walsh et al. (1991), Cordery and Pilgrim (1983), Harvey
(1982), Dyer et al. (1994), Linsley et al. (1982), Cordery (1970a),
NERC, (1975), Kumar and Jain (1982), U.S. Soil Conservation
Service, (1985).Clark (1980), Moore(1985).
Design Event Approach
The design Event Approach
considers the probabilistic nature
of rainfall depth but ignores the
probabilistic behaviour of other
inputs/parameters including
losses. This method uses a
representative value (either mean
or median) from a large number
of observed values.
ARR (I.E. Aust., 1958, 1977, 1987,1998), Beran (1973), Ahern and
Weinmann (1982), Walsh et al. (1991), Hill and Mein (1996), Overney
et al. (1995) Russel et al. (1979), Viessman et al. (1989).
Modelling losses in flood estimation
2-24
Table 2.1 Summary of previous studies on losses continued
Joint Probability
Approach
Considers outcomes of events of all
possible combination of input values and,
if necessary, their correlation structure, to
estimate design floods. This method uses
probabilistic losses in flood estimation.
ARR (I.E. Aust., 1987), Ahern and Weinmann.(1982), Beran and
Sutcliffe (1972), Eagleson (1972), Hughes (1977), Goyen (1983),
Laurenson (1974), Rahman et al. (2000, 2001a, b). Haan (1977), Haan
and Schulze (1987), Weinmann et al. (1998).
Continuous
Simulation
These Models aim to represent the major
processes responsible for converting the
catchment rainfall inputs into flood
outputs. This method estimates losses
from a continuous water balance.
Bras et al.(1985), Bradley and potter (1992), Porter and McMahon
(1971), Crawford and Linsley (1966), Chapman (1968), Boughton and
Carroll (1993), Linsley and Crawford (1974), James and Robinson
(1986), Hubber et al. (1986), Linsley et al. (1988), Rahman et al.
(1998), Russell (1977), Lumb and James (1976), Ahern and
Weinmann (1982), Boughton,W.C (2001), James and Robinson
(1982).
Modelling losses in flood estimation
2-25
2.6.2 Losses recommended in ARR (I. E. Aust., 1998)
The recommended design values in ARR (I. E. Aust., 1998) for Queensland, NSW
and Victoria are shown in Table 2.2.
There is virtually no information on the initial loss–proportional loss model and the
recommended value of initial loss varies over a large range. It is important to note
that for a large portion of Victoria (North and West of Great Dividing Range), the
only recommendation for design loss is that it is “probably as for similar areas of
NSW”. It is found in many studies that the use of design losses with the design
rainfall recommendation in ARR (I. E. Aust., 1987) result in the overestimation of
design peak flows, when compared with a frequency analysis of recorded peak flows
(Hill and Mein, 1996; and Walsh et al., 1991). This indicates that, for many
catchments, the design losses recommended in ARR (I. E. Aust., 1987; I. E. Aust.,
1998) are too low.
There are two inadequacies in the current ARR (I. E. Aust., 1998) loss values, which
were derived from the analysis of large flood events. These are, the selection of high
runoff events for loss derivation is biased towards wet antecedent conditions (i.e.
losses tend to be too low) and storm losses do not account for the nature of design
rainfall information of ARR (I. E. Aust., 1998), which has been derived from rainfall
bursts within longer duration storms (i.e. losses tend to be too high). It is recognised
in ARR (I. E. Aust., 1998) that these two adequacies should have opposite effects.
However it is implicitly assumed by users of the current design loss values that they
fully compensate each other.
Waugh (1991) examined the effect of the first inadequacy. From a study of five
Western Australian catchments, he concluded that the selection of runoff events for
the estimation of design loss underestimates the design loss value, and can result in
over-estimation of the design flood magnitude by up to 20%. This is because many
sizable summer storms yielded little or no runoff, due to dry antecedent conditions,
and were not included in the analysis. Srikanthan and Kennedy (1991) examined the
effect of the second inadequacy. They examined the degree to which the rainfall
burst used to generate design temporal patterns was embedded within longer duration
Modelling losses in flood estimation
2-26
storms, and found that, for a given annual exceedance probability (AEP), antecedent
rainfall prior to storm bursts decreased with increasing storm duration. This is
because, as the duration of the bursts increases, more bursts represent complete
storms.
Table 2.2 Currently recommended Design Losses in ARR (I. E. Aust., 1998)
Modelling losses in flood estimation
2-27
The critical duration of design rainfall for a catchment is that duration which gives
the largest design peak flow. It depends on catchment characteristics and on the
variation of the rainfall magnitude and temporal pattern with duration. It would be
reasonable to expect that the critical duration would increase with catchment size, as
the surface runoff will take longer duration to reach the outlet of the catchment but a
number of studies have found the critical duration does not depend on catchment
area. For example, in a study of 22 NSW catchments (ranging in areas from 25 to
6,560 km 2 ), Walsh et al. (1991) found that the critical duration was independent of
catchment size. Some investigators (Walsh et al., 1991; Hill and Mein, 1996) have
noted that, when the recommended design losses in ARR (I. E. Aust., 1987) are
applied to the design rainfall, the resulting critical duration is excessively long.
Hill and Mein (1996) mentioned the excessively long critical durations, when
applying the ARR (I. E. Aust., 1987) recommended loss values to 8 Victorian
catchments with catchment areas ranging from 32 to 153 km2. For all but one
catchment in temporal pattern Zone 1, the 36 hour temporal pattern was critical. In
addition, the 72 hour temporal pattern was found to be critical for one catchment in
Zone 2, which has a catchment area of only 60 km2. Hill and Mein (1996) also noted
that the excessively long critical durations are caused by:
(i) Temporal patterns which include sub-intervals with excessively heavy
rainfall, and which are not consistent with the estimated average recurrence
interval (ARI) of the complete pattern. This sub-interval with excessively
heavy rainfall can be alleviated by fully filtering the temporal patterns to
ensure that no sub-interval has an ARI which is greater than the ARI of the
complete storm.
(ii) Another reason which causes the excessively long critical duration is the use
of the same design value of initial loss for all durations. In order to be
consistent with the derivation of design rainfalls in ARR (I. E. Aust., 1987),
the design initial loss should increase with increasing duration.
Modelling losses in flood estimation
2-28
(iii) The time increment used in ARR (I. E. Aust., 1987) for the longer duration
temporal patterns is too long, and for the smaller catchments, it approaches
the response time of the catchment.
2.7 Summary
The point infiltration equations discussed in Section 2.2 of this chapter are suitable
for small catchments and each theoretical equation has its own limitations. It can be
stated that, theoretically based point infiltration models should be considered to
determine the infiltration capacity of different types of soil and soil topography.
Larger catchments tend to be heterogeneous and as a result do not have the same
infiltration capacity throughout the catchments. Hence, these approaches are not
particularly appropriate for design flood estimation for larger catchments. The IL-CL
loss model has been widely used in Australia (I. E. Aust., 1998), due both to its
simplicity and to its generally better approximation of the temporal pattern of storm
losses than the constant loss rate approach.
The AWBM rainfall runoff model and other runoff routing models discussed here use
a representative value of losses in design flood estimation. The design loss values
which were used in these models were usually from ARR (I.E. Aust., 1998)
recommended loss values, which have wide ranges and it is generally difficult to
select a value that will preserve the AEP of input rainfall depth in the final flood
output.
Loss models adopted for real-time flood forecasting are usually simple lumped loss
models. In these models the temporal variation of losses are not considered. There
are some advanced conceptual loss models for flood forecasting, which are capable
of modeling temporal variation but there is limitation in their use because of the
difficulties in parameter optimization and in updating.
The existing two inadequacies in ARR (I. E. Aust., 1998) loss values were derived
from analysis of large flood events. These are, the selection of high runoff events for
loss derivation is biased towards wet antecedent conditions (i.e. losses tend to be too
Modelling losses in flood estimation
2-29
low) and storm losses do not account for the nature of design rainfall information of
ARR (I. E. Aust., 1998), which has been derived from rainfall bursts within longer
duration storms (i.e. losses tend to be too high).
These inadequacies result in over estimation of design floods. Ultimately the
overestimation of design floods lead to the construction of over design and expensive
hydraulic structures. It is therefore important to estimate suitable design losses which
would ultimately save millions of dollars in construction of hydraulic structures.
Hence it is necessary to develop a new loss function to estimate floods. A suitable
flood estimation method among the existing methods should be selected to examine
the effect of losses in flood estimation. A detailed discussion is performed about the
selection of flood estimation method in the next chapter.
Modelling losses in flood estimation 3-1
CHAPTER 3
FLOOD ESTIMATING METHODS
3.1 General
Design flood estimation methods can be broadly classified into two groups: rainfall
based methods and streamflow based methods. Rainfall based methods can be
subdivided into event based methods and continuous simulation. Design Event
Approach is an example of event based methods. The currently recommended
method for design flood estimation in ARR (I. E. Aust., 1998) is based on Design
Event Approach in that a hypothetical design storm is defined as an input to a runoff
routing model. In recent years, significant developments have been made in the
application of more advanced techniques in flood estimation e.g. Joint Probability
Approach (Rahman et al., 2001a, b) and Continuous Simulation (Boughton, 2001).
This chapter describes about all three types of flood estimation methods.
3.2 Design Event Approach
ARR (I. E. 1998), Ahern and Weinmann (1982), Beran (1973) described the steps
involved with the Design Event Approach. The estimation of design flood of a
specified annual exceedance probability (AEP) by Design Event Approach can be
described in the following steps:
i. Select a number of design storm durations D1, D2, …and for each of these,
obtain a streamflow hydrograph following the steps given below.
ii. Obtain an average rainfall depth from the intensity-frequency-duration (IFD)
curve, given the design location, specified AEP and duration.
iii. Distribute the total rainfall within the duration to form the gross rainfall
hyetograph.
iv. Select loss parameters and compute rainfall excess hyetograph.
v. Formulate catchment response model.
Modelling losses in flood estimation 3-2
vi. Select catchment response parameters.
vii. Select design baseflow.
viii. Compute stream flow hydrograph.
ix. The rainfall duration giving maximum peak flood is taken as critical duration,
and the corresponding peak is taken as the design flood of the specified AEP.
The key assumption involved in this approach is that the representative design values
of the inputs/ parameters (such as duration, temporal pattern, loss shown in Figure
3.1) at different steps can be defined in such a way that they are ”AEP neutral” i.e.
they result in a flood output that has the same AEP as the rainfall input. The success
of this approach is crucially dependent on how well this assumption is satisfied.
There are no definite guidelines on how to select the appropriate values of the
inputs/parameters in the above steps that are likely to convert a rainfall depth of
particular AEP to the design flood of the same AEP. There are many methods to
determine an input value. Likewise, other inputs to the design such as critical storm
duration, spatial and temporal distributions of the design storm, and baseflow values
can also be determined by many methods, the choice of which is totally dependant on
various assumptions and preferences of the individual designer. As an example, a
designer may follow any of the many possible combinations of the inputs (as shown
in the Figure 3.1). The critical storm duration is an important factor in converting
rainfall input to a flood output of design AEP. Critical storm duration depends on
storm factors, loss factors, catchment characteristics. Selection of inappropriate
values for any of these factors will result in different critical durations and assuming
that the resulting flood has the same AEP as the input rainfall depth is not correct.
Due to non-linearity of the transformation in the rainfall-runoff process, it is
generally not possible to know a priori how a representative value for an input should
be selected to preserve the AEP (Rahman et al., 1998).
In summary, the current Design Event Approach considers the probabilistic nature of
rainfall depth but ignores the probabilistic behaviour of other inputs/parameters such
as rainfall duration, losses and temporal pattern. The assumption regarding the
probability of the flood output i.e. a particular AEP rainfall depth will produce a
flood of the same AEP is unreasonable in many cases. The arbitrary treatment of the
various flood producing variables, as done in the current Design Event Approach, is
Modelling losses in flood estimation 3-3
likely to lead to inconsistencies and significant bias in flood estimates for a given
AEP.
Figure 3.1 Schematic diagram of the Design Event Approach
Duration
Temporal pattern
Spatial pattern
Loss
Catchment response model and parameters
Baseflow Design flood AEP = ?
AEP = 1 in Y
D2 D4D3
T1 T2 T3
S1 S2 S3
L1 L2 L3
P1
B1 B2 B3
D1 D5 Total rainfall depth
T4 T5
S4 S5
L4 L5 Rainfall excess hyetograph
P2 P3
Surface hydrograph
F
Modelling losses in flood estimation 3-4
3.3 Joint Probability Approach
The Joint Probability Approach recognises that any design flood characteristics (e.g.
peak flow) could result from a variety of combinations of flood producing factors,
rather than from a single combination, as in the Design Event Approach (Eagleson,
1972, Beran, 1973, Laurenson, 1974, Russell et al., 1979). For example, the same
peak flood could result from a small storm on a saturated basin or a large storm on a
dry basin. Thus it is clear that the Joint Probability Approach, which considers the
outcomes of events with all possible combinations of input, values and if necessary
their correlation structure is most likely to lead to better estimates of design flows
(Rahman et al., 1998, Weinmann et al., 1998). The flood output, consequently, will
also have a probability distribution instead of a single value.
By using the same component models as the current Design Event Approach but
treating inputs and parameter values to the design as random variables, the Joint
Probability Approach obviously attempts to eliminate subjective criteria in
specifying input values.
Joint Probability Approach is theoretically superior to the Design Event Approach
and regarded as an attractive design method (I.E. Aust., 1998). The Joint Probability
Approach is based on the Theorem of Total Probability that can be expressed in three
dimensions by the following equation:
)(),|()( ,111
xkixki
t
x
m
k
n
iDCBPDCBAPAP ∩∩= ∑∑∑
===
(3.1)
In applying the Theorem of Total Probability to the calculation of flood probability,
the explanations of the terms involved in Equation (3.1) are as follows:
• P (A) is the unconditional probability of a flood (to be exceeded in any given
year);
• B, C, D are random variables to the design, for example temporal pattern, losses
and storm duration.
Modelling losses in flood estimation 3-5
The method of combining probability distributed inputs to form a probability
distributed output is known as the derived distribution approach. A derived
probability distribution can be solved in two ways:
1) Analytical methods
2) Approximate methods
It depends upon the analytical skills of the designer and the computer resources
available to select a method to compute derived probability distribution. In the
analytical methods the probability distribution of the dependent variable is found by
directly applying principles of probability. The cumulative density function of the
dependent variable should be determined. The analytical methods have been found to
be difficult to apply under practical situations mainly due to the mathematical
complexity involved and difficulties in parameter estimation.
In approximate methods a solution is obtained by substituting numerical values for
variables and parameters. For example, Rahman et al. (2001a, b) developed a Monte
Carlo simulation technique (a type of approximate method), as illustrated in Figure
3.2, to solve Equation (3.1). This technique is easy to apply and has been found to
provide more precise flood estimates than the Design Event Approach. The
approximate methods can be applied relatively easily. These methods are applied
when an analytical approach to the problem becomes difficult or impossible.
In recent applications of Joint Probability Approach, Rahman et al. (2001a, b) have
used initial loss-continuing loss model without investigating the appropriateness of
other loss functions. Also, they assumed continuing loss to be a constant throughout
the storm duration. This could be a tenuous assumption for long duration storms.
Also they ignored the random variability of continuing loss from storm to storm.
The steps involved in the Monte Carlo Simulation technique illustrated in Figure 3.2
are as follows:
• Draw a random value of duration from the rainfall duration (D).
Modelling losses in flood estimation 3-6
• Given the random value of duration, draw a random value of rainfall intensity
from the conditional distribution of rainfall intensity (I).
• Given the random value of duration draw a random temporal pattern from the
conditional distribution of temporal pattern (TP).
• Given the random value of duration draw a random value of initial loss from
the conditional distribution of initial loss (IL).
• Run the randomly selected variables through the runoff generation and runoff
routing models to simulate a flood hydrograph.
• Add the baseflow to the simulated flood hydrograph and note the flood peak,
repeat the above steps in the order of thousands.
• Use the simulated flood peaks to determine the derived flood frequency curve
in a distribution free manner using rank-order statistics.
Modelling losses in flood estimation 3-7
Rainfall events
Identify component distributions
Rainfall duration(D )
Rainfall intensity (I )
Rainfall temporal pattern (TP )
Losses(IL )
D I TP IL
Rainfall excesshyetograph
Route through runoff routing model
Runoff model calibration (m, k )
Baseflow analysis
Peak of simulatedstreamflow hydrograph
Construct derived flood frequency curve
Repeat in theorder of thousands
Figure 3.2 Schematic diagram of the Joint Probability Approach (Monte Carlo
simulation technique, probability-distributed components shaded)
Modelling losses in flood estimation 3-8
3.4 Continuous Simulation Continuous Simulation is the alternative to the Design Event Approach using
deterministic catchment models or rainfall-runoff process models (I.E. Aust., 1998).
Examples of this approach can be found in Crawford and Linsley (1966); Linsley and
Crawford (1974); Huber et al. (1986) and Linsley et al. (1988).
Continuous Simulation models aim to represent the major processes responsible for
converting the catchment rainfall inputs into flood outputs. They generate outflow
hydrographs over a long period of time from inputs of historical rainfall series,
potential evaporation and other climatological data. Time steps used in these models
are usually from one hour to one day, sometimes may be as short as 5 or 15 minutes,
and the simulation period is often up to hundreds of years. It is difficult to use the
continuous modelling for design flood estimation in rural catchments because of the
complexity involved in calibration of the model.
ARR (I.E. Aust., 1998) states that: “although Continuous Simulation models provide
probably the best means of estimating runoff sequences from rainfall, considerable
care and expertise are required to obtain useful results”.
The main problems with the Continuous simulation arise from the difficulties in
adequately modeling the soil moisture balance, synthesising long records of rainfall
and evaporation at the appropriate temporal and spatial resolution, and accounting
for correlations between inputs. There are also some other difficulties associated with
Continuous Simulation as described by Lumb and James (1976), Ahern and
Weinmann (1982), James and Robinson (1982), Institution of Engineers (1998):
• Loss of sharp events by the use of relatively long time steps.
• Time and effort are required in gathering the precipitation and other
climatologic data needed for simulation of a long continuous sequence
(Extensive data requirements).
• Management of large amount of time series output (data management).
Modelling losses in flood estimation 3-9
• Expertise is required to determine parameter values which best reproduce
historical hydrographs (model calibration effort).
3.5 Selection of flood estimation method
The main problems with the Continuous Simulation arise from the difficulties in
adequately modeling the soil moisture balance, synthesising long records of rainfall
and evaporation at the appropriate temporal and spatial resolution, and accounting
for correlations between inputs. Hence Design Event Approach and Joint Probability
Approach were used in this research for flood estimation and flood frequency
analysis.
3.6 Summary
The currently recommended method for design flood estimation in ARR (I. E. Aust.,
1998) is based on Design Event Approach in that a hypothetical design storm is
defined as an input to a runoff routing model. The key assumption involved in this
approach is that the representative design values of the inputs/ parameters (such as
duration, temporal pattern, loss shown in Figure 3.1) at different steps can be defined
in such a way that they are ”AEP neutral” i.e. they result in a flood output that has
the same AEP as the rainfall input. The success of this approach is crucially
dependent on how well this assumption is satisfied.
In recent years, significant developments have been made in the application of more
advanced techniques in flood estimation e.g. Joint Probability Approach (Rahman et
al., 2001a, b) and Continuous Simulation (Boughton, 2001). Thus it is clear that the
Joint Probability Approach, which considers the outcomes of events with all possible
combinations of input, values and if necessary their correlation structure is most
likely to lead to better estimates of design flows. Joint probability Approach is
theoretically superior to the Design Event Approach and regarded as an attractive
design method (I. E. Aust., 1998).
Modelling losses in flood estimation 3-10
Although the continuous modelling approach is conceptually sounder than event
based approaches, it is difficult to use the continuous modelling for design flood
estimation in rural catchments because of the complexity involved in calibration of
the model. The main problems with the Continuous Simulation arise from the
difficulties in adequately modeling the soil moisture balance, synthesising long
records of rainfall and evaporatioin at the appropriate temporal and spatial resolution,
and accounting for correlations between inputs. Hence Design Event Approach and
Joint Probability Approach were used in this research as the flood estimation
methods.
To estimate the design flood of a catchment using the Design Event Approach and
Joint Probability Approach some catchments are needed to be selected, which are
discussed in the next Chapter.
Modelling losses in flood estimation 4-1
CHAPTER 4
STUDY CATCHMENTS 4.1 General For the proposed study, selection of study catchments was an important step. This
chapter describes the process of how study catchments were selected. Firstly, the
criteria to select the study catchments are discussed. Then mapping catchment
boundaries and all selected stations are described. Then the discussions on selected
catchments are presented.
4.2 Criteria to select study catchments
This study was aimed at deriving new improved design losses for Queensland
catchments. Initially from the whole of Queensland gauging stations, 78 catchments
were selected from a book (DNRM, 2000) published by the Department of Natural
Resources and Mines. This initial selection was performed depending upon the
catchment area, regulation and record length of the catchment. But there were some
other important criteria which were to be considered to finalise the catchment
selection. These were the location of the pluviograph station, daily rainfall station,
streamflow station and the catchment boundary. All of these criteria are discussed in
this Chapter.
4.2.1 Catchment area
It was important to know whether the catchment is small or large in size. The loss
(IL-CL) model which was used in this research is only suitable for small to medium
size catchments, and not suitable to compute the loss values for the larger
catchments. The reason is that the process of computing loss values for larger
catchments is different from the process of computing loss values for smaller
catchments as described in section 6.7. The definition of initial loss (IL) computation
adopted in this research section 6.3 does not satisfy large catchments, hence the
proposed study was based on small to medium sized catchments. Australian Rainfall
Modelling losses in flood estimation 4-2
and Runoff (ARR) (I. E. Aust, 1998) suggests the catchment area with an upper limit
of 1000 km2 can be considered as a small to medium sized catchments, which was
taken as a guide to selecting the study catchments.
4.2.2 Regulation
To select the study catchments consideration was given to whether the study
catchments were regulated or unregulated. As major regulation affects the rainfall
runoff relationship significantly, gauging stations subject to major regulation (such as
dams, back water effect etc) were not included in this research. Also urbanisation
affects the catchment hydrology, so no urban catchment was selected. Only
unregulated rural catchments were selected for this research work.
4.2.3 Record length
The record length of the rainfall and streamflow data should be long so that more the
number of rainfall and streamflow events can be selected for the analysis of any
catchment. The highest record length of streamflow data is 48 years and the mean
and median value of streamflow record length are 30 and 31 years. Some streamflow
data with record lengths up to 11 years were also selected. Hence the recorded lower
limit of streamflow data was 11 years for this analysis. In this study, if the stream
gauging record length for any catchment is less than 11 years, the catchment was not
selected as a candidate catchment.
Considering the above criteria, a total of 48 catchments were selected from all over
the Queensland. Streamflow data were obtained from the Department of Natural
Resources and Mines (DNRM) Queensland. Topographic Maps of Australia
(1:100000) were consulted to investigate the nature of streamflow network and
nature of regulation in the selected catchments. Also the gauging authority was
consulted to know about any recent changes of regulation and land use in the
selected catchments. A total of 132 pluviograph stations and 338 daily rainfall
stations were selected from and near the selected catchments. The rainfall data were
obtained from the Bureau of Meteorology (BoM), Australia. The streamflow and
Modelling losses in flood estimation 4-3
rainfall data graded as ‘good quality’ by the data authorities were accepted and poor
quality data were not included in the study.
4.2.4 Catchment boundary
All the 78 catchment boundaries were collected from the Department of Natural
Resources and Mines in electronic forms. Mapinfo professional version 5.0 (mapping
software) was used to select the catchment boundary for the study catchment.
After mapping the catchment boundary, an electronic layer of stream gauging
stations were laid over the catchment boundary. The catchments whose location of
the stream gauging station in the map was found away from the catchment boundary,
that catchment was not selected as study catchment. Catchments were selected, when
there was one or more pluviograph station or daily rainfall stations within the
catchment boundary. To select a rainfall streamflow event to estimate loss values the
temporal pattern of the rainfall over the catchment is to be obtained. Catchments with
only one pluviograph station but no daily rainfall station within the catchment
boundary were selected as candidate catchments. As the catchments were small to
medium in size, it was assumed that the temporal pattern of the pluviograph data
were the representative temporal pattern of the whole catchment, provided the
pluviograph station was located well inside the catchment boundary. But catchments
with no pluviograph station inside or within 50 km of the catchment boundary were
not selected as study catchments, though there was daily rainfall station within or
near the catchment boundary. Again catchments having a pluviograph station close
to the boundary and with daily rainfall stations within the catchment boundary, were
selected as study catchments. Because, it was assumed that when the pluviograph
station and daily rainfall station are closely located, the temporal pattern of the daily
rainfall station and the pluviograph station were same. Hence the pluviograph data
can be used to proportion the daily rainfall data to obtain the representative temporal
pattern of rainfall within the catchment.
Modelling losses in flood estimation 4-4
4.3 Mapping catchment boundaries and the selected pluviograph,
daily rainfall and streamflow stations
Finally a total of 48 study catchments were selected out of 78 catchments. A
mapping computer software known as Mapinfo professional version 5.0 (Mapinfo,
1998) was used in this study to select the representative pluviograph stations and the
daily rainfall stations. In Mapinfo for each of catchments, the selected catchment
boundaries, pluviograph stations and the daily rainfall stations were laid in electronic
layers one above another. Pluviograph stations and the daily rainfall stations which
were representative of the respective catchments were selected from this electronic
diagram. A single catchment boundary and its stream gauging station with all its
pluviograph and daily rainfall stations are shown in Figure 4.1 as an example. The
Figure 4.1 also shows that all the three pluviograph stations are inside the catchment
boundary and a number of daily rainfall stations are also located inside the catchment
boundary.
Modelling losses in flood estimation 4-5
Figure 4.1 Example of the Major Creek catchment with a stream gauging station
119006A, pluviograph and daily rainfall stations
The symbols used in Figure 4.1 are illustrated below:
● Stream gauging station (119006A)
■ Pluviograph station
□ Daily rainfall station
33,166
33,171
33,173
119006A
33,074 33,151
33,159 33,171
33,173
33,226
33,166
33,171
33,173
119006A
33,074 33,151
33,159 33,171
33,173
33,226
33,166
33,171
33,173
119006A
33,074 33,151
33,159 33,171
33,173
33,226
33,166
33,171
33,173
119006A
33,074 33,151
33,159 33,171
33,173
33,226
Modelling losses in flood estimation 4-6
Each catchment had only one stream gauging station. Within the catchment boundary
there could be more than one pluviograph and daily rainfall stations. The gauging
station which has shown in Figure 4.1 is 119006A.
119 - Basin number
0 - Sub-basin number
06 - Station number
A - Site location code
4.4 Selected catchments
Of the selected 48 study catchments which were selected from the whole of
Queensland, most of these catchments were selected from the coastal areas because
of their long streamflow record. There were only a few catchments in western
Queensland having a reasonably long streamflow record. The selected catchments
were mainly unregulated and rural and have reasonably long rainfall and streamflow
records. In Queensland there are 5 drainage divisions and there are many basins in
each drainage division. The distribution of the drainage divisions and basins are
shown in Figure 4.2.
Modelling losses in flood estimation 4-7
Figure 4.2 Distribution of the drainage divisions and the basins in Queensland
Modelling losses in flood estimation 4-8
The distribution of the candidate catchments selected from all over the Queensland is shown
in Figure 4.3. The locations of the study catchments were identified by the electronic layer of
catchment boundaries with in the Queensland boundary using Mapinfo professional version
5.0.
Figure 4.3 Locations of the 48 study catchments in Queensland
Each study catchment is represented by a stream gauging station. A list of selected stream
gauging stations numbers, streamflow names, location of stream gauging stations, latitude and
longitude of stream gauging stations, catchment area and streamflow record length (start and
finish date) is shown in Table 4.1.
Modelling losses in flood estimation 4-9
Table 4.1 Stream gauge station number, their location , catchment area and streamflow record length of the study catchments Serial
Number Basin ID Streamflow name Location of stream
gauging station Lat. of stream
gauge Long. of
stream gaugeCatchment area (km2)
Start date of streamflow
Finish date of streamflow
1 102101A Pascoe River Fall Creek 12.87 142.97 635 1/10/1967 Continue 2 104001A Stewart River Telegraph Road 14.17 143.38 480 18/01/1970 " 3 105105A E. Normanby River Development Road 15.77 145.00 300 24/02/1969 " 4 107001B Endeavour River Flaggy 15.42 145.05 310 1/10/1967 " 5 107003A Anna River Beesbike 15.68 145.20 247 9/03/1990 " 6 112003A N. Johnston River Glen Allyn 17.37 145.65 173 1/10/1958 " 7 112101B S. Johnston River Upstream Central Meal 17.60 145.97 400 1/10/1974 " 8 114001A Murray River Upper Murray 18.10 145.80 155 26/05/1970 " 9 116008B Gowrie Creek Abergowrie 18.43 145.83 124 1/10/1953 "
10 116015A Blunder Creek Wooroora 17.73 145.43 127 20/10/1966 " 11 116017A Stone River Running Creek 18.77 145.95 157 30/06/1970 " 12 118003A Bohle River Hervey Range Road 19.32 146.70 143 1/04/1985 " 13 119006A Major Creek Damsite 19.67 147.02 468 4/05/1978 " 14 120014A Broughton River Oak Meadows 20.17 146.32 182 5/11/1970 13/04/1999 15 120216A Broken River Old Racecourse 21.18 148.43 78 1/06/1969 " 16 124002A St. Helens Creek Calen 20.90 148.75 129 7/02/1973 " 17 125005A Blacks Creek Whitefords 21.32 148.82 505 12/12/1973 " 18 130207A Sande Creek Clermont 22.78 147.57 409 21/01/1965 " 19 136108A Monal Creek Upper Monal 24.60 151.10 92 15/07/1962 " 20 137101A Gregory River Burrum Highway 25.08 152.23 454 10/02/1966 " 21 138110A Mary River Bellbird Creek 26.62 152.70 486 1/10/1959 " 22 141009A N. Maroochy River Eumundi 26.48 152.95 38 15/02/1982 " 23 143110A Bremer River Adams Bridge 27.82 152.50 125 30/09/1968 " 24 143212A Tenhill Creek Tenhill 27.55 152.38 447 18/03/1968 " 25 145003B Logan River Forest Home 28.20 152.77 175 1/10/1953 "
Modelling losses in flood estimation 4-10
Serial Number Basin ID Streamflow name
Location of stream gauging station
Lat. of stream gauge
Long. of stream gauge
Catchment area (km2)
Start date of streamflow
Finish date of streamflow
26 145010A Running Creek 5.8km Deickmans
Bridge 28.23 152.88 128 26/11/1965 " 27 145011A Teviot Brook Croftby 28.13 152.57 83 7/02/1966 " 28 146014A Back Creek Beechmont 28.12 153.18 7 5/06/1971 " 29 145101D Albert River Lumeah Number 2 28.05 153.03 169 1/10/1953 " 30 416410A Macintyre Brook Barongarook 28.43 151.45 465 15/06/1967 " 31 422321B Spring Creek Killarney 28.35 152.32 35 1/10/1972 " 32 422338A Canal Creek Leyburn 28.02 151.58 395 27/03/1972 " 33 422394A Cadamine River Elbow Vally 28.37 152.13 325 2/12/1972 " 34 913005A Paroo Creek Damsite 20.33 139.52 305 20/11/1968 1/10/1988 35 913009A Gorge Creek Flinders Highway 20.68 139.63 248 13/11/1970 Continue 36 915205A Malbon River Black Gorge 21.05 140.06 425 1/10/1970 1/10/1988 37 916002A Norman River Strathpark 19.53 143.25 285 1/10/1969 30/09/1988 38 916003A Moonlight Creek Alehvale 18.27 142.33 127 1/10/1969 10/04/1989 39 917005A Agate Creek Cave Creek Junction 18.93 143.47 228 1/07/1969 30/09/1988 40 917007A Percy River Ortana 19.15 143.48 445 2/09/1969 30/09/1988 41 917107A Elizabeth Creek Mount Surprise 18.13 144.30 585 23/07/1968 Continue 42 917114A Routh Creek Beef Road 18.28 143.70 81 11/12/1972 30/09/1988 43 919201A Palmer River Goldfields 16.10 144.77 530 11/12/1967 Continue 44 919205A North Palmer River 4.8 km 16.00 144.28 430 16/10/1973 30/09/1988 45 921001A Holroyd River Ebagoola 14.23 143.15 365 19/01/1970 17/05/1988 46 922101B Coen River Racecourse 13.95 143.17 166 10/11/1967 Continue 47 926002A Dulhunty River Dougs Pad 11.83 142.42 325 18/11/1970 " 48 926003A Bertie Creek Swordgrass Swamp 11.82 142.50 130 10/11/1972 "
Table 4.1 (Continued)
Modelling losses in flood estimation 4-11
The distribution of the selected catchment area is shown in Figure 4.4. The areas of
the catchment were ranging from 7 km2 – 645km2. The median and mean of the
catchment areas were respectively 248 km2 and 273 km2. Only 8.0% of the
catchments were greater than 500 km2 and 15.0% were smaller than 100 km2.
Figure 4.4 Distribution of catchment areas of the study catchments in Queensland.
4.5 Summary
Initially 78 catchments were selected on the basis of catchment size, regulation and
streamflow record length. But after plotting all these catchment boundaries,
pluviograph, rainfall and stream gauging stations in electronic forms in Mapinfo
professional version 5.0, only 48 catchments were selected for loss computation and
eventually to estimate the design flood.
592.0528.0
464.0400.0
336.0272.0
208.0144.0
80.016.0
Freq
uenc
y
12
10
8
6
4
2
0
592.0528.0
464.0400.0
336.0272.0
208.0144.0
80.016.0
Freq
uenc
y
12
10
8
6
4
2
0
592.0528.0
464.0400.0
336.0272.0
208.0144.0
80.016.0
Freq
uenc
y
12
10
8
6
4
2
0
592.0528.0
464.0400.0
336.0272.0
208.0144.0
80.016.0
Freq
uenc
y
12
10
8
6
4
2
0
592.0528.0
464.0400.0
336.0272.0
208.0144.0
80.016.0
Freq
uenc
y
12
10
8
6
4
2
0
592.0528.0
464.0400.0
336.0272.0
208.0144.0
80.016.0
Freq
uenc
y
12
10
8
6
4
2
0
592.0528.0
464.0400.0
336.0272.0
208.0144.0
80.016.0
Freq
uenc
y
12
10
8
6
4
2
0
592.0528.0
464.0400.0
336.0272.0
208.0144.0
80.016.0
Freq
uenc
y
12
10
8
6
4
2
0
Catchment area (km2)
Modelling losses in flood estimation 4-12
To select a rainfall streamflow event the temporal pattern of the rainfall over the
catchment is required. To estimate loss values the selection of rainfall streamflow
event was required. Catchments with only one pluviograph station but no daily
rainfall station within the catchment boundary were selected as candidate
catchments. As the catchments were small to medium in size, it was assumed that the
temporal pattern of the pluviograph data was the temporal pattern of the whole
catchment, provided the pluviograph station was located well inside the catchment
boundary.
Most of these catchments were selected from coastal areas because there were only a
few catchments in western Queensland having a reasonably long streamflow record.
The selected catchments were mainly unregulated, rural and with reasonably long
rainfall and streamflow records. Once the catchments were selected it was important
to compile and analyse the data to compute losses for design flood estimation. So
data compilation and analysis is discussed in the next chapter.
Modelling losses in flood estimation 5-1
CHAPTER 5
DATA ANALYSIS 5.1 General
All the streamflow data for this research were collected from the Department of
Natural Resources and Mines (DNRM, 2000) for this research. Pluviograph data and
daily rainfall data were collected from the Bureau of Meteorology (BOM). Only
good data were used in this research, as the quality and error of the data were
checked by the respective authority before the data was collected. This chapter
describes the collation and formatting of streamflow, pluviograph and daily rainfall
data. It also describes how to estimate the catchment average rainfall and finally how
to select the rainfall streamflow events.
Generally it is found that the srteamflow hydrograph tend to be high peaked when
there is little or no delay between the commencement of rainfall excess and the start
of rise of the streamflow hydrograph. Again the hydrograph tend to be broad crested
when a considerable delay will occur between the beginning of rainfall excess and
the start of surface runoff at the catchment.
5.2 List of stream gauge, pluviograph and daily rainfall stations
The list of the selected stream gauge stations, their location, streamflow record
lengths, selected pluviograph stations and daily rainfall stations is provided in Table
5.1. The Basin ID is the unique identifier of stream gauging stations used by the
Queensland Department of Natural resources and Mines (DNRM) for Queensland
catchments. Table 5.1 shows that for each catchment only one stream gauge station
was selected but 5 pluviograph stations and 5 daily rainfall stations were selected.
Table 5.1 shows the data range of all the selected 48 stream gauging stations. The
pluviograph and daily rainfall stations were selected on the basis of their locations
and the record length of data as shown in Figure 5.1.
Modelling losses in flood estimation
5-2
Table 5.1 List of the selected stream gauge, pluviograph and daily rainfall stations
SL BASIN ID NAME LOCATION LA.. LOG.AREA(Km2) ST. DATE FIN. DATE PluSta 1 PluSta 2 PluSta 3 Plusta 4 plusta 5 DaRaSta 1 DaRaSta2 DaRaSta3 DaRaSta4 DaRaSta5
1 102101A Pascoe River Fall Creek 12.87 142.97 635 1/10/1967 Continue 27006 27045 27042 27022 _ 27024 28008 28001 27064 27014
2 104001A Stewart River Telegraph Road 14.17 143.38 480 18/01/1970 " 27006 27045 27042 31017 _ 27005 27028 28017 27006 _
3 105105A East Normanby River Development Road 15.77 145.00 300 24/02/1969 " 31016 31017 31055 28004 _ 31112 31074 31009 31106 31003
4 107001B Endeavour River Flaggy 15.42 145.05 310 1/10/1967 " 31017 31016 28004 31055 _ 31129 31107 31030 31017 31137
5 107003A Anna River Beesbike 15.68 145.20 247 9/03/1990 " 31016 31017 31055 28004 _ 31112 31074 31007 31003 31106
6 112003A North Johnston River Glen Allyn 17.37 145.65 173 1/10/1958 " 31034 31053 32061 32021 31046 31038 31153 31050 31149 31163
7 112101B South Johnston River Upstream Central Meal 17.60 145.97 400 1/10/1974 " 32061 31053 31083 32021 32076 32161 32138 31186 31128
8 114001A Murray River Upper Murray 18.10 145.80 155 26/05/1970 " 32004 32042 31083 32052 32018 32109 32035 32144 32060 32006
9 116008B Gowrie Creek Abergowrie 18.43 145.83 124 1/10/1953 " 32052 32004 32086 32043 32042 32123 32153 32147 32050 32170
10 116015A Blunder Creek Wooroora 17.73 145.43 127 20/10/1966 " 31083 31053 31046 32018 32042 31078 31085 31142 31083 31018
11 116017A Stone River Running Creek 18.77 145.95 157 30/06/1970 " 32043 32064 32078 32086 32052 32039 32011 32186 32129 32043
12 118003A Bohle River Hervey Range Road 19.32 146.70 143 1/04/1985 " 32040 33171 33173 33166 32064 32134 32116 32127 32065 32120
13 119006A Major Creek Damsite 19.67 147.02 468 4/05/1978 " 33171 33173 33166 32040 33002 33171 33151 33159 33226 33173
14 120014A Broughton River Oak Meadows 20.17 146.32 182 5/11/1970 13/04/1999 34036 34002 34084 _ _ 34056 34037 34040 34036 34002
15 120216A Broken River Old Racecourse 21.18 148.43 78 1/06/1969 " 33172 33099 33191 33152 33021 33172 33197 33182 33108 33183
16 124002A St. Helens Creek Calen 20.90 148.75 129 7/02/1973 " 33191 33152 33172 33099 33021 33080 33191 33010 33109 33102
17 125005A Blacks Creek Whitefords 21.32 148.82 505 12/12/1973 " 33090 33172 33152 33191 33145 33090 33178 33182 33301 33172
18 130207A Sande Creek Clermont 22.78 147.57 409 21/01/1965 " 35010 35104 36047 35147 35098 35252 35011 35171 35158 35183
19 136108A Monal Creek Upper Monal 24.60 151.10 92 15/07/1962 " 39148 39104 39330 39089 39334 39129 39202 39119 39175 39292
20 137101A Gregory River Burrum Highway 25.08 152.23 454 10/02/1966 " 40624 39303 39128 40126 40451 39027 39159 39207 40624 39188
21 138110A Mary River Bellbird Creek 26.62 152.70 486 1/10/1959 " 40106 40386 40102 40062 40139 40051 40698 40018 40298 40118
22 141009A North Maroochy River Eumundi 26.48 152.95 38 15/02/1982 " 40059 40386 40106 40282 40496 40078 40059 40561 40053 40257
23 143110A Bremer River Adams Bridge 27.82 152.50 125 30/09/1968 " 40135 40004 40677 40082 41018 40400 40183 40675 40392 40447 Basin ID – Unique identifier of stream gauging stations used by Queensland Department of Natural resources and Mines (NRM) for Queensland catchments
Modelling losses in flood estimation
5-3
Table 5.1 List of the selected stream gauge, pluviograph and daily rainfall stations (continued)
SL BASIN ID NAME LOCATION LA.. LOG.AREA(Km2) ST. DATE FIN. DATE PluSta 1 PluSta 2 PluSta 3 Plusta 4 plusta 5 DaRaSta 1 DaRaSta2 DaRaSta3 DaRaSta4 DaRaSta5
24 143212A Tenhill Creek Tenhill 27.55 152.38 447 18/03/1968 " 40082 40270 40135 41018 40004 40570 40716 40114 40079 40835
25 145003B Logan River Forest Home 28.20 152.77 175 1/10/1953 " 40677 40676 40135 41056 40192 40597 40394 40832 40619 40012
26 145010A Running Creek 5.8km Deickmans Bridge 28.23 152.88 128 26/11/1965 " 40192 40676 40757 40677 40606 40363 40006 40653 40116 40362
27 145011A Teviot Brook Croftby 28.13 152.57 83 7/02/1966 " 40677 40135 41056 40676 41044 40490 40485 40030 40876 40267
28 145101D Albert River Lumeah Number 2 28.05 153.03 169 1/10/1953 " 40676 40192 40750 40197 40606 40413 40044 40866 40407 40182
29 146014A Back Creek Beechmont 28.12 153.18 7 5/06/1971 " 40192 40750 40606 40584 40197 40742 40015 40162 40173 40882
30 416410A Macintyre Brook Barongarook 28.43 151.45 465 15/06/1967 " 41457 41175 41060 41465 41413 41377 41014 41034 41447
31 422321B Spring Creek Killarney 28.35 152.32 35 1/10/1972 " 41056 41465 41044 40677 40135 41085 41134 41208 41165 41093
32 422338A Canal Creek Leyburn 28.02 151.58 395 27/03/1972 " 41060 41063 41457 41018 41044 41084 41118 41344 41315 41104
33 422394A Cadamine River Elbow Vally 28.37 152.13 325 2/12/1972 " 41056 41465 41044 40677 40135 41056 41463 41381 41186 41233
34 913005A Paroo Creek Damsite 20.33 139.52 305 20/11/1968 1/10/1988 29127 29130 29009 37010 29090 29109 29107 29118 29064 29128
35 913009A Gorge Creek Flinders Highway 20.68 139.63 248 13/11/1970 Continue 29127 29130 29009 37010 29090 29126 29130 29125 29127 29128
36 915205A Malbon River Black Gorge 21.05 140.06 425 1/10/1970 1/10/1988 29130 29127 29009 29090 37010 29133 29138 29161 29119 37110
37 916002A Norman River Strathpark 19.53 143.25 285 1/10/1969 30/09/1988 30122 30121 30082 30112 _ 30150 30019 30060 30056 30086
38 916003A Moonlight Creek Alehvale 18.27 142.33 127 1/10/1969 10/04/1989 29102 29101 29103 _ _ 29102 29103 29101 29096 29012
39 917005A Agate Creek Cave Creek Junction 18.93 143.47 228 1/07/1969 30/09/1988 30121 30122 30112 30082 30126 30122 30121 30104 30086 30060
40 917007A Percy River Ortana 19.15 143.48 445 2/09/1969 30/09/1988 30122 30121 30112 30082 30126 30104 30086 30060 30114 30019
41 917107A Elizabeth Creek Mount Surprise 18.13 144.30 585 23/07/1968 Continue 30036 30126 32018 31046 30120 30116 30154 30036 30126 31128
42 917114A Routh Creek Beef Road 18.28 143.70 81 11/12/1972 30/09/1988 30120 30036 30126 30112 _ 30111 30018 30103 30110 30013
43 919201A Palmer River Goldfields 16.10 144.77 530 11/12/1967 Continue 31055 28004 31016 31017 _ 28013 31114 31009 31110 31019
44 919205A North Palmer River 4.8 km 16.00 144.28 430 16/10/1973 30/09/1988 28004 31017 31016 31055 _ 28002 28021 28013 31110 31009
45 921001A Holroyd River Ebagoola 14.23 143.15 365 19/01/1970 17/05/1988 27006 28004 27045 27042 31017 27028 27005 28017 28014 27032
46 922101B Coen River Racecourse 13.95 143.17 166 10/11/1967 Continue 27006 27029 27042 _ _ 27005 27006 28017 27028 27049
47 926002A Dulhunty River Dougs Pad 11.83 142.42 325 18/11/1970 " 27042 27045 27022 _ _ 27050 27013 27030 27012 27015
48 926003A Bertie Creek Swordgrass Swamp 11.82 142.50 130 10/11/1972 " 27042 27045 27022 _ _ 27050 27013 27030 27012 27015
Modelling losses in flood estimation
5-4
To compute the weighting average rainfall easily this list was prepared. Because in
many investigations it was found that more than one pluviograph station or daily
rainfall station jointly represent the catchment average rainfall. Sometimes it was
found that there are only 3 or 4 pluviograph stations which are within or near the
candidate catchments, in that case only 3 or 4 pluviograph stations are illustrated in
this list instead of 5 pluviograph stations.
5.3 Streamflow data
In each of the selected study catchments, there was a stream gauging station at the
catchment outlet. A total of 48 stream gauging stations were selected for this
research work. The distribution of streamflow records for the selected 48 stream
gauging stations is shown in Figure 5.1. It also shows that the standard deviation and
the mean value of the stream flow data were 8.82 and 29.5 years respectively.
Record length of stream gauge (Year)
50.045.040.035.030.025.020.015.010.0
Freq
uenc
y
16
14
12
10
8
6
4
2
0
Std. Dev = 8.82 Mean = 29.5
N = 48.00
Figure 5.1 Histogram showing the distribution of streamflow record lengths
Modelling losses in flood estimation
5-5
A cumulative plot illustrating the number of stream gauging station and their stream
flow record length is shown in Figure 5.2. The horizontal line with in the cumulative
plot indicates the median value. The range of stream flow data varies from 11 years
to 48 years. In Figure 5.2 it is shown that the median value of the stream flow data
was 31 years.
All the range of streamflow data were put together and their data range at 10.0%,
25.0% and 90.0% are described. The following statistics were observed from the
streamflow data range.
• 10.0% of the range of streamflow stations has record length more than 39
years.
• 25.0% of the range of streamflow stations has record length more than 34
years.
• 90.0% of the range of streamflow stations has record length more than 17
years.
0
10
20
30
40
50
60
1 5 9 13 17 21 25 29 33 37 41 45
No of stream gauge stations
Stre
am g
auge
reco
rd le
ngth
(Yea
rs)
Median Value
Figure 5.2 Cumulative plot showing the range and median value of
streamflow data record lengths
Modelling losses in flood estimation
5-6
5.4 Pluviograph data
A total of 132 pluviograph stations were selected depending upon the location and
record length of data from the whole of Queensland. All these 132 pluviograph
stations were within or near the selected catchment boundaries. The procedure which
was adopted to select the representative pluviograph station of the selected
catchments is described below.
An electronic layer of 132 pluviograph stations was laid on an electronic layer of
selected Queensland catchment boundaries. This process was performed by MapInfo
(1998) GIS software. After selecting the representative pluviograph station for each
catchment it was found that a total of 82 pluviograph stations were needed to be used
in this research project. The distribution of these 82 pluviograph data record lengths
is shown in Figure 5.3 in the form of a histogram. Figure 5.3 shows the standard
deviation and the mean value of the pluviograph data were 11 and 19 years
respectively.
Modelling losses in flood estimation
5-7
Record length of pluviograph data (Year)
45.042.5
40.037.5
35.032.5
30.027.5
25.022.5
20.017.5
15.012.5
10.07.5
5.0
Freq
uenc
y
12
10
8
6
4
2
0
Std. Dev = 10.91 Mean = 19.3
N = 82.00
Figure 5.3 Histogram showing the distribution of pluviograph record lengths
A cumulative plot illustrating the number of pluviograph station and the pluviograph
data record length is shown in Figure 5.4. The range of the pluviograph data record
length varies from 4 years to 46 years. The horizontal line within the cumulative plot
of Figure 5.4 shows the median value of pluviograph data, which was 18 years.
All the range of pluviograph data were put together and their data range at 10.0%,
25.0% and 90.0% are described below:
• 10.0% of the range of pluviograph stations has record length more than 33
years.
• 50.0% of the range of pluviograph stations has record length more than 18
years.
• 90.0% of the range of pluviograph stations has record length more than 5
years.
Modelling losses in flood estimation
5-8
05
101520253035404550
1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 81
No of pluviograph stations
Plu
viog
raph
reco
rd le
ngth
(Yea
rs)
Median value
Figure 5.4 Cumulative plot showing the range and median value of the
pluviograph data record lengths
5.5 Daily rainfall data
A total of 338 daily rainfall stations were selected depending upon the location and
record length of data from the whole of Queensland. An electronic layer of 338 daily
rainfall stations was laid on electronic layers of selected Queensland catchment
boundaries and selected pluviograph stations. This process was executed using
MapInfo professional version 5.0, GIS software. A total of 24 daily rainfall stations
were selected for this analysis. The daily rainfall stations data were needed to be
disintegrated by the temporal pattern of the nearest pluviograph station. The
distribution of the selected 24 daily rainfall data record lengths is shown in Figure
5.5 in the form of a histogram. All these selected 24 daily rainfall stations were
selected from within or near the 48 unregulated rural catchment boundaries. Figure
5.5 also shows that the standard deviation and the mean value of the daily rainfall
data were 28 and 41 years respectively.
Modelling losses in flood estimation
5-9
Record length of daily rainfall data (Year)
110.0100.0
90.080.0
70.060.0
50.040.0
30.020.0
10.0
Freq
uenc
y
7
6
5
4
3
2
1
0
Std. Dev = 28.30
Mean = 41.2
N = 24.00
Figure 5.5 Histogram showing the distribution of daily rainfall record lengths
A cumulative plot illustrating the number of daily rainfall stations and their daily
rainfall data record length is shown in Figure 5.6. The range of the daily rainfall data
record lengths varies from 7 years to 112 years. In Figure 5.6 it is shown that the
median value of the daily rainfall data was 30.5 years.
All the range of daily rainfall data were put together and their data range at 10.0%,
25.0% and 90.0% are described below:
• 10.0% of the range of daily rainfall stations has record length more than 82
years.
• 50.0% of the range of daily rainfall stations has record length more than 30
years.
• 90.0% of the range of daily rainfall stations has record length more than 9
years.
Modelling losses in flood estimation
5-10
0
20
40
60
80
100
120
1 2 3 4 5 6 7 8 9 101112131415161718192021222324
No of daily rainfall stations
Dai
ly ra
infa
ll re
cord
leng
th (Y
ears
)
Median value
Figure 5.6 Cumulative plot showing the range and the median value of daily
rainfall record lengths
5.6 Formatting stream flow, rainfall and pluviograph data
The stream flow data which was obtained from the Department of Natural Resources
and Mines (DNRM) had to be formatted for use in this research project. Special
purpose programs (FORTRAN PowerStation 4.0) were developed in this research to
format and analyse the voluminous streamflow and rainfall data efficiently and
rapidly and to provide the results in the required format. The stream flow data was
supplied in two stages, both the sets of streamflow data were formatted using two
specially developed computer programs for this study. Both the computer programs
SFFOR1 and SFFOR2 are illustrated in Appendix A and Appendix B.
The pluviograph data obtained from the Bureau of Meteorology (BOM) was
processed using the software called Hydsys Version 8.0 (Hydsys, 2002). The daily
rainfall data obtained from the Bureau of Meteorology and was needed to format this
data to use in this research project. The daily rainfall data was formatted using a
specially developed computer program (FORTRAN PowerStation 4.0)
RAINDATA3 illustrated in Appendix C. The daily rainfall data was used in places
where there was no pluviograph data which was representative to the catchment
Modelling losses in flood estimation
5-11
rainfall. The catchment weighting average rainfall (CWAR) was estimated for
catchments with more than one pluviograph or daily rainfall stations as was
discussed in section 5.7.
5.7 Process to estimate the weighting average rainfall of a catchment
for more than one pluviograph and daily rainfall stations
A computationally simpler approximation to the conventional procedure of Thiessen
polygons was used in this research. Depending on the size of the catchment, grids
ranging from 40×40 to 80×80 over the catchment, were used to divide the
catchments into small square elements. Example of the grid lines is shown in Figure
5.7. Each element located within the catchment boundary was assigned to the nearest
pluviograph or daily rainfall station. Only the stations having data for the particular
time period for which weighting factors are sought were considered. Weighting
factors were calculated by taking the proportion of elements assigned to each station.
Given the uncertainty about the true areal variation of rainfall over the catchment, the
errors associated with this approximation were considered negligible (Siriwardena,
L., & Weinmann, P. E., 1996).
Modelling losses in flood estimation
5-12
Figure 5.7 Example of the grids ranging from 40×40 to 80×80 depending on the
catchment size
S1 = Stream gauging station
D1 = Daily rainfall station site1
P1 = Pluviograph station site 1
The Figure 5.7 shows the example of a catchment with more than one pluviograph
and daily rainfall stations. For this type of catchments weighting average rainfall for
the catchment was estimated. To compute the weighting average rainfall, the daily
rainfall data was disintegrated by the temporal pattern of the nearest pluviograph data
1 1 2 3
4 5 6
7 8 9
* s1
*P2 1
*D1
1 2 3
*D2
4 5 6
* P1
7 8 9
* s1 *D3
Modelling losses in flood estimation
5-13
as it was assumed that the temporal patterns of the daily rainfall data would be
similar to the temporal pattern of the nearest pluviograph data. The data of daily
rainfall stations D1 and D2 were disintegrated using temporal pattern of pluviograph
station P2, as P2 and is closer to D1 and D2 than any other pluviograph station.
Similarly the daily rainfall data of D3 was disintegrated using the temporal pattern of
pluviograph station P1, as P1 and is closer to D3 than any other pluviograph station as
shown in Figure 5.7.
As it is mentioned earlier that each small element located within the catchment
boundary was assigned to the nearest pluviograph or daily rainfall station. Figure 5.7
illustrated that, the small square elements of area 4, area 7, area 8 and most of the
small elements of area 5 were assigned to pluviograph station P1. Small elements of a
rea 1 and most of the small elements of area 2 were assigned to daily rainfall station
D1. Similarly the small square elements of area 6 and area 9 were assigned to daily
rainfall station D2 and the small square element of area 3 was assigned to pluviograph
station P2. CWAR was expressed in the form of equation 5.1:
CWAR = P1 * 0.50 + D1 * 0.25 + D2 * 0.125 + P2 * 0.125 (5.1)
Where
CWAR = Catchment weighting average rainfall
Modelling losses in flood estimation 5-14
5.8 Selection of rainfall and streamflow events
The streamflow, pluviograph and daily rainfall data for any specific catchment were
made concurrent to select the rainfall and streamflow events. The ARR (I. E. Aust.,
1998) losses are based on larger flood events as mentioned in Section 2.6.2. That is,
many heavy rainfall events that did not produce significant runoff due to higher
losses were not considered. This resulted in smaller recommended loss values. In
contrast, Hill et al. (1996a, b) and Rahman et al. (2001a) considered rainfall events
that have higher relative intensities irrespective of associated runoff in the
computation of losses. This approach is more logical and results in more appropriate
design loss values and was adopted in this study. In the approach of Rahman et al.
(2001a) partial series ‘complete storm’ events that have the potential to produce
significant runoff were selected. Figure 5.8 shows an example of a rainfall event
where the ‘complete storm’ is between the start of storm and end of gross storm.
A complete storm was defined by Rahman et al. (2001a) as follows:
• Step 1: A ‘gross’ storm is a period of rain starting and ending by a non-dry hour
(i.e. hourly rainfall > C1 mm/h), preceded and followed by at least six dry hours
(Figure 5.8), where C1 mm/h is a rainfall intensity.
• Step 2: ‘Insignificant rainfall’ periods at the beginning or at the end of a gross
storm, if any, are then cut off from the storm, the remaining part of the gross
storm being called the ‘net’ storm. [A period is defined as ‘having insignificant
rainfall’ if all individual hourly rainfalls ≤ C2 (mm/h), and average rainfall
intensity during the dry period ≤ C1 (mm/h)], where C2 mm/h is a rainfall
intensity.
• Step 3: The net storms, from now on simply referred to as complete storms, are
then evaluated in terms of their potential to produce significant storm runoff. This
is performed by assessing their rainfall magnitudes, i.e. by comparing their
average intensities with threshold intensities. A net storm is only selected for
further analysis if the average rainfall intensity during the entire storm duration
Modelling losses in flood estimation 5-15
(RFID) or during a sub-storm duration (RFId), satisfies one of the following two
conditions:
RFI F ID D≥ ×1 2
RFI F Id dmax ≥ ×2 2
where
2ID = 2 year ARI intensity for the selected storm duration D
2Id = Corresponding intensity for the sub-storm duration d
F1 = Areal reduction factors
F2 = Areal reduction factors
The values of 2ID and 2Id were estimated from the design rainfall data in ARR (I. E.
Aust., 1998). In the above event definition, the use of appropriate areal reduction
factors F1 and F2 allows the selection of only those events that have the potential to
produce significant storm runoff. The use of smaller values of F1 and F2 captures a
relatively larger number of events; appropriate values need to be selected such that
events of very small average intensity are not included. In ARR (I. E. Aust., 1998)
the areal reduction factor is typically 0.9 for this range of catchment sizes and storm
durations. But the recommendation of areal reduction factor value in ARR (I. E.
Aust., 1998) was estimated from the investigation performed for United States
catchments not for Australian catchments. In this study, the following parameters
which are the areal reduction factors and hourly rainfall values were considered same
like Rahman et al. (2000). Rahman et al. (2001b) used these parameter values F1 =
0.4, F2 = 0.5, C1 = 0.25 (mm/h), and C2 = 1.2 (mm/h) for flood frequency analysis of
Victorian catchments. This typically resulted in 3 to 6 rainfall events, on average,
being selected per year for a catchment.
Modelling losses in flood estimation 5-16
Figure 5.8 Rainfall events: complete storms and storm-cores
5.9 Summary
A total of 48 streamflow gauging stations were selected. A total of 82 pluviograph
stations and 24 daily rainfall stations were selected within and near the catchment
boundaries on the basis of location and record length of data range. The range of
stream flow data was 11 to 48 years. The pluviograph data range was from 4 to 48
years and similarly the daily rainfall data range was from 7 to 112 years.
The stream flow data was collected from the Department of Natural Resources and
Mines in two different stages and the data sets were compiled in two different
formats. Both the stream flow data sets were formatted using two different computer
programs SFFOR1 and SFFOR2 enclosed in Appendix A and Appendix B. Similarly
the pluviograph data received from the Bureau of Meteorology were extracted using
Hydsys version 8.0. The daily rainfall data supplied from Bureau of Meteorology
were formatted using computer program RAINDATA3 enclosed in Appendix C.
01234
56789
10
1 5 9 13 17 21 25 29 33 37 41 45
Time (h)
Rai
nfal
l int
ensi
ty (m
m/h
) Start of storm
Storm-core
End of net storm
End of gross storm
01234
56789
10
1 5 9 13 17 21 25 29 33 37 41 45
Time (h)
Rai
nfal
l int
ensi
ty (m
m/h
) Start of storm
Storm-core
End of net storm
End of gross storm
01234
56789
10
1 5 9 13 17 21 25 29 33 37 41 45
Time (h)
Rai
nfal
l int
ensi
ty (m
m/h
) Start of storm
Storm-core
End of net storm
End of gross storm
01234
56789
10
1 5 9 13 17 21 25 29 33 37 41 45
Time (h)
Rai
nfal
l int
ensi
ty (m
m/h
) Start of storm
Storm-core
End of net storm
End of gross storm
Complete storm
Modelling losses in flood estimation 5-17
The pluviograph data and the disintegrated daily rainfall data were not only used but
also the weighting average rainfall was used to obtain the representative rainfall of
the catchment. The streamflow, pluviograph and daily rainfall data were made
concurrent before selecting the rainfall streamflow event and were investigated as to
how streamflow varies with the variation of rainfall.
Modelling losses in flood estimation 6-1
CHAPTER 6
METHODOLOGY FOR LOSS COMPUTATION
6.1. General
Losses during rainfall events can be represented in several ways. There are
mathematical equations to estimate losses, however, in practice, losses are frequently
represented by conceptual models. These types of conceptualised models do not
consider the spatial variability or the actual temporal pattern of storm losses.
Consideration of the model parameters is obtained using the total catchment response
i.e. runoff. These, spatially lumped loss models are widely used because of their
simplicity and ability to approximate catchment runoff behaviour. Some of the most
frequently used methods for spatially lumped losses include (i) constant loss rate, (ii)
initial loss-continuing loss, (iii) proportional loss rate, (iv) antecedent precipitation
index and (v) SCS curve number procedure.
Initial loss and continuing loss are two parameters simplified from the loss concept to
represent the average situation of a storm over the catchment with time. This method
is widely used in Australian practice and hence adopted in this study. This Chapter
describes how loss values were computed in this research. It also describes how loss
values vary with the catchment size.
The initial loss value was computed for all the selected rainfall steamflow events. It
was observed that less than 1.0% of total selected events based on initial loss were
more than 200 mm. Similarly, it was observed that less than 1.5% of the total
selected events based on continuing loss were more than 20 mm/h. Hence a upper
limit of initial loss 200 mm and continuing loss 20 mm/h was imposed in this
research.
Modelling losses in flood estimation 6-2
6.2 Runoff generation mechanism
There are two types of runoff generating mechanisms reported in literature (I. E.
Aust., 1998). Runoff being produced by infiltration excess i.e. when the rainfall
intensity exceeds the infiltration capacity, the rainfall excess which occurs at the
ground surface produce storm runoff. The other runoff generating mechanisms is
runoff being produced by saturated overland flow i.e. when the surface horizon of the
part of a catchment becomes saturated as a result of either the build-up of a saturated
zone above a soil horizon of lower hydraulic conductivity, or due to the rise of a
shallow water table to the surface causing saturated overland flow. Further rain on
the saturated soil then converts 100% surface runoff whereas little or no runoff may
occur on areas where the surface is not saturated.
There is another type of runoff known as throughflow. The storm runoff that occurs
due to water that infiltrates into the soil and percolates rapidly, largely through macro
pores such as cracks, root holes, worm holes and animal holes, and then moves
laterally in a temporarily saturated zone above a layer of low hydraulic conductivity
is known as throughflow. It reaches the stream channel relatively quickly and differs
from other subsurface flow by the rapidity of its response and possibly by its
relatively large magnitude.
Use of a large storm from which runoff is likely to occur from the whole catchment
would be the most appropriate design case. Therefore the concept of runoff produced
by infiltration excess was adopted in this study.
6.3 Method of Initial loss computation
The rainfall which occurs prior to the commencement of surface runoff was
considered to be as initial loss (IL) is shown in Figure 6.1. The same procedure was
adopted in this research to compute the initial loss i.e. amount of rainfall saturated
the catchment before the surface runoff. The estimation of initial loss from observed
storms was a somewhat arbitrary and subjective procedure, mainly due to difficulties
in the determination of the beginning and end time of initial loss. In general, most
Modelling losses in flood estimation 6-3
hydrographs are characterised by an initial rise with a very mild gradient followed by
a steep rising limb. The end time for initial loss computation was considered as the
start of the steep rising limb, as shown in Figure 6.1.
Figure 6.1 Initial loss – Continuing loss (IL-CL) model
In computing initial loss, a surface runoff threshold value equal to 0.01 mm/h was
used, similar to Hill et al. (1996a). It was considered that surface runoff commences
when the surface runoff threshold was exceeded. The initial loss occurs in the
beginning of the storm, prior to the commencement of the surface runoff. Figure 6.1
shows the example of an initial loss-continuing loss (IL-CL) model. As IL-CL model
computes the loss values from rainfall and streamflow data. This model can not
identify the start of surface runoff of a catchment for any rainfall event. The hourly
streamflow data was observed and compared with the surface runoff threshold value
in volume. When the difference of streamflow volume exceeds the surface runoff
threshold value volume of rainfall, then the rainfall that occurred before the rise of
streamflow in a rainfall event was considered as the initial loss. Figure 6.1 shows a
12 hours rainfall event, out of this rainfall event the first 5 hours were considered as
initial loss as there was no rise of streamflow occurred.
02468
1012141618
1 2 3 4 5 6 7 8 9 10 11 12
Rainfall duration (h)
IL (m
m) a
nd C
L (m
m/h
)/ st
ream
flow
(m3/
s)
IL
Streamflowhydrograph
02468
1012141618
1 2 3 4 5 6 7 8 9 10 11 12
Rainfall duration (h)
IL (m
m) a
nd C
L (m
m/h
)/ st
ream
flow
(m3/
s)
IL
Streamflowhydrograph
CL
Modelling losses in flood estimation 6-4
Surface runoff threshold values equal to 0.02 mm/h and 0.005 mm/h was also used
for all the selected rainfall events in this research to compute the variation of loss
values with different threshold values. The results of the loss variations due to the
change of threshold values are discussed in chapter 8. A lower limit of 0.0mm and an
upper limit of 200.0mm were imposed for the initial loss computation and events
outside of this range were excluded from the analysis.
When the catchment was extremely saturated i.e. when the initial loss was very low,
it was observed that sometimes the initial loss values were lower than continuing loss
values. It was also observed in this analysis that in a few rainfall streamflow events,
the initial loss value was found lower than the continuing loss value not because the
catchment was extremely saturated but due to overlapping of two rainfall events
upon each other in one complete rainfall event. Events that overlapped each other in
one complete event were excluded in this analysis.
In the IL-CL model losses are estimated from observed rainfall and stream flow
events. The same procedure was adopted in this research work for the estimation of
losses. Also in the IL-CL model, they do not consider spatial variability or real
temporal pattern of losses. In this research work the spatial variability or real
temporal pattern of losses were not considered.
6.4 Method of continuing loss computation
In ARR (I. E. Aust., 1998) the continuing loss is defined as the loss that occurs at a
constant rate after the commencement of the surface runoff. The continuing loss in
ARR (I. E. Aust., 1998) is the average rate of loss that occurred during the remainder
of the storm. The procedure which was adopted in this analysis to compute the
continuing losses was the same as the procedure adopted in ARR (I. E. Aust., 1998)
i.e. the continuing loss is the rate of loss that occurred during the remainder of the
storm.
The rates of continuing loss are constant as recommended in ARR (I. E. Aust., 1998),
but the value could be decreasing with time depending upon the soil cover and
Modelling losses in flood estimation 6-5
duration of the storm. It was investigated in this analysis whether continuing loss rate
is constant in nature or continuing loss rate decays with the duration of the storm.
The continuing loss rate is a decaying function of duration, which is an assumption.
In this analysis initial loss and continuing loss model was used to compute the initial
loss and continuing loss values from the rainfall and streamflow events. In ARR (I.
E. Aust., 1988) the recommended design median initial loss ranging from 15.0 mm –
35.0mm and the recommended design median continuing loss 2.5 mm/h for eastern
catchments. Similarly for western Queensland catchments the recommended median
continuing loss is 1.4 mm/h. This shows that the recommended design initial loss
varies with the duration of the storm but the recommended design continuing loss
does not vary but remain constant throughout the duration of the storm.
Table 8.1 shows that the derived median continuing loss values for all the 48 selected
catchments are from 0.71mm/h – 5.8 mm/h. Hence it is observed that for Queensland
catchments the continuing loss varies with the duration of the storm rather than it
remains constant throughout the storm.
The rate of loss that occurred during the remainder of the storm after the initial loss
was satisfied was considered as continuing loss. In Figure 6.1 (section 6.3) the loss
that started from the 6th hour and continued up to the 12th hour was the continuing
loss. The continuing loss was computed and averaged over the time period for the
remainder of the storm at hourly timesteps.
In actual rainfall events the continuing loss continues until the end of surface runoff.
As this research was performed with the real pluviograph and daily rainfall data, it
was easy to compute the duration of surface runoff of the same rainfall event. In this
research continuing loss was also estimated up to the end of the surface runoff, to
investigate how much actual continuing loss values varied with the design
recommended Queensland continuing loss values. It was observed that usually the
surface runoff from a rainfall event continues even after the end of the storm.
Therefore the continuing loss which was averaged over the time period computed up
to the end of the storm always estimates higher loss value than the continuing loss
value which was averaged over the time period computed up to the end of surface
Modelling losses in flood estimation 6-6
runoff. In some of the rainfall streamflow events, the duration of the remainder of the
storm is longer than the duration of the surface runoff i.e. the rainfall continued even
after the end of the surface runoff. In these events when the rainfall was very low in
intensity and did not show any variation in the resulted streamflow, then the duration
of the remainder of the storm and the duration of the surface runoff for those same
rainfall streamflow events were considered equal. For these types of rainfall
streamflow events the continuing loss values computed were the same under both the
conditions. Some of the events from stream gauge stations 105015A, 107001B and
107003A were examples of these types of events.
The water balance equation from the start of a rainfall event and the end of a runoff
event may be expressed as:
R = IL + CL*t + QF (6.1)
where R = Total rainfall of the event expressed in average depth of rainfall in mm
over the catchment.
QF = Quickflow, assumed to be resulted from the rainfall event, expressed in mm.
t = Time elapsed between the start of the surface runoff and end of the rainfall event
(hours).
Since, QF is the total streamflow (SFT) minus baseflow (BF), equation (6.1) may be
written as:
R = IL + CL*t + SFT – BF (6.2)
Where both SFT and BF are expressed in mm.
As IL-CL model does not consider the temporal variability of losses. From equation
(6.1) CL may be expressed as:
CL = (R – IL – QF)/ t (6.3)
Modelling losses in flood estimation 6-7
To estimate QF in equation (6.3), separation of baseflow from total streamflow was
required, as discussed in chapter 7. A lower limit of 0.0mm/h and an upper limit of
20.0mm/h were imposed for the continuing loss computation and events outside of
this range were excluded from this analysis. As continuing loss value more than 20.0
mm/h, needs more detailed investigation.
6.5 Method of proportional loss computation
In ARR (I. E. Aust., 1998) the proportional loss is defined as an initial loss followed
by a loss consisting of a constant fraction of the rainfall in the remaining time period.
The proportional loss is assumed to be a constant fraction of the rainfall after the
commencement of surface runoff. The proportional loss was computed as:
Proportional loss = 1 – [(Vol2) / (rainstorm – Ilstorm)] (6.4)
Rainstorm = Total rainfall volume of the storm
Ilstorm = Volume of initial loss for the storm
Vol2 = Total surface runoff volume for the storm
To evaluate the loss values and investigate how the loss value varies with the
catchment, the proportional loss was also computed for each and every selected
rainfall and streamflow events. Then these computed proportional losses were
observed and compared with the continuing loss values of the same rainfall
streamflow events. The loss model which was used in this research analysis is the
initial loss-continuing loss model. Appendix E shows the results of all the
proportional losses of 48 selected catchments.
6.6 Method of volumetric runoff coefficient computation
The volumetric runoff coefficient was computed as the ratio of the volume of surface
runoff to the total volume of rainfall of the storm (Hill et al., 1996). The same
Modelling losses in flood estimation 6-8
equation was used in this research to compute the volumetric runoff coefficient. The
volumetric runoff coefficient was computed as:
Volumetric runoff coefficient = (Vol2) / (rainstorm) (6.5)
Vol2 = Total surface runoff volume for the storm
Rainstorm = Total rainfall volume of the storm
All the abbreviations were the same as mentioned in Section 6.5. In this research
work the volumetric runoff coefficient was also computed for all the selected rainfall
streamflow events of the catchments. Then these computed volumetric runoff
coefficients were observed and compared with the continuing loss as well as
proportional loss values. Appendix E shows the results of all the volumetric runoff
coefficient losses of 48 selected catchments.
The volumetric runoff co-efficient can not be rational since the runoff co-efficient
must vary with the both loss values and rainfall volume. For catchment where the
soil is more pervious the runoff co-efficient would be smaller. As the impervious
area increases the runoff co-efficient approaches unity. Again when the soil is
saturated even in the case of rural catchments, the runoff co-efficient value will also
be higher depends upon the degree of saturation of the catchment.
6.7 Losses versus catchment size
Catchment size influences storm runoff in a number of direct and indirect ways.
Various aspects of runoff such as total yield, flood peaks, direct storm runoff and
losses are usually affected by the catchment size accordingly.
The resulting streamflow hydrograph of a streamflow depends upon the size of the
catchment. When the catchment size is small to medium, the source area is close to
the rainfall gauging station. The streamflow hydrographs of these catchments tend to
be high peaked with little or no delay between the commencement of rainfall excess
and the start of rise of the streamflow hydrograph. For large catchments the source
Modelling losses in flood estimation 6-9
area is far away from the rainfall gauging station. Hence a considerable delay will
occur between the beginning of rainfall excess and the start of surface runoff at the
catchment outlet, hence the hydrographs tend to be broad crested. Laurenson and
Pilgrim (1963) mentioned that peak discharge is one of the factors affecting the loss
rate magnitude.
It was observed that for larger catchments there is lack of uniformity in catchment
characteristics than in smaller to medium sized catchments. Laurenson and Pilgrim
(1963) mentioned that catchment characteristic is a factor which affects the loss
value.
The rainfall events which are potential to produce significant surface runoff for large
catchments, the rainfall excess would not occur from all over the catchment at the
same time. The rainfall excess would occur only on some portion of the catchment.
This ‘partial area runoff’ can more probably occur in larger catchments because of
the variation of catchment soil conditions, soil moisture content or variation in
rainfall. Care should therefore be exercised in using these values in determining a
design loss rate, especially if the type and condition of soil and soil cover vary
appreciably over the catchment.
From the above discussion it is evident that loss computation for large catchments is
different from small to medium sized catchments. For that reason only small to
medium sized catchments were used in this research to derive the loss values for
design flood estimation.
6.8 FORTRAN program development
To compute losses (continuing loss, proportional loss and volumetric runoff
coefficient) the volume of runoff resulting from the rainfall streamflow event needs
to be estimated. The streamflow data recorded, consists of runoff resulting from
rainfall event and baseflow. A computer program was developed to separate the total
streamflow hydrograph into baseflow and surface runoff, which is illustrated in
Appendix D. The output of this specially developed computer program was checked
Modelling losses in flood estimation 6-10
with the readily available software packages. Then the computed program was used
to compute losses for 48 selected catchments. The program first computed initial loss
then computed continuing loss, proportional loss and volumetric runoff coefficient
using the definitions mentioned in sections 6.3, 6.4 and 6.5.
6.9 Surface runoff threshold value
Initially a surface runoff threshold value equal to 0.01 mm/h was used in computing
all these losses. The surface runoff commenced when the surface runoff threshold
value was exceeded. From this analysis it was observed that with the variation of the
threshold value the loss value also varied. To observe how much loss value varied
with the change of threshold value two more surface runoff threshold values were
used in this research (Table 8.6). There was nothing called most suitable threshold
value for loss computation of different size of catchments. The two other surface
runoff threshold values which were used in this research work were 0.02mm/h and
0.005mm/h respectively. The analysis of how the loss value varied with the change
of the threshold value is discussed in chapter 8.
6.10 Reasons for the variability of calculated loss values
From this investigation it was observed that there are few reasons which affected the
variability in calculated loss values from event to event. These reasons are discussed
below:
• Catchment condition is a factor which affects the loss values from event to
event. For example from the same rainfall event a saturated catchment will
produce larger floods than the dry catchment.
• In a rainfall event, when the rainfall intensity is not uniform throughout the
catchment. A partial area runoff occurs which is another reason for the
variability in calculated loss values from event to event.
Modelling losses in flood estimation 6-11
• The variability of calculated loss values from event to event also occurs with
the variation of the surface runoff threshold value.
6.11 SUMMARY
Initial loss and continuing loss are two parameters simplified from the loss concept to
represent the condition of storm over the catchment with time. The initial loss-
continuing loss method is widely used in Australian practice and hence adopted in
this study. The most appropriate design case would usually involve the use of a large
storm from which runoff is likely to occur from the whole catchment. Therefore the
concept of runoff produced by infiltration excess was adopted in this study.
In ARR (I. E. Aust., 1998) the continuing loss rate is assumed to be constant during
the remainder of the storm. The procedure to derive the continuing loss values, which
was adopted in this research were the same as the procedure adopted in ARR (I. E.
Aust., 1998). In the IL-CL model losses are estimated from observed rainfall and
stream flow events. The same procedure was adopted in this research work for the
estimation of losses. Also in the IL-CL model, they do not consider spatial variability
or real temporal pattern of losses. In this research work the spatial variability or real
temporal pattern of losses were not considered. The proportional loss and the
volumetric runoff coefficient are computed in this analysis using the equations 6.4
and 6.5. Appendix E shows the results of all the initial losses, continuing losses,
proportional losses and volumetric runoff coefficient losses of 48 selected
catchments.
The resulting streamflow hydrograph depends upon the size of the catchment. When
the catchment size is small to medium, the source area is close to the rainfall gauging
station. The streamflow hydrographs of these catchments tend to be high peaked with
little or no delay between the commencement of rainfall excess and the start of rise
of the streamflow hydrograph. For large catchments the streamflow hydrographs tend
to be broad crested. Also the rainfall excess that occurs only on some portion, not the
entire of the catchment, that is, ‘partial area runoff’ situations can occur in larger
Modelling losses in flood estimation 6-12
catchment. For that reason only small to medium sized catchments were used in this
research to derive the loss values for design flood estimation.
The streamflow data recorded, consists of runoff resulting from rainfall event and
baseflow. A computer program was developed to separate the total streamflow
hydrograph into baseflow and surface runoff. Three different surface runoff threshold
values were used in this research to examine the effect of loss values with the
variation of threshold values. There is nothing called most suitable threshold value
for loss computation of different size of catchments.
Modelling losses in flood estimation 7-1
CHAPTER 7
METHOD OF BASEFLOW SEPARATION
7.1 General
The streamflow data measures total flow, which consists of surface runoff (rainfall
excess), interflow or throughflow and baseflow. In this analysis surface runoff was
considered to be the water that enters the stream primarily by way of overland flow
across the ground surface. Baseflow which is groundwater flow included other
components such as interflow or throughflow. It was required to separate the total
streamflow into surface runoff and baseflow to compute the storm losses. This
Chapter describes various methods of baseflow separation and the method that was
selected for this research work.
7.2 Review of baseflow separation process
There have been a number of studies performed regarding the separation of
streamflow on the basis of physical properties, chemical properties or travel time of
the streamflow in order to determine the proportions of the surface runoff and
baseflow at a given time. O’Loughlin et al. (1982), Hill (1993), Nathan and
McMahon (1990) separated the flow components not according to physical sources
of runoff but on the basis of travel times. Separation of surface runoff from baseflow
has been important for methods of real time and design flood estimation. For
example Dickinson et al. (1967) and Hall (1971) have reviewed the streamflow
separation problem from the viewpoint of flood analysis. Also separation of
streamflow is needed for the purpose of synthesising the hydrological behaviour of
catchments in rainfall runoff models. For example Boughton (1987) and Lyne and
Hollick (1979) illustrate the use of streamflow partitioning as a basis for rainfall
runoff modelling.
Modelling losses in flood estimation 7-2
Shirmohammadi et al. (1984) used rainfall data in conjunction with streamflow data
to determine periods of surface runoff. Their method involved making prior
assumptions about a threshold amount of rainfall was needed to initiate surface
runoff and also assumptions were made about the duration of surface runoff.
Pilgrim et al (1979) and Kobayashi (1986) studied the use of specific electrical
conductance of stream waters as a means of estimating the proportion of different
flow components. Kobayashi (1985) also used stream temperatures for partitioning
flows in an area where snowmelt formed a major part of the flow. Hino and Hasebe
(1985) used isotopes of oxygen as a means of separation. Boughton (1988)
mentioned that the data used in these studies are not readily available for routine
partitioning of streamflow, which is the limitation for practical application of these
methods.
Nathan and McMahon (1990) compared two baseflow separation techniques, one
based on a digital filter and the other on a simple smoothing and separation
technique. They noted that compared to the smoothed minima technique, the digital
filter method is better suited to low baseflow conditions.
Lyne (1979) separated streamflow into quick and slow response components using a
recursive digital filter. Also O’Loughlin et al. (1982), Chapman (1987), Nathan and
McMahon (1990) and Hill (1993) have used a similar method for baseflow
separation. The filter is of the simple form:
kf = 1. −kfa +
21 a+ ( ky - )1−ky (7.1)
where,
kf = Fitted quick response at the kth sampling instant
ky = Total streamflow;
a = Filter parameter ( or factor).
Modelling losses in flood estimation 7-3
Boughton (1988) mentioned that the recursive digital filter method (Equation 7.1)
uses only a single parameter (usually in the range of 0.75 to 0.90). Hill et al. (1996)
mentioned that two restrictions are placed on the digital filter: the separated slow
flow is never to be negative and nor it is greater than the original streamflow. The
other mathematical filtering methods used in Australia for partitioning streamflow
are more complex and use several parameters.
Bethlahmy (1971) used a method for baseflow separation using only streamflow
data. In that method, the rate of baseflow at any time (Bi) is made equal to the sum of
the baseflow rate at the previous time (Bi-1) and an incremental value (Ui).
Bi = Bi-1 + Ui (7.2)
Boughton (1988) mentioned that the rationale behind the calculations of the
incremental values (Ui) is not clear and the applicability of this method to Australian
catchments seems dubious.
In this research analysis two simple models of computer separation of baseflow and
surface runoff were compared in Figure 7.1. Both the models allow for user
identification of the point on a hydrograph at which the separation of flow
components was most clear. Figure 7.1 shows a streamflow hydrograph, where point
A and point B indicate the start and end of surface runoff hydrograph. Two models to
separate the baseflow from the streamflow are assumed by joining these two points
either by a straight line (Model 1) or by a smooth curve (Model 2). The methods of
baseflow separation from streamflow hydrograph were performed in using hourly
streamflow data. These methods use manual identification of one or more points that
mark the end of surface runoff. They differ in the assumption of how baseflow
discharge increases with the surface runoff.
Model 1 which was tested is based on the following operational assumptions:
• The maximum increase in baseflow discharge in one time step is limited to a
preset value,
Modelling losses in flood estimation 7-4
• Surface runoff commences when the total streamflow exceeds the baseflow
of the previous day by the preset maximum increase in baseflow discharge;
the new baseflow is made equal to the previous baseflow plus the preset
constant value; surface runoff is the difference between the total flow and the
baseflow,
• Surface runoff ends when the total streamflow does not exceed the baseflow
of the previous day by the preset maximum increase in baseflow discharge,
Model 2 allows for the increase in baseflow discharge to vary depending on the rate
of increase in total flow. The operation of the model is based on the following
assumptions:
• Baseflow increases whenever there is an increase in the rate of total flow,
• The rate of increase in baseflow is calculated as a fraction of the difference
between the total flow and the rate of baseflow on the previous day; the
difference between the total flow and the new baseflow is surface runoff,
• Surface runoff ends when the total flow is less than the baseflow of the
previous day
Model 1 used a constant rate of baseflow increase at each time step i.e. the increase
in baseflow and hence the rate of recharge of baseflow storage was time dependent.
In Model 2 the increase in rate of baseflow by a fraction of the surface runoff i.e.
baseflow increase and hence baseflow recharge was dependent on runoff volume.
Modelling losses in flood estimation 7-5
0
5
10
15
20
25
30
35
40
45
0 5 10 15 20 25Days
Dai
ly d
isch
arge
- M
L
Model 2
Model 1
A
B
Figure 7.1 Comparison of baseflow separation by two different models
Boughton (1988) has shown that both models give similar results in simulating big
runoff events. The main difference between these two models is that Model 1
estimates more surface runoff and less baseflow than Model 2 for the big events.
Also Model 1 estimates less surface runoff and more baseflow than Model 2 for
small runoff events. Model 2 estimates some surface runoff at every rise in the
hydrograph while Model 1 treats many small rises in the hydrograph as increases in
baseflow shown in Figure 7.2.
In general, the results obtained from Model 1 are more sensitive to change in the
value of the constant than the results of Model 2 to change in the value of the surface
runoff fraction (Boughton, 1988).
Modelling losses in flood estimation 7-6
Figure 7.2 Example of comparison of baseflow separation models in large and small
runoff Events (Boughton, 1988)
Dickson et al. (1967) illustrated some different ideas about the pattern of increase of
baseflow during periods of surface runoff. Regarding the numerous small rises in the
streamflow hydrograph, he mentioned that because of the rain falling directly on to
flowing water it should be partitioned as surface runoff in the manner of Model 2.
Dickson et al. (1967) also mentioned that the partition of surface runoff should be in
the manner as shown in of Model 2. It seems unlikely that baseflow discharge would
rise and fall as quickly as is calculated by Model 1 in these small rises in the
hydrograph.
Lyne and Hollick (1979) in Western Australia; and Bethlahmy (1971) in Idaho, USA
used the rate of increase of the baseflow proportional to the rate of surface runoff as
in Model 2. Also it seems highly unlikely that baseflow discharge would rise and fall
as quickly as is calculated by Model 1 in these small rises in the hydrograph.
For these reasons in this study Model 2 was selected as a method for baseflow
separation, i.e. the rate of increase of baseflow depends on the fraction of the surface
runoff (α ). The rate of baseflow at any time step is i (BFi) may be expressed as
Modelling losses in flood estimation 7-7
equal to the baseflow in the previous timestep i-1(BFi-1) plus α times the difference
of total streamflow at step i (SFi) and baseflow at step i-1 (BFi-1).
The Equation, which was used to separate the baseflow from the total streamflow in
this research, is of the form:
BFi = BFi-1+ α (SFi - BFi-1) (7.3)
where
α = The fraction of the surface runoff
BFi = The rate of baseflow at any time step i
BFi-1 = The rate of baseflow in the previous timestep i-1
SFi = Total streamflow at timestep i
7.3 Results For the baseflow separation of rainfall streamflow events of 48 selected catchments it
was necessary to determine an acceptable baseflow separation co-efficient (α). In
this analysis the baseflow separation co-efficient was selected by trial and error
method. Also the sensitivity analysis of the α value was performed using 3 to 4
rainfall streamflow events.
The method of baseflow separation in this study was based on hourly streamflow and
rainfall data. The baseflow separation method Model 2 was applied to all the selected
48 study catchments to determine the value of α for each catchment and finally to
compute their baseflow and continuing loss values. The α value was not attached
with the relative streamflow event size, but it was a fraction of the surface runoff.
The example of the methods used to select the α value for different catchments are
discussed in this Chapter. Two catchments were selected randomly from Queensland
as an example to show how α values were estimated for this research with different
methods. These catchments were the Bremer River (143110A, catchment area 130 sq
km2) and the Tenhill Creek (143212A, catchment area 447 sq km2).
Modelling losses in flood estimation 7-8
Four streamflow events were selected for each catchment. An appropriate value of α
needed to be selected that allows acceptable separation of baseflow for all the events
of the Bremer River and Tenhill Creek catchments. In this research a semi-
logarithmic plot of the recession curve method was applied (I. E. Aust., 1998) to find
the end of surface runoff. In normal scale plot a change from quickflow to baseflow
in a streamflow hydrograph can not be identified. Hence streamflow hydrographs
were plotted on a semi-logarithmic graph (discharge to logarithmic scale, time to
linear scale), as the recession curves on semi-logarithmic graph approximated as
straight line segments. The baseflow separation line which starts from the rising limb
of the streamflow hydrograph (point A in Figure 7.1) and finishes at the start of the
recession curve (point B in Figure 7.1), was considered as the acceptable baseflow
separation line. For the Bremer River catchment a streamflow event (Event 1) was
selected and four different values of α were used to determine a suitable α value,
which would be used for the baseflow separation. In Figure 7.3 it is shown that when
α = 0.005 and α = 0.008 the baseflow separation curve joined above the start of the
straight line segment of the streamflow hydrograph. Hence any of these two values
of α did not provide acceptable baseflow separation. Again, in Figure 7.3 when α =
0.003 the baseflow separation curve joined below the start of the straight line
segment of the streamflow hydrograph. Hence α = 0.003 did not provide acceptable
baseflow separation. The results shown in Figure 7.3 indicate that a value of α =
0.004 provided acceptable baseflow separation for Event 1.
-4
-3
-2
-1
0
1
2
0 50 100 150
Time step
LogQ
α = 0.003
Modelling losses in flood estimation 7-9
-4
-3
-2
-1
0
1
2
0 50 100 150
Time stepLo
gQ
α = 0.004
-4
-3
-2
-1
0
1
2
0 50 100 150
Time step
LogQ
α = 0.005
-4
-3
-2
-1
0
1
2
0 50 100 150
Time step
LogQ
α = 0.008
Figure 7.3 Separation of streamflow components in a semi-log graph for Event 1,
when α = 0.003, 0 .004, 0.005, 0.008 (Bremer River)
Modelling losses in flood estimation 7-10
It was worth examining how sensitive the derived loss values were on the selection
of α values. In ARR (I. E. Aust., 1998) it is mentioned that the maximum baseflow
discharge is well below 10.0% of the maximum discharge. Table 7.1 shows that for
Event 1 of the Bremer River catchment when α = 0.004, CL = 1.16, if α was
increased by 25.0%, the value of CL varied by 1.11%, if α was decreased by 25.0%,
the value of CL varied by 1.38%. A variation of 100.0% in the value of α resulted in
only 4.0% variation in CL. This showed that a small error in selecting appropriate
value of α would not affect the value of CL significantly. For Event 1 in the Bremer
River catchment, it appeared that an α = 0.004 was acceptable.
Table 7.1 Effects of changing α value on CL, PL (proportional loss) and Volumetric runoff co-efficient Catchment ID = 143110A Catchment name = Bremer River
Event No α CL % differ. PL % differ. Vol.r.c % differ.
Event 1 0.003 1.147 1.380 0.624 1.270 0.112 2.750
0.004 1.163 0.632 0.109
0.005 1.176 1.118 0.641 1.424 0.107 1.835
0.008 1.211 4.127 0.659 4.272 0.101 7.339
Event 2 0.003 0.010 71.430 0.007 72.000 0.381 1.870
0.004 0.035 0.025 0.374
0.005 0.057 62.857 0.041 64.000 0.368 1.604
0.008 0.114 225.714 0.082 228.000 0.352 5.882
Event 3 0.003 0.179 17.510 0.118 15.710 0.614 3.020
0.004 0.217 0.140 0.596
0.005 0.251 15.668 0.163 16.429 0.581 2.517
0.008 0.344 58.525 0.223 59.286 0.539 9.564
Event 4 0.003 0.915 1.290 0.639 1.240 0.144 2.860
0.004 0.927 0.647 0.140
0.005 0.938 1.187 0.655 1.236 0.137 2.143
0.008 0.967 3.092 0.675 4.328 0.129 7.857
It was then examined whether the best value selected for Event 1 (α = 0.004), gives
an acceptable baseflow separation for other events of the same catchment. Three
more events were selected from the Bremer river catchment and α = 0.004 was used
for all three events to investigate the acceptable baseflow separation. Figure 7.4
Modelling losses in flood estimation 7-11
shows that the value of α = 0.004 indeed provided acceptable baseflow separation for
the other 3 events of the same catchment.
Thus, it appeared that for the Bremer River catchment a value of α = 0.004 was used
for baseflow separation for all streamflow events.
-0.5
0
0.5
1
1.5
2
0 20 40 60 80
Time step
Log
Q
-3
-2
-1
0
1
2
3
0 50 100 150
Time step
Log
Q
Event 2 Event 3
-3.5-3
-2.5-2
-1.5-1
-0.50
0.51
1.52
0 50 100 150
Time step
Log
Q
Event 4
Figure 7.4 The baseflow separation with α = 0.004 for 3 events in the Bremer
River catchment.
For the catchments where a single α value did not provide the acceptable baseflow
separation, another method was used to select an appropriate α value for the
acceptable baseflow separation. ARR (I. E. Aust., 1998) recommends that the
median value is to be used, where a single value is to be selected from a range of
values. For Events 1, 2, 3 and 4 in the Tenhill Creek catchment, a value of α equals
Modelling losses in flood estimation 7-12
to 0.010, 0.003, 0.008 and 0.002 respectively provided acceptable separation of
baseflow, as shown in Figure 7.5. It was observed that the selected α value for each
event showed a wide variation and none of these values provided acceptable
separation for all four events.
0
0.5
1
1.5
2
0 50 100 150Time step
Log
Q
0
0.5
1
1.5
2
2.5
3
0 50 100 150 200Time step
Log
Q
Event 1 when α = 0.010 Event 2 when α = 0.003
-2
-1
0
1
2
3
0 20 40 60 80
Time step
LogQ
-1-0.5
00.5
11.5
22.5
3
0 50 100 150 200
Time step
Log
Q
Event 3 when α = 0.008 Event 4 when α = 0.002
Figure 7.5 Baseflow separation of four events for Tenhill Creek catchment.
It was then examined whether the median of these four values (0.0055) provided
acceptable separation for all four events. In Figure 7.6, Event 1 showed that the
separation curve between baseflow and surface runoff joined the streamflow
hydrograph in the middle of the recession curve. In Event 2, the baseflow separation
curve joined the streamflow hydrograph at the start of the recession curve. In Event
4, the baseflow separation curve joined the streamflow hydrograph before the start of
the recession curve. Figure 7.6 shows that baseflow separation with α = 0.0055
which was the median value provided a reasonable baseflow separation for all four
events for the Tenhill Creek catchment.
Modelling losses in flood estimation 7-13
0
0.5
1
1.5
2
0 50 100 150Time step
Log
Q
0
0.5
1
1.5
2
2.5
3
0 50 100 150 200Time step
Log
Q
Event 1 Event 2
-2
-1
0
1
2
3
0 20 40 60 80
Time step
LogQ
0
50
100
150
200
250
300
350
30 40 50 60 70
Time step
Q
Event 3(log scale) Event 3 (natural scale)
-1-0.5
00.5
11.5
22.5
3
0 50 100 150 200
Time step
Log
Q
0
50
100
150
200
250
300
350
400
0 50 100 150 200
Time step
Q
Event 4(log scale) Event 4 (natural scale) Figure 7.6 Baseflow separation for the four events with α =0.0055 (Tenhill Creek) The value α (0.0055 was rounded to 0.005) was varied to observe how CL value was
affected. Table 7.2 shows that for Event 1 in the Tenhill Creek catchment when α =
0.005, CL = 1.71, if α was increased by 20%, the value of CL varied by 2.78%, if α
was decreased by 20%, the value of CL varied by 1.46%. Thus a variation of 60% in
the value of α resulted in only 4.8% variation in the CL value. For the Event 2 and
Modelling losses in flood estimation 7-14
Event 3, the variation in α value by about 50% caused about 13% and 2% variation
in CL value. However for Event 4, the CL was more sensitive to the change of α
value. In general it was found that a smaller error in selecting appropriate value of α
did not affect the value of CL significantly for the Tenhill Creek catchment. A single
flood event with total rainfall, flood hydrograph and baseflow separation is shown in
Appendix H.
Table 7.2 Effects of changing value on CL, PL (proportional loss) and Volumetric
runoff coefficient
Catchment ID = 143212A Catchment name = Tenhill Creek
Event No α CL % differ. PL % differ. Vol.r.c % differ.
Event 1 0.004 1.689 1.460 0.757 1.560 0.091 4.600
0.005 1.714 0.769 0.087
0.006 1.736 2.783 0.779 2.906 0.083 8.791
0.008 1.771 4.855 0.795 5.020 0.077 15.385
Event 2 0.004 1.475 4.283 0.503 4.373 0.334 4.702
0.005 1.541 0.526 0.319
0.006 1.593 8.000 0.543 7.952 0.307 8.084
0.008 1.671 13.288 0.57 13.320 0.289 13.473
Event 3 0.004 2.063 0.578 0.694 0.857 0.111 1.835
0.005 2.075 0.700 0.109
0.006 2.087 1.163 0.703 1.297 0.108 2.703
0.008 2.111 2.327 0.711 2.450 0.105 5.405
Event 4 0.004 0.17 24.107 0.111 24.490 0.508 4.312
0.005 0.224 0.147 0.487
0.006 0.27 58.824 0.177 59.459 0.47 7.480
0.008 0.346 103.529 0.226 103.604 0.442 12.992
7.4 Summary
The recursive digital filter method used by many of the researchers for the baseflow
separation, uses only a single parameter (usually in the range of 0.75 to 0.90). Two
restrictions are placed on the digital filter: the separated slow flow is never to be
Modelling losses in flood estimation 7-15
negative and nor it is greater than the original streamflow. The other mathematical
filtering methods used in Australia for partitioning streamflow are more complex and
use several parameters.
Bethlahmy (1971) used a method for baseflow separation using only streamflow
data. In that method, the rate of baseflow at any time (Bi) is made equal to the sum of
the baseflow rate at the previous time (Bi-1) and an incremental value (Ui). Boughton
(1988) mentioned that the rationale behind the calculations of the incremental values
(Ui) is not clear and the applicability of this method to Australian catchments seems
dubious.
Two simple models of computer separation of baseflow and surface runoff were
compared in this study. Both the models allow for user identification of point on a
hydrograph at which the separation of flow components was most clear. The method
of baseflow separation from streamflow was performed in using hourly streamflow
data. The two models of baseflow separation were compared to select a suitable
baseflow separation method for this research work. The baseflow separation method
Model 2 was selected for this analysis and was applied to all the selected 48 study
catchments to determine the α value for each catchment and finally to compute their
continuing losses, proportional losses and volumetric runoff coefficients. The results
of only two catchments are discussed in this Chapter as an example.
The following conclusions were drawn from the study:
• An acceptable baseflow separation coefficient (α) was selected for the
catchments by trial and error and also their sensitivity analysis was performed
using 3 to 4 rainfall streamflow events.
• The continuing loss was not very sensible to the selection of α value in the
baseflow separation. It was found that a change in α value by about 50% made
less than about 10% variation in the value of continuing loss. The results of the
loss values derived from this research are discussed in the next chapter.
Modelling losses in flood estimation 8-1
CHAPTER 8
RESULTS AND DISCUSSIONS 8.1 General
The design loss values which were derived from this research are presented and
discussed in this chapter. A total of 969 rainfall and streamflow events were
identified for the 48 unregulated rural catchments as shown in Table 8.1. A lower
limit of 0.0mm and an upper limit of 200.0mm were imposed for the initial loss
computation i.e. rainfall streamflow events which estimated initial loss values
outside of this range (0.0mm-200.0mm) were excluded from this analysis. Similarly
a lower limit of 0.0mm/h and an upper limit of 20.0mm/h were imposed for
continuing loss i.e. rainfall streamflow events which estimated continuing loss values
outside of this range (0.0mm/h–20.0mm/h) were excluded from this analysis. Less
than 1.0% of total events based on initial loss and less than 1.5% of total events
based on continuing loss were excluded as the range of the loss values were
considered. This chapter describes the fixed losses used in design flood estimation,
followed by a discussion on stochastic losses. It also covers regionalization of losses.
Finally the validation of results is described in this chapter.
8.2 Fixed losses
The losses where only a single loss value for a catchment either the median or mean
value was used are described as fixed losses in this analysis. The derived storm
losses were compared with ARR (I. E. Aust., 1998) recommended loss values and
the interaction of these derived losses were observed with other loss variables such as
how continuing loss should be estimated, surface runoff threshold values and
durations of the storm. This section compares the derived loss values obtained from
the 48 selected catchments with ARR (I. E. Aust., 1998) recommended losses and
also discusses about how the derived loss values vary with the change of loss
variables.
Modelling losses in flood estimation 8-2
ARR (I. E. Aust., 1998) is a design guide for flood estimation. The design loss values
are recommended in ARR (I. E. Aust., 1998), which is used by the designers for
design flood estimation. In ARR (I. E. Aust., 1998) there is no recommended
proportional loss and volumetric runoff co-efficient values. Hence the results of the
derived proportional loss and volumetric runoff co-efficient were not compared.
8.2.1 Comparison was made between derived computed losses and ARR (I. E.
Aust., 1998) recommended losses
The descriptive statistics of all the selected 969 rainfall streamflow events of IL and
CL values are shown in Table 8.1. Out of all the 48 selected catchments in Table 8.1,
the catchments which have the streamflow towards the east of Queensland were
considered as eastern catchments. A total of 25 catchments were considered as
eastern catchments. For the eastern catchments the IL values range was 0.0mm –
189.37mm and the CL values range was 0.01mm/h – 18.31mm/h and the median IL
values range was from 4.84mm – 71.77mm and the median CL values range was
from 0.71mm/h – 4.74mm/h respectively. The derived IL values from the study
catchments were much higher than the recommended IL values in ARR (Table 3.6,
Book Two, P.48, I. E. Aust., 1998 for eastern Queensland). The median IL values
from the selected 627 rainfall events of eastern catchments showed a range of
4.84mm – 71.77mm (average range of the median value: 37.98 mm) was compared
with the median IL values in ARR (I. E. Aust., 1998) for eastern catchments, which
was 15.0mm – 35.0mm (average of the range of median value: 25 mm). The details
of all the results of 969 rainfall streamflow events of 48 selected Queensland
catchments with a threshold value 0.01mm/h are presented in Appendix E.
Modelling losses in flood estimation 8-3
Table 8.1 Descriptive statistics of the computed IL and CL values (N = number of events)
Catchment ID Name Area (km2) N Storm initial losses (IL ), mm Storm continuing losses (CL ) mm/hRange Median Range Median
102101 Pascoe River 635 93 0.2 - 174.88 37.61 0.05 - 13.67 1.91104001 Stewart River 480 7 3.89 - 78.3 42.17 0.9 - 6.59 1.42105105 E. Norman. R. 300 3 10.92 - 18.39 11.92 0.92 - 1.43 1.3107001 Endeavour R. 310 3 16.35 - 114.03 71.77 0.57 - 3.07 0.71107003 Anna River 247 3 12.24 - 36.46 14 0.94 - 2.63 1.49112003 N. Johnston R. 173 15 3.3 - 108.55 34.04 0.3 - 7.79 2.69112101 S. Johnston R. 400 3 31.52 - 112.72 41.66 2.68 - 4.48 3.34114001 Murray River 155 23 1.6 - 159.22 65.75 0.05 - 8.44 4.74116008 Gowrie River 124 61 0.29 - 155.01 21.74 0.01 - 10.29 2.63116015 Blunder Creek 127 48 1.92 - 189.37 70.53 0.07 - 11.3 1.46116017 Stone River 157 55 0.26 - 161.71 33.23 0.09 - 14.52 2.54118003 Bohle River 143 24 0.11 - 93.2 28.8 0.66 - 7.63 2.26119006 Major Creek 468 4 10.27 - 79.87 35.35 0.33 - 1.20 1.15120014 Broughton R. 182 19 2.0 - 71.0 18.42 0.16 - 8.39 2.06120216 Broken River 78 11 29.35 - 123.37 64.26 0.56 - 9.11 1.7124002 St. Helens Ck. 129 11 11.56 - 154.62 53.71 0.33 - 6.04 1.62125005 Blacks Creek 505 35 0.8 - 144.39 57.63 0.22 - 15.39 3.15130207 Sande Creek 409 14 3.84 - 97.04 27.74 0.18 - 8.99 2.68136108 Monal Creek 92 12 2.71 - 48.2 13.08 0.18 - 9.12 1.21137101 Gregory River 454 8 3.57 - 123.05 29.81 0.12 - 5.74 2.235138110 Mary River 486 23 0.6 - 126.09 29.95 0.1 - 4.01 1.02141009 N. Maroochy R. 38 22 1.52 - 113.26 42.27 0.16 - 3.71 0.89143110 Bremer River 125 37 0.24 - 116.98 39.04 0.02 - 12.55 1.17143212 Tenhill Creek 447 24 6.86 - 125.46 43.48 0.01 - 7.58 1.16145003 Logan River 175 42 0.2 - 99.01 30.82 0.07 - 18.31 1.46145010 Running Creek 128 20 0.0 - 80.57 31.86 0.01 - 10.17 1.18145011 Teviot Brook 83 37 1.5 - 91.9 29.7 0.01 - 6.99 1145101 Albert River 169 35 0.59 - 165.84 43.46 0.1 - 6.95 1.52146014 Back Creek 7 10 0.0 - 49.55 4.84 0.52 - 2.92 1.87416410 Macintyre Brk. 465 28 0.05 - 93.34 28.73 0.15 - 15.06 1.77422321 Spring Creek 35 6 0.24 - 40.41 4.29 0.05 - 1.76 0.73422338 Canal Creek 395 27 0.07 - 116.41 24.24 0.13 - 6.84 1.58422394 Cadamine R. 325 21 8.99 - 89.35 40.51 0.08 - 3.2 0.92913005 Paroo Creek 305 6 1.06 - 25.64 9.31 0.64 - 5.58 2.38913009 Gorge Creek 248 9 0.03 - 46.23 6.2 0.17 - 5.39 1.06915205 Malbon River 425 5 9.17 - 59.79 34.21 0.56 - 14.61 3.95916002 Norman River 285 9 0.66 - 102.63 16.61 0.54 - 5.63 3.2916003 Moonlight Ck. 127 7 0.51 - 60.58 28.93 0.45 - 10.4 2.4917005 Agate Creek 228 19 0.14 - 34.67 13.9 0.23 - 7.12 2.83917007 Percy River 445 8 0.11 - 24.85 24.49 0.04 - 5.02 1.92917107 Elizabeth Ck. 585 8 2.74 - 42.09 27.25 0.25 - 4.03 2.04917114 Routh Creek 81 7 6.57 - 61.03 29.55 0.67 - 4.06 1.44919201 Palmer River 530 5 1.82 - 55.86 38.31 0.08 - 8.62 2.2919205 N. Palmer R. 430 7 0.8 - 46.19 14.51 0.3 - 10.95 5.8921001 Holroyd River 365 16 1.91 - 90.31 39.29 0.26 - 16.04 1.19922101 Coen River 166 59 0.26 - 81.89 24.52 0.08 - 9.45 2.16926002 Dulhunty River 325 12 0.0 - 16.29 3.39 0.03 - 5.91 1.6926003 Bertie Creek 130 8 0.0 - 5.03 0.7 0.18 - 6.85 1.64Average 20 0.0 - 189.37 0.7 - 71.77 0.01 - 18.31 0.71 - 5.8
Modelling losses in flood estimation 8-4
In this research the median initial loss values of all the selected 48 catchments were
computed as shown in Table 8.1. ARR (I. E. Aust., 1998) presented only the range of
median initial loss values, not all median initial loss values for the catchments. Hence
the average of the range of median loss values was compared in this research, not the
median value of the median initial losses of Queensland catchments. Table 8.2 shows
that the derived median IL which was computed from the 25 selected eastern
catchments was 52.0% higher than the ARR (I. E. Aust., 1998) recommended
median initial loss value. The use of smaller initial losses recommended in ARR (I.
E. Aust., 1998) is likely to result in significant overestimation of design floods. The
computed median CL value for the 25 eastern catchments was 1.52 mm/h, which was
compared with ARR (I. E. Aust., 1998) recommended eastern catchments median CL
value (2.5 mm/h). From Table 8.2 it is observed that the computed median CL value
for eastern catchments was 39.2% less than that of ARR (I. E. Aust., 1998)
recommended median continuing loss value. The low continuing loss values for
Queensland catchments was not because of the high initial loss values for the
catchments, but due to longer storm durations and less significant rainfall intensities.
In ARR (I. E. Aust., 1998) the median initial loss is shown by the upper and lower
limit, but the median continuing loss is illustrated by a fixed value. The derived
single median continuing loss value from this study was compared with ARR (I. E.
Aust., 1998) recommended single median continuing loss value.
Table 8.2 Comparison of computed median initial and median continuing loss values
with ARR (I. E. Aust., 1998) recommended median loss values for eastern
Queensland catchments
Typesof losses
Medianlosses
Average of the range of median losses
% Variation for loss values
Computed IL value 4.84 mm - 71.77 mm 37.98 mm 52.0
ARR (I. E. Aust., 1998) recommended IL value 15.0 mm - 35.0 mm 25.00 mm
Computed CL value 1.52 mm/h -39.2
ARR (I. E. Aust., 1998) recommended CL value 2.50 mm/h
Modelling losses in flood estimation 8-5
The derived initial loss values obtained from this research work was compared to the
computed initial loss values for Victorian catchments (Rahman et al., 2000). It was
observed that the losses in Queensland exhibit a much greater variability, as shown
in Table 8.3. The lower limit of initial loss for both the Queensland and Victorian
catchments are 0.0mm. The upper limit of initial loss and median initial loss values
for Queensland catchments are respectively 32.43% and 29.39% greater than that of
Victorian catchments. The possible reason for higher loss values and greater
variability in Queensland as compared to Victoria may be due to higher evaporation
rate in Queensland and variation in soil characteristics and yearly rainfall.
Table 8.3 Comparison of initial loss values between Queensland and Victorian
catchments
Initial loss Queensland (mm) Victoria (mm)Variation
(% higher for Qld)
Lower limit 0.00 0.00
Upper limit 189.37 143.00 32.43
Median 29.76 23.00 29.39 In ARR (I. E. Aust., 1998) the whole Queensland catchments is divided into eastern
and western catchments. In this research there were 5 catchments, which were
considered as the western catchments. Other catchments were considered as northern
catchments, though in ARR (I. E. Aust., 1998) there are no loss values recommended
for northern catchments. Table 8.4 shows that the average of the derived median IL
which was computed from the 5 selected western catchments was 44.8mm and the
ARR (I. E. Aust., 1998) recommended median initial loss value is 0.0mm. The use of
smaller initial losses recommended in ARR (I. E. Aust., 1998) is likely to result in
significant overestimation of design floods. The computed median CL value for the 5
western catchments was 1.58 mm/h, which was compared with ARR (I. E. Aust.,
1998) recommended western catchments median CL value (1.4 mm/h). Table 8.4
shows that the computed median CL value for western catchments was 12.86%
higher than that of ARR (I. E. Aust., 1998) recommended median continuing loss
value. For the Western catchments the IL value range was from 0.07mm –
Modelling losses in flood estimation 8-6
116.41mm, but in ARR (I. E. Aust., 1998) the IL value range is from 0.0mm-
80.0mm. The reason of computed IL and CL values of this research were well
different from recommended loss values, because in ARR (I. E. Aust., 1998) the
recommended CL values used were derived from a very few Queensland catchments
and also recommended IL values used were derived using the NSW catchments (not
Queensland catchments). The detail of this existing recommended loss values in
ARR (I. E. Aust., 1998) is discussed in section 8.4.2.
Table 8.4 Comparison of computed median initial and median continuing loss values
with ARR (I. E. Aust., 1998) recommended median loss values for
western Queensland catchments
Typesof losses Loss range
Medianlosses
Average of the rangeof median losses
% Variation for loss values
Computed IL value 0.07 mm - 116.41 mm 4.29 mm - 40.51 mm 44.8 mmARR (I. E. Aust., 1998) recommended IL value 0.0 mm - 80.0 mm 0.0mm 0.0 mm
Computed CL value 1.58 mm/h 12.86ARR (I. E. Aust., 1998) recommended CL value 1.4 mm/h
8.2.2 Interactions of losses with other variables
8.2.2.1 Different methods of computing continuing losses and their comparison
In ARR (I. E. Aust., 1998) it is mentioned that after satisfying the IL, the CL in a
rainfall event occurs until the end of the storm. In other words the recommended
value for the continuing loss is computed up to the end of the remainder of the storm.
The assumption of the estimation of continuing loss in ARR (I. E. Aust., 1998) is
discussed in equation 6.3 (Chapter 6). This assumption is applicable where the
design storm is used to compute the design loss values.
This assumption is not accurate for actual rainfall events, because in actual rainfall
events continuing loss occur up to the end of the surface runoff. In this research
actual rainfall and actual streamflow data were used to compute the loss values.
Modelling losses in flood estimation 8-7
Hence it was possible to compute the duration of surface runoff from the surface
runoff hydrograph. The continuing loss occurred up to the end of surface runoff was
computed by the equation 8.1 and was compared with the computed CL values (CL
values up to the end of the storm) of this research work. As IL-CL model does not
consider the temporal variability of losses, to compute continuing loss the following
equation 8.1 was adopted in this research.
CL1 = (R – IL – QF)/ t1 (8.1)
Where,
R = Total rainfall of the event expressed in average depth of the rainfall over the
catchment in mm
IL = Initial loss in mm
QF = Quickflow, total streamflow minus baseflow in mm
CL1 = Continuing loss occurred within the time t1 (h)
t1 = Time elapsed between the start of the surface runoff and end of the surface
runoff (hour)
All the derived CL and CL1 values obtained from this research are shown in Table
8.5. Also some important statistics of the derived CL and CL1 values from the study
catchments are summarized in Table 8.5. The range of CL and CL1 values were
found 0.01mm/h – 18.31mm/h and 0.0mm/h – 10.40mm/h respectively and the
median values for CL and CL1 were 1.67mm/h and 1.49mm/h respectively.
In most of the rainfall events the surface runoff continues even after the end of the
storm. A few events in some catchments like stream gauging station numbers
116015A, 124002A, 141009A, 146014A, 422321B and 913009A the duration (t) is
greater than the duration (t1). It means that the surface runoff was ceased prior to the
completion of the rainfall event. This usually happens when there is a long duration
rainfall event with a very little rainfall intensity at the end of the storm. For those
rainfall events the derived median CL value was found less than the derived median
CL1 value.
Modelling losses in flood estimation 8-8
A few events in 6 catchments out of 48 selected catchments (which is 12.5%), it was
observed that continuing loss does not continue upto the end of the storm as the
surface runoff was ceased prior to the completion of the rainfall event. Again most of
the events of 4 catchments out of 48 selected catchments (which is 8.33 %), it was
observed that these events do not satisfy the requirements of IL – CL model to
compute the losses as the surface runoff started after the end of the rainfall event.
Hence continuing loss does not occur upto the end of the storm.
In this research design continuing loss rate (CL) and actual continuing loss rate (CL1)
both were computed from observed rainfall event. The continuing loss rate CL1 can
be used as recommended continuing loss rate for real time flood estimation.
Modelling losses in flood estimation 8-9
Table 8.5 Descriptive statistics of the computed CL and CL1 values (N = number of events) Catchment Name Area (km2) N Storm continuing losses (CL ) mm/h Storm continuing losses (CL 1 )mm/h
Range Median Range Median102101 Pascoe River 635 93 0.05 - 13.67 1.91 0.06 - 9.15 1.68104001 Stewart River 480 7 0.9 - 6.59 1.42 0.45 - 3.76 1.07105105 E. Norman. R. 300 3 0.92 - 1.43 1.3 0.92 - 1.43 1.3107001 Endeavour R. 310 3 0.57 - 3.07 0.71 0.57 - 3.07 0.71107003 Anna River 247 3 0.94 - 2.63 1.49 0.94 - 2.63 1.49112003 N. Johnston R. 173 15 0.3 - 7.79 2.69 0.1 - 7.79 2.33112101 S. Johnston R. 400 3 2.68 - 4.48 3.34 2.49 - 3.20 2.63114001 Murray River 155 23 0.05 - 8.44 4.74 0.04 - 7.3 3.81116008 Gowrie River 124 61 0.01 - 10.29 2.63 0.01 - 9.33 2.41116015 Blunder Creek 127 48 0.07 - 11.3 1.46 0.05 - 8.24 1.58116017 Stone River 157 55 0.09 - 14.52 2.54 0.08 - 7.3 2.14118003 Bohle River 143 24 0.66 - 7.63 2.26 0.66 - 4.16 2.08119006 Major Creek 468 4 0.33 - 1.20 1.15 0.21 - 1.06 0.64120014 Broughton R. 182 19 0.16 - 8.39 2.06 0.04 - 7.52 1.5120216 Broken River 78 11 0.56 - 9.11 1.7 0.26 - 9.11 1.49124002 St. Helens Ck. 129 11 0.33 - 6.04 1.62 0.23 - 6.02 2.32125005 Blacks Creek 505 35 0.22 - 15.39 3.15 0.16 - 13.7 3.08130207 Sande Creek 409 14 0.18 - 8.99 2.68 0.02 - 6.36 1.95136108 Monal Creek 92 12 0.18 - 9.12 1.21 0.03 - 7.82 0.85137101 Gregory River 454 8 0.12 - 28.74 2.56 0.12 - 3.59 1.84138110 Mary River 486 23 0.1 - 4.01 1.02 0.07 - 3.72 0.83141009 N. Maroochy R. 38 22 0.16 - 3.71 0.89 0.16 - 2.43 0.94143110 Bremer River 125 37 0.02 - 12.55 1.17 0.01 - 3.81 0.44143212 Tenhill Creek 447 24 0.01 - 7.58 1.16 0.01 - 7.58 0.75145003 Logan River 175 42 0.07 - 18.31 1.46 0.01 - 5.18 0.43145010 Running Creek 128 20 0.01 - 10.17 1.18 0.0 - 6.73 0.63145011 Teviot Brook 83 37 0.01 - 6.99 1 0.01 - 3.93 0.8145101 Albert River 169 35 0.1 - 6.95 1.52 0.06 - 6.3 0.99146014 Back Creek 7 10 0.52 - 2.92 1.87 0.3 - 2.92 2.16416410 Macintyre Brk. 465 28 0.15 - 15.06 1.77 0.09 - 8.03 1.22422321 Spring Creek 35 6 0.05 - 1.76 0.73 0.02 - 1.76 0.78422338 Canal Creek 395 27 0.13 - 6.84 1.58 0.07 - 5 57 0.81422394 Cadamine R. 325 21 0.08 - 3.2 0.92 0.04 - 1.93 0.52913005 Paroo Creek 305 6 0.64 - 5.58 2.38 0.43 - 3.28 1.72913009 Gorge Creek 248 9 0.17 - 5.39 1.06 0.16 - 2.53 1.42915205 Malbon River 425 5 0.56 - 14.61 3.95 0.36 - 3.76 1.93916002 Norman River 285 9 0.54 - 5.63 3.2 0.25 - 4.03 2.18916003 Moonlight Ck. 127 7 0.45 - 10.4 2.4 0.3 - 10.4 2.32917005 Agate Creek 228 19 0.23 - 7.12 2.83 0.23 - 4.9 2.05917007 Percy River 445 8 0.04 - 5.02 1.92 0.03 - 4.96 1.81917107 Elizabeth Ck. 585 8 0.25 - 4.03 2.04 0.25 - 2.93 1.42917114 Routh Creek 81 7 0.67 - 4.06 1.44 0.21 - 3.69 0.94919201 Palmer River 530 5 0.08 - 8.62 2.2 0.08 - 7.2 2.16919205 N. Palmer R. 430 7 0.3 - 10.95 5.8 0.19 - 4.38 3.4921001 Holroyd River 365 16 0.26 - 16.04 1.19 0.21 - 6.42 1.11922101 Coen River 166 59 0.08 - 9.45 2.16 0.03 - 7.46 1.58926002 Dulhunty River 325 12 0.03 - 5.91 1.6 0.01 - 5.6 0.71926003 Bertie Creek 130 8 0.18 - 6.85 1.64 0.18 - 2.69 1.18Average 20 0.01 - 18.31 1.67 0.0 - 10.40 1.49
Modelling losses in flood estimation 8-9
For some catchments only 3-5 rainfall streamflow events were selected to compute
loss values (Basin ID no: 112101b, 137101a, 119006a and 915205a), as in those
catchments most of the events do not satisfy the requirements of IL-CL model to
compute the losses. For example, rainfall events where there was no surface runoff
occurred throughout the duration of the storm, such rainfall events were not
considered for the analysis. When the duration of the storm continues even after the
completion of the surface runoff and the rainfall that occurs after the completion of
the surface runoff did not show any impact in the streamflow, then to compute
continuing losses the duration of the remainder of the storm (t) and the duration up to
the end of the surface runoff (t1) were considered equal.
In the case of consecutive rainfall events, separated by a short time interval, if the
surface runoff generated by an individual event could not be distinguished explicitly
and hence these types of events were excluded from the analysis. Events where
surface runoff started after the end of the rainfall event were not considered in this
research work, as this kind of rainfall events do not satisfy the requirements of IL-
CL model.
To examine the variation of continuing loss due to difference of time t and t1 a
research was performed on the results of all the selected 969 rainfall events shown in
Table 8.6. The computed median CL value for time t which is 1.67 mm/h is 12.1%
higher than the computed median CL1 value for time t1 which is 1.49 mm/h. Table
8.6 also shows that by comparing the range of derived CL and CL1 values and by
taking the average of the range of both the loss values, CL value is 76.15% higher
than the CL1 value. The 25 catchments from Table 8.5, which is the eastern
Queensland catchments showed that the derived median CL value was 1.52mm/h and
the derived median CL1 value was 1.49mm/h. This result indicates that although the
initial losses for the study catchments were relatively higher than the ARR (I. E.
Aust., 1998) recommended initial losses (Table 8.2), but the derived CL and CL1
values were relatively smaller than the ARR (I. E. Aust., 1998) recommended
continuing losses. This indicates that once the catchments surface get saturated or
near saturated, the subsequent loss rate decreases. This is likely to produce relatively
greater volume of runoff particularly, at the latter part of the rainfall event.
Modelling losses in flood estimation 8-10
Table 8.6 Comparison between computed CL and CL1 values for all the selected
catchments
Typesof losses
Continuing loss range (mm/h)
Average of the range of continuing loss value (mm/h)
Variation(% higher for CL loss values)
Median continuingloss (mm/h)
Variation(% higher for CL loss values)
Computed CL 0.01 - 18.31 9.16 76.15 1.67 12.1
Computed CL 1 0.0 - 10.40 5.2 1.49
8.2.2.2 Comparison of derived loss values with different surface runoff
threshold values
The surface runoff threshold value is the minimum amount of water required above
the catchment to start the surface runoff. It is usually found that for a specific rainfall
event if the surface runoff threshold value increases, the initial loss also increases.
Hence it is important to select a more appropriate surface runoff threshold value to
compute the storm losses. In this research the surface runoff threshold value was
used as 0.01mm/h. To examine how storm loss values vary with the change of
surface runoff threshold values in loss computation, three different surface runoff
threshold values were used in this research work. These threshold values were
0.005mm/h, 0.01mm/h and 0.02mm/h respectively. Table 8.7 shows the change of
loss values of median IL, CL and CL1 with the change of surface runoff threshold
values.
Modelling losses in flood estimation 8-11
Table 8.7 Comparison of computed IL, CL and CL1 values with different surface runoff threshold values Catchment Name Area (km2) Events Surface runoff threshold = 0.005 mm/h Surface runoff threshold = 0.01mm/h Surface runoff threshold = 0.02 mm/h
No median IL median CL median CL 1 median IL median CL median CL 1 median IL median CL median CL 1
102101 Pascoe River 635 93 30.75 1.83 2.415 37.61 1.91 1.68 39.29 1.78 2.15104001 Stewart River 480 7 38.52 1.64 1.435 42.17 1.42 1.07 42.17 1.42 0.86105105 E. Norman. R. 300 3 11.92 1.3 1.3 11.92 1.3 1.3 11.92 1.3 1.3107001 Endeavour R. 310 3 71.77 0.71 5.21 71.77 0.71 0.71 53.68 0.705 5.28107003 Anna River 247 3 14 1.49 1.49 14 1.49 1.49 72 1.49 1.49112003 N. Johnston R. 173 15 22.19 3.005 2.47 34.04 2.69 2.33 47.275 3.325 2.635112101 S. Johnston R. 400 3 21.93 2.76 2.47 41.66 3.34 2.63 49.72 3.7 2.9114001 Murray River 155 23 38.14 2.91 3.32 65.75 4.74 3.81 64.805 2.445 3.46116008 Gowrie River 124 61 14.405 2.56 2.37 21.74 2.63 2.41 24.54 2.61 2.46116015 Blunder Creek 127 48 61.46 1.36 1.7 70.53 1.455 1.58 68.815 1.63 1.89116017 Stone River 157 55 31.21 2.49 2.4 33.23 2.54 2.14 32.51 2.63 2.37118003 Bohle River 143 24 27.685 2.355 1.925 28.8 2.255 2.08 37.71 2.17 1.86119006 Major Creek 468 4 35.345 1.18 1.03 35.345 1.145 0.64 47.865 0.995 0.525120014 Broughton R. 182 19 15.67 1.78 1.5 18.42 2.06 1.5 19.83 1.925 1.53120216 Broken River 78 11 40.4 1.08 0.93 64.26 1.7 1.49 52.33 2.015 1.655124002 St. Helens Ck. 129 11 47.11 2.525 2.15 53.71 1.62 2.32 57.895 2.485 3.31125005 Blacks Creek 505 35 36.7 3.08 2.78 57.63 3.15 3.08 48.2 3 3.92130207 Sande Creek 409 14 22.815 2.61 1.79 27.74 2.675 1.945 33.97 2.56 1.99136108 Monal Creek 92 12 11.31 1.21 0.85 13.075 1.205 0.85 25.66 1.2 0.85137101 Gregory River 454 8 31.22 2.235 1.575 29.805 2.56 1.84 37.51 2.45 1.04138110 Mary River 486 23 31.29 1.335 0.945 29.95 1.02 0.83 36.62 1.13 0.85141009 N. Maroochy R. 38 22 14.155 1.035 0.93 42.27 0.89 0.935 42.45 0.71 0.9143110 Bremer River 125 37 35.82 1.19 0.5 39.04 1.17 0.44 37.885 1.18 0.545143212 Tenhill Creek 447 24 39.19 1.26 1.14 43.475 1.155 0.75 42.21 1.23 0.73145003 Logan River 175 42 25.46 1.81 0.46 30.815 1.455 0.425 30.535 1.28 0.425145010 Running Creek 128 20 31.72 1.18 0.52 31.86 1.18 0.63 31.735 1.115 0.645145011 Teviot Brook 83 37 27.07 1.32 0.86 29.695 1 0.8 30.22 1.245 0.82145101 Albert River 169 35 36.89 1.5 0.86 43.46 1.515 0.99 43.46 1.41 1.03146014 Back Creek 7 10 4.84 1.56 2.16 4.84 1.87 2.16 4.84 1.64 2.6416410 Macintyre Brk. 465 28 28.73 1.49 1.125 28.73 1.765 1.22 28.91 1.55 1.19422321 Spring Creek 35 6 4.29 0.53 0.78 4.29 0.73 0.78 35.01 0.54 0.92422338 Canal Creek 395 27 26.955 1.345 0.88 24.24 1.58 0.81 25.73 1.5 0.83422394 Cadamine R. 325 21 38.83 0.74 0.52 40.51 0.92 0.52 33.67 0.9 0.52913005 Paroo Creek 305 6 4.455 2.295 1.205 1.84 2.38 1.715 2.23 1.51 1.19913009 Gorge Creek 248 9 11.155 1.075 0.615 6.2 1.06 1.42 9.49 1.31 0.65915205 Malbon River 425 5 34.21 1.24 1.24 46.69 3.945 1.925 15.65 2.33 1.33916002 Norman River 285 9 16.61 2.99 1.65 16.61 3.2 2.18 16.315 3.285 1.98916003 Moonlight Ck. 127 7 24.695 3.85 2.465 28.93 2.4 2.32 32.525 1.555 0.655917005 Agate Creek 228 19 14.205 3.06 1.995 13.9 2.83 2.05 18.36 3.03 1.995917007 Percy River 445 8 20.38 2.2 1.81 24.49 1.92 1.81 26.04 1.94 1.81917107 Elizabeth Ck. 585 8 23.47 2.14 1.24 27.25 2.035 1.415 31.55 1.92 1.43917114 Routh Creek 81 7 26.04 3.83 1.08 29.55 1.44 0.94 33.86 1.3 0.64919201 Palmer River 530 5 31.29 2.305 2.16 38.31 2.2 2.16 45.69 2.16 1.185919205 N. Palmer R. 430 7 14.51 6.97 3.4 14.51 5.8 3.4 11.685 7.15 3.83921001 Holroyd River 365 16 42.01 1.28 1.06 39.29 1.185 1.11 46.885 1.13 0.625922101 Coen River 166 59 14.78 2.23 1.62 24.52 2.16 1.58 25.175 2.185 1.475926002 Dulhunty River 325 12 0.72 1.59 0.7 1.64 1.6 0.705 4.145 2.3 1.95926003 Bertie Creek 130 8 0.31 1.44 1.71 0.7 1.635 1.175 4.58 2.3 2.3Average 20 26.5 1.62 1.46 29.75 1.67 1.49 33.77 1.64 1.38
Modelling losses in flood estimation 8-12
Table 8.8 shows the comparison of median loss values with different threshold
values. Table 8.8 also shows that, when the threshold value was 0.005mm/h the
median initial loss was 26.5mm. When the threshold value for surface runoff of all
the selected events was increased to 0.01mm/h, the median initial loss was 29.75mm
i.e. the median initial loss values increased 12.3%. When the threshold value was
increased to 0.02mm/h from 0.005mm/h, the median initial loss was 33.77mm i.e.
the median initial loss values increased 27.4%. The median CL value was 1.62mm/h,
when the surface runoff threshold value was 0.005mm/h. By increasing the threshold
value from 0.005mm/h to 0.01mm/h, the median CL value was 1.67mm/h i.e. the
median continuing loss increased 3.1%. Again by increasing the threshold value from
0.005mm/h to 0.02mm/h, the median CL value was increased to 1.64mm/h i.e. the
median continuing loss increased 1.2%. Similarly the median CL1 value was
1.46mm/h, when the surface runoff threshold value was 0.005mm/h. By increasing
the threshold value from 0.005mm/h to 0.01mm/h, the median CL1 value was
1.49mm/h i.e. the CL1 value was increased to 2.1%. Again by increasing the
threshold value from 0.005mm/h to 0.02mm/h, the median CL1 value was decreased
to 1.38mm/h i.e. the CL1 value was decreased to 5.5%.
Care should be taken before the selection of the threshold value, as the threshold
value depends on the rainfall input and the size of the catchment. For larger
catchments the smaller threshold value would compute more accurate loss values.
This is because for larger catchments a small increase in threshold value would
convert a big volume of surface runoff into initial loss. For Victorian catchments Hill
et al., (1996a) used a surface runoff threshold value of 0.01mm/h.
From this result it is observed that the proper selection of a threshold value is
important for the computation of storm loss values. The result also indicates that the
median initial loss of a rainfall event is quite sensitive to the selection of a surface
runoff threshold value. Again, the derived median CL and CL1 values were not very
sensitive to the selection of a surface runoff threshold value.
Modelling losses in flood estimation 8-13
Table 8.8 Comparison of loss values with different threshold values
Loss valuesThreshold
values (mm/h)Medianvalues
% variationof losses for different
threshold values
0.005 26.5
IL (mm) 0.01 29.75 12.3
0.02 33.77 27.4
0.005 1.62
CL (mm/h) 0.01 1.67 3.1
0.02 1.64 1.2
0.005 1.46
CL 1 (mm/h) 0.01 1.49 2.1
0.02 1.38 -5.5
Modelling losses in flood estimation 8-14
8.2.2.3 Variation of continuing losses with duration
In ARR (I. E. Aust., 1998) the continuing loss is assumed to occur up to the end of
the rainfall event. To compare the effect of duration on continuing losses equation
6.3 was used:
CL = (R – IL – QF)/ t
Where,
R = Total rainfall of the event expressed in average depth of rainfall in mm over the
catchment.
IL = Initial loss in mm
QF = Quickflow, assumed to result from the rainfall event, expressed in mm.
t = Time elapsed between the start of the surface runoff and end of the rainfall event
(hour)
To examine the effect of duration on continuing loss an analysis was performed on
all the selected 969 rainfall events of this research using threshold value 0.01mm/h.
In ARR (I. E. Aust., 1998) it is recommended that for any rainfall event, the
continuing loss is a constant after satisfying the required initial loss. To examine how
continuing loss varies with the duration, the continuing losses of all the selected 969
rainfall events were plotted against their duration (duration between the end of initial
loss and the end of the rainfall event) of all the events as shown in Figure 8.1. The
continuing loss for each catchment was examined against their durations of the
remainder of the storm. It was observed that the continuing loss decays with duration
i.e. it was not a single fixed value as recommended in ARR (I. E. Aust., 1998). The
result of the continuing loss values for each catchment is shown in Appendix F. The
catchments with more than 12 rainfall events were considered for this investigation.
In Queensland the loss value varies with the location of the catchments. To examine
the effect of duration in loss values for different regions of Queensland, the
Queensland catchments were divided into two categories to compute storm losses
such as eastern catchments and western catchments. The initial losses in western
catchments are sometimes higher because the catchments are dryer than the eastern
Modelling losses in flood estimation 8-15
catchments. An investigation was executed to examine the effect of duration on
continuing losses for different locations of Queensland catchments. Out of all
selected 48 Queensland catchments 11 eastern catchments, 5 western catchments and
12 northern Queensland catchments were selected to examine the effect of duration
on continuing loss values.
y = -0.5454Ln(h) + 4.0791
0
5
10
15
20
25
30
0.1 1 10 100 1000Duration (h)
CL
(mm
/h)
Figure 8.1 Variation of continuing loss values with duration in all 48 selected
Queensland catchments
To examine how continuing loss varies with duration, the continuing loss and the
duration of 270 rainfall events of 11 eastern Queensland catchments were plotted as
shown in Figure 8.2. It shows that, the continuing loss is not constant with storm
duration but rather it decays with the duration. The equation of the decaying curve is
shown in Figure 8.2.
Modelling losses in flood estimation 8-16
y = -0.5873Ln(h) + 3.4464
0
5
10
15
20
25
30
35
1 10 100 1000Duration (h)
CL
(mm
/h)
Figure 8.2 Variation of continuing loss values with duration in 11 eastern
Queensland catchments.
In Figure 8.3 the continuing losses of 96 rainfall events of 5 western catchments in
Queensland are plotted against their respective durations to examine the effect of
duration on continuing losses. Figure 8.3 shows that the continuing loss is not
constant in respect of duration, but it decays with respect to duration of the rainfall
event. The equation of the decaying curve is also shown in Figure 8.3.
y = -0.5893Ln(h) + 3.6065
0
2
4
6
8
10
12
14
16
1 10 100 1000Duration (h)
CL
(mm
/h)
Figure 8.3 Variation of continuing loss values with duration in 5 western Queensland
catchments.
Modelling losses in flood estimation 8-17
In Figure 8.4 the continuing losses of 340 rainfall events of 12 northern catchments
of Queensland are plotted against their respective durations to examine the effect of
duration on continuing losses. Figure 8.4 shows that the continuing loss is not
constant in respect of duration, but that it decays with respect to duration of the
rainfall event. The equation of the decaying curve is shown in Figure 8.4.
y = -0.8883Ln(h) + 5.7003
0
2
4
6
8
10
12
14
16
1 10 100 1000Duration (h)
CL
(mm
/h)
Figure 8.4 Variation of continuing loss values with duration in 12 northern
Queensland catchments.
ARR (I. E. Aust., 1998) recommended that continuing loss rate is constant
throughout the duration of the storm. The derived continuing loss values of this
research were plotted in the diagram against their durations as shown in Figures 8.1,
8.2, 8.3 and 8.4. Also the result of the continuing loss against their duration for
individual catchments is shown in Appendix F. All these diagrams show that the
derived continuing loss value decreases with the increase in the duration of the
rainfall event i.e. CL value was not a fixed single value for a catchment as
recommended in ARR (I. E. Aust., 1998) but it decays with the increase in the
duration of the storm. Hence it was observed that the continuing loss of the
Queensland catchments can be described as probability distributed losses.
Modelling losses in flood estimation 8-18
8.3 Stochastic losses
8.3.1 Initial losses
Fixed losses are discussed in Section 8.2. Stochastic or probability distributed losses
is discussed in this section. In Queensland there is no data available on probability
distributed losses. It was investigated whether a theoretical distribution can be used
to describe the initial loss distributions. To undertake the analysis of probability
distributed losses, a total of 15 catchments ranging from 83 km2 to 486 km2 (average:
254 km2) were selected along the coastal region of Queensland as there were very
few western catchments that have long streamflow records to compute storm losses.
Out of these 15 catchments, 12 catchments were previously selected to compute the
fixed losses and 3 new catchments were randomly selected for this analysis. The
selected catchments were unregulated and rural and have reasonably long rainfall and
streamflow records. The locations of the 15 study catchments are shown in Figure
8.5.
Modelling losses in flood estimation 8-19
Figure 8.5 Locations of the 15 study catchments in Queensland
A total of 388 complete storm rainfall events were selected that have the potential to
produce significant surface runoff. For each complete storm, a storm-core was
identified, defined as the most intense rainfall burst within a complete storm.
The available IFD information in ARR (I. E. Aust., 1998) is not based on complete
storm but on periods of intense rainfall within complete storm, called burst. If this
existing information is to be used with the proposed new approach, it is more useful
to undertake the design rainfall analysis in terms of the storm bursts. However as the
duration of the bursts in the ARR (I. E. Aust., 1998) analysis was predetermined
rather than random, it is necessary to consider a new storm burst definition that will
produce randomly distributed storm burst durations. These newly defined storm
bursts are referred to as storm-cores, same as mentioned in Rahman et al. (1998).
The computed complete storm initial loss (ILs) and storm-core initial loss (ILc) values
of the selected catchments are shown in Table 8.9. Considering all the 388 rainfall
events, the mean ILs and ILc values were respectively 34.7mm and 29.4mm, that is
Modelling losses in flood estimation 8-20
mean ILc value is 18.0% smaller than mean ILs value. The median values of ILs and
ILc are respectively 28.7mm and 27.3mm. The standard deviations were 28.0mm and
19.8mm respectively, which were relatively high showing a high degree of
variability in computed loss values. Also the loss values show a positive skewness
with an average value of over 0.7. This shows that the losses in Queensland, in
general are much higher and having greater variability. Given the degree of
variability and wide range of ILs and ILc values for Queensland catchments, it
appears to be unreasonable to adopt a single representative value (either mean or
median) of losses for flood estimation, as adopted with the Design Event Approach.
Table 8.9 Loss statistics for 15 Queensland catchments (N = number of events, SD =
standard deviation)
Station ID N Complete storm loss (ILs ) Storm-core loss (ILc )
Range Mean Median SD Skew Range Mean Median SD Skew
117002 21 4.59-152.32 45.63 41.14 37.33 1.24 0.0-87.58 31.62 34.42 26.39 0.29
117003 14 0.11-154.01 43.24 30.77 46.02 1.37 0.0-154.01 41.86 38.98 35.23 0.96
120014 19 2.0-71.41 24.86 18.42 21.15 1.24 0.0-71.41 22.55 15.42 19.63 1.27
136112 12 0.47-88.31 31.38 23.57 28.51 0.92 0.0-84.68 30.46 22.87 16.88 0.39
138110 22 0.6-95.19 38.15 29.95 32.12 1.4 0.0-95.19 31.64 32.05 21.11 0.29
143110 37 0.24-116.98 42.23 39.04 28.73 0.81 0.0-109.75 17.81 38.35 22.01 0.96
145003 42 0.2-99.01 33.97 30.82 24.44 0.71 0.0-98.72 33.31 29.76 17.3 1.38
145011 37 1.5-91.90 30.48 29.7 21.19 0.98 0.0-91.90 28.74 27.15 18.22 0.81
145101 35 0.59-165.84 53.04 43.46 46.78 0.97 0.58-156.39 45.08 39.24 29.44 1.4
416410 34 0.31-93.34 35.77 28.73 26.61 0.69 2.215-88.38 33.93 28.14 16.33 1.07
422338 26 0.07-116.41 35.28 25.73 31 1.35 17.16-94.45 34.71 25.49 16.06 1.2
422394 21 8.99-89.35 47.3 40.51 26.62 0.38 8.99-85.56 34.88 32.13 16.27 1.21
913009 6 0.03-31.86 12.99 7.57 14.44 0.74 3.24-31.86 9.8 5.36 11.35 0.06
916002 8 0.66-46.37 20.48 16.32 15.93 0.47 0.0-43.18 20.24 15.09 13.8 -0.75
922101 53 0.26-97.81 25.6 24.86 19.82 0.58 0.0-83.41 24.9 24.66 17.51 0.14
Average 26 0.03-165.84 34.69 28.72 28.05 0.92 0.0-156.39 29.44 27.27 19.84 0.71
The individual histogram of ILs and ILc values of all the 15 catchments were
examined (examples are shown in Figures 8.6 and 8.7), and it is hypothesised that a
four-parameter Beta distribution can be used to describe the initial loss distributions
for Queensland catchments. The individual histogram of ILs and ILc values of most of
the 15 catchments are shown in Appendix G. A four-parameter Beta distribution is
Modelling losses in flood estimation 8-21
fitted to individual site’s ILs data using observed values of lower limit (LL), upper
limit (UL), mean and standard deviation of ILs data. The fitting of a theoretical
distribution to ILs data is considered only because ILc value was estimated from
simple relationship between ILs and ILc. The ILc was estimated from ILs following
the approach of Rahman et al. (2000) shown in equation 8.2.
ILc = ILs [0.5 + 0.25log10 (Dc)] (8.2)
Where
ILs = Complete storm initial loss
ILc = Storm-core initial loss
Dc = Storm-core duration
This relation gave ILc = ILs at Dc = 100 hour, and ILc = 0.50 × ILs at Dc =1 hour.
Figure 8.6 Histogram showing the distribution of ILs for catchment 143110.
01020304050607080
0 -20 21 -40 41 -60 61 -80 81 -100101 -120Initial loss (mm)
Freq
uenc
y
01020304050607080
0 -20 21 -40 41 -60 61 -80 81 -100101 -120Initial loss (mm)
Freq
uenc
y
01020304050607080
0 -20 21 -40 41 -60 61 -80 81 -100101 -120Initial loss (mm)
Freq
uenc
y
01020304050607080
0 -20 21 -40 41 -60 61 -80 81 -100101 -120Initial loss (mm)
Freq
uenc
y
Modelling losses in flood estimation 8-22
Figure 8.7 Histogram showing the distribution of ILc for catchment 143110.
For each catchment, a total of 10,000 values of ILs were generated from the fitted
four-parameter Beta distribution. The statistics of the observed and generated ILs data
were compared in Table 8.10, which shows that the generated data preserves the
statistics of the observed loss value very well with respect to mean value (variation is
in the range 0.17-3.9% with an average of 1.83%) but for Gorge Creek (Catchment
ID 913009) the variation in mean value was 30.3% on the basis of generated data,
which needs some more investigation. Table 8.10 also shows that the generated data
preserves the statistics of the observed loss value very well with respect to the
standard deviation (variation in the range 0-1.62% with an average of 0.23%). The
lower limits of both the generated and observed data varies in the range of 0.0-
2.73mm (which is close to 0.0mm) with an exception of two values 4.59mm for
Black River and 8.99mm for Cadamine River. In the case of the upper limit of both
the generated and observed data the variation is in the range of 0.0-0.80% with an
average of 0.19%. Table 8.10 shows that the lower limit and upper limit of the
generated data preserves the statistics of the observed loss value very well.
From the above analysis it was found that the initial losses in Queensland can be
approximated by a four-parameter Beta distribution. The generated initial losses from
the fitted Beta distribution preserve the lower limit, upper limit, mean and standard
0102030405060708090
0 - 20 21 - 40 41 - 60 61 - 80 81 - 100 101 - 120
Initial loss (mm)
Freq
uenc
y
0102030405060708090
0 - 20 21 - 40 41 - 60 61 - 80 81 - 100 101 - 120
Initial loss (mm)
Freq
uenc
y
Modelling losses in flood estimation 8-23
deviation of the observed losses very well of the observed ILs data. Thus, four-
parameter Beta distribution can be used to approximate ILs distributions in the
selected 15 catchments. The generated ILs data from the fitted Beta distribution can
be used to obtain derived flood frequency curves in the Queensland catchments using
Joint Probability Approach (Monte Carlo Simulation) (Rahman et al., 2001a,b).
Table 8.10 Comparison at-site observed and generated ILs data for the 15 catchments Catchment ID Name of the stream Area (km2 ) LL(mm) UL(mm) Mean(mm) SD(mm)
117002 Black River 256 Observed 4.59 152.32 45.63 37.33Generated 4.00 152.09 44.73 37.30
117003 Bluewater Creek 86 Observed 0.11 154.01 43.24 46.02Generated 0.00 154.00 42.80 45.90
120014 Broughton River 182 Observed 2.73 73.53 24.86 21.15Generated 2.00 73.52 24.02 21.14
136112 Burnett River 370 Observed 0.47 88.31 31.38 28.51Generated 0.00 88.30 31.00 28.49
138110 Mary River 486 Observed 0.60 126.09 38.15 32.12Generated 0.01 125.90 37.80 32.12
143110 Bremer River 125 Observed 0.24 116.98 42.23 28.73Generated 0.01 116.04 41.80 28.63
145003 Logan River 175 Observed 0.20 99.01 33.97 24.44Generated 0.01 98.72 32.70 24.40
145011 Teviot Brook 83 Observed 1.50 91.90 30.48 21.19Generated 1.01 91.31 29.75 21.14
145101 Albert River 169 Observed 0.59 165.84 53.04 46.78Generated 0.00 165.81 52.66 46.76
416410 Macintyre Brook 465 Observed 0.31 93.34 35.77 26.61Generated 0.00 93.30 34.70 26.57
422338 Canal Creek 395 Observed 0.07 116.41 35.28 31.00Generated 0.00 116.00 34.70 30.97
422394 Cadamine River 325 Observed 8.99 89.35 47.30 26.62Generated 8.00 89.35 47.08 26.66
913009 Gorge Creek 248 Observed 0.03 31.86 12.99 14.44Generated 0.00 31.86 18.64 14.21
916002 Norman River 285 Observed 0.66 46.37 20.48 15.93Generated 0.00 46.40 20.10 15.93
922101 Coen River 166 Observed 0.26 81.89 25.60 19.82Generated 0.00 81.60 24.80 19.80
8.3.2 Continuing losses
It was investigated whether a theoretical distribution can be used to describe the
continuing loss distributions for the Queensland catchments. To examine the
probability distribution of the CL values, a total of 20 events from Bremer River
(Catchment ID: 143110), 9 events from Tenhill Creek (Catchment ID: 143212) and
27 events from Logan River (Catchment ID: 145003) were selected and a histogram
was plotted for all the 56 events as shown in Figure 8.8. The plot shows that CL
distribution can be approximated by an exponential distribution. Stochastic CL
Modelling losses in flood estimation 8-24
values can be generated from the fitted distribution, which can be used to obtain
derived flood frequency curves using a Joint Probability Approach. The impact of
stochastic CL values on design flood estimates is left to the future research efforts.
Figure 8.8 Distribution of observed CL values
8.4 Regionalisation of derived loss values
8.4.1 General
ARR (I. E. Aust., 1998) provided the best possible information for design flood
estimation. Previous editions were published in 1987, 1977 and 1958. ARR (I. E.
Aust., 1998) published the recommended loss values for the different states of
Australia. Table 8.2 and Table 8.4 compared between the existing recommended loss
values in ARR (I. E. Aust., 1998) and the computed loss values from this research.
The basis of Queensland existing recommended design losses in ARR (I. E. Aust.,
1998) is discussed in this section.
CL (mm/h)
6.515.514.513.512.511.51.51
Freq
uenc
y40
30
20
10
0
Std. Dev = 1.46 Mean = 1.40N = 56.00
CL (mm/h)
6.515.514.513.512.511.51.51
Freq
uenc
y40
30
20
10
0
Std. Dev = 1.46 Mean = 1.40N = 56.00
Modelling losses in flood estimation 8-25
8.4.2 Basis of Queensland existing recommended design losses in ARR (I. E. Aust.,
1998)
8.4.2.1 Recommended loss values in ARR (I. E. Aust., 1998)
The recommended design loss value in ARR (I. E. Aust., 1998) for Queensland is
divided into eastern Queensland and western Queensland. For western Queensland
the recommended loss values are considered as same as the Northern Territory; this
must be an assumption as there is no reference to this comment in ARR (I. E. Aust.,
1998). It means that no investigation was performed on western Queensland
catchments to compute the recommended loss values. For eastern Queensland the
three different recommended loss values like initial loss, median initial loss and
median continuing loss were selected from three single papers published by different
Authors (I. E. Aust., 1998), which are discussed below:
8.4.2.2 Recommended median initial loss for eastern Queensland
The recommended median initial loss in ARR (I. E. Aust., 1998) is 15.0-35.0mm has
been selected on the results of the research publication of Cordery (1970b). In this
paper a total of fourteen catchments were selected and their median initial loss values
were computed. These computed loss values are considered as a recommended
median initial loss values for Queensland. It is important to note that all of these
fourteen selected catchments are situated on or east of the Great Dividing Range in
New South Wales. None of them is situated within the boundary of Queensland. The
range of the median initial loss values for these fourteen catchments is from 0.91
inches (2.31 cm) to 1.72 inches (4.37 cm). The areas of these catchments are ranging
from 0.024 sq miles (.06 km2) to 97 sq miles (251.23 km2).
8.4.2.3 Recommended median continuing loss for eastern Queensland
In ARR (I. E. Aust., 1998) for eastern Queensland catchments, the recommended
median continuing loss 2.5 mm/h has been selected on the basis of the research
publication of Cordery and Pilgrim (1983). The Queensland Water Resources
Commission provided 37 values for six basins from Queensland. All of them were
Modelling losses in flood estimation 8-26
small to medium size catchments except the Boyan river catchment, which has an
area of 4105 km2. The other five catchment areas ranged from 350 km2 to 545 km2.
In term of land cover three of the catchments have dense forest, two have medium
forest and the Boyan River catchment has scattered forest. The mean loss rates for
these six catchments range from 0.8-10.0mm/h. The median continuing loss rates for
six catchments were in the range of 0.9-8.0mm/h. From this range, a loss value of
2.5mm/h has been used as the recommended median continuing loss rate in ARR (I.
E. Aust., 1998).
8.4.2.4 Recommended initial loss for eastern Queensland
The recommended initial loss value 0.0-140.0mm in ARR (I. E. Aust., 1998) was
selected on the basis of the research publication of Queensland Water Resources
Commission (1982). A total of four catchments were selected to compute the initial
loss for eastern Queensland. The areas of these catchments range from 340 km2 to
1375 km2. No initial loss values were published in this paper, and it is noted in ARR
(I. E. Aust., 1998) that, the initial loss values were obtained through private
communication.
8.4.3 Comments
In this research analysis, 48 catchments were selected from the whole of Queensland
to compute the loss values. Out of these 48 catchments the IL and CL values of 25
eastern catchments and 5 western catchments were computed. The range of the
median initial loss values of this research analysis can be used as recommended
median initial loss for eastern Queensland catchments as well as western catchments
to obtain more accurate design flood estimation. Also the median continuing loss
value of this research analysis would provide more accurate flood estimation than the
recommended median continuing loss value in ARR (I. E. Aust., 1998). Out of 48
selected catchments a total of 18 catchments were selected from the northern
Queensland. These loss values could be used as the recommended loss values for
different parts of Queensland catchments to obtain more accurate flood estimation
for Queensland.
Modelling losses in flood estimation 8-27
8.5 Validation of results
8.5.1 Calibration of runoff routing model
A catchment response model is needed to convert the rainfall excess hyetograph
produced by the loss model into surface runoff hydrograph. The proposed modelling
frame work was intended for application with a non-linear, semi-distributed runoff
routing model like RORB (Laurenson and Mein, 1997), as the runoff and streamflow
routing processes are generally non-linear and spatially variable. However for this
research a simpler conceptual runoff routing model (with a single non-linear storage
concentrated at the catchment outlet) was adopted to reduce computational effort.
This model can account for catchment non-linearity but not for the distributed nature
of catchment storage. For medium size catchment up to 500 km2, this model provides
an indication of what could be achieved with a semi-distributed model such as
RORB. For the adopted runoff routing model, the storage discharge relationship was
in the form shown in equation 8.3:
S = k Q m (8.3)
Where
S = Storage (m3)
k = Storage delay coefficient (hour)
Q = Streamflow discharge (m3/s)
m = Non-linearity parameter
The objective of this model calibration was to determine a value of k that results in
satisfactory fit for a range of recorded rainfall and runoff events at the catchment
outlet. In ARR (I. E. Aust., 1998) it is mentioned that the value of m is virtually
always less than one. Rahman et al. (2001b) used m = 0.8 for Victorian catchments.
Weeks and Stewart (1978) derived the m value for Queensland catchments and it was
found that the average m value for Queensland catchments was 0.73. As this research
was performed with Queensland catchments, hence m = 0.73 was used in this
analysis. The storage delay coefficient k was estimated in this analysis separately for
Modelling losses in flood estimation 8-28
each flood, then a global k value was selected for all events (e.g. a median or a mean
value) giving appropriate weight to the values from individual events depending on
data quality, purpose of modeling. The example of the quality of fit of the observed
and computed streamflow is shown in Figure 8.9. In Figure 8.9 event 27/12/97 for
the Black River is illustrated as an example. The difference in peak of the
hydrograph was 0.007% and the difference in volume of the hydrograph was 0.21%
of the observed hydrograph.
0
5
10
15
20
25
1 3 5 7 9 11 13 15 17 19 21 23Time (h)
Surf
ace
runo
ff (m
3/s)
ObservedComputed
Figure 8.9 Fitting the single non-linear storage model for the Black River catchment (event 27/12/97)
8.5.2 Comparison of flood frequency curves with the observed losses and the
derived losses from this research
Three catchments Burnett River, Bluewater Creek and Black River catchments were
selected randomly from Queensland for the validation of the storm loss results
obtained from this analysis. Three different k values were computed for three
different catchments as mentioned in section 8.5.1. Their catchment sizes were
370km2, 86km2, 256km2 respectively. A Joint Probability Approach flood estimation
Modelling losses in flood estimation 8-29
method was used to construct the flood frequency curves. To obtain the flood
frequency curves of these catchments, a set of values of duration, intensity, temporal
pattern and initial loss was generated to define the rainfall excess hyetograph, which
then routed through the calibrated runoff routing model to produce a corresponding
streamflow hydrograph. From this generated data the simulation of streamflow
hydrographs was drawn. The CL value was used as a constant value rather than
probability distributed value for the flood frequency curves. The peak of each of the
simulated hydrograph was used to construct a derived flood frequency curve. The
resulting observed flood frequency curves (partial series) for Burnett River,
Bluewater Creek and Black River catchments are compared in Figures 8.10, 8.11 and
8.12 respectively with the derived IL and CL loss values of the 48 catchments of this
research (following the empirical distribution approach using Cunnane’s plotting
position formula). This comparison of flood frequency analysis was performed by
using a FORTRAN program (Rahman, 1999). It is hypothesised that a four-
parameter Beta distribution can be used to describe the initial loss distributions for
Queensland catchments. A four-parameter Beta distribution is fitted to individual
site’s ILs data using observed values of lower limit (LL), upper limit (UL), mean and
standard deviation of ILs data (discussed in section 8.3). All parameters in the
observed and derived flood frequency curves were kept the same except the four
parameters of initial loss (upper limit, lower limit, mean and standard deviation) and
the median continuing loss for Burnett River and Black River.
To compare the flood frequency curves FORTRAN program (Rahman, 1999) needs
the four parameters of initial loss values (upper limit, lower limit, mean and standard
deviation), but ARR (I. E. Aust., 1998) did not illustrate the four parameters of the
recommended initial loss values. Hence for Bluewater Creek the comparison of the
flood frequency curves were performed between the observed floods, floods for CL
value of the ARR (I. E. Aust., 1998) and derived floods for the CL value of 48
catchments. The horizontal segment of these curves represents simulated events with
zero runoff.
Modelling losses in flood estimation 8-30
0
500
1000
1500
2000
2500
3000
3500
4000
0.1 1 10 100 1000 10000ARI (years)
Q (m
3 /s)
Observed floods
Derived floods with IL and CLof 48 catchments
Figure 8.10 Comparison of derived flood frequency curve with the losses of 48
selected catchments and ARR (I. E. Aust., 1998) for Burnett River.
In Figure 8.10 the comparison of the flood frequency curves was performed between
the observed floods of Burnett River and derived floods with the IL and CL loss
results of the 48 selected Queensland catchments. For both the flood frequency
curves, the generated data for the duration, temporal pattern and the intensity were
kept at same value for all the flood results. Only the IL and CL were changed, where
the IL values were probability distributed and the CL value was median loss value.
The result shows that the flood frequency curve from the derived loss values of 48
catchments fitted well with the flood frequency curve of Burnett River up to the
range of 900 years ARI.
In Figure 8.11 the comparison of the flood frequency curves was performed between
the observed floods of Bluewater Creek, computed floods with CL value of ARR (I.
E. Aust., 1998) derived floods with CL loss results of the 48 selected Queensland
catchments. For all the flood frequency curves, the generated data for the duration,
temporal pattern, intensity and IL were kept at same value for all the flood results.
Only the CL values were changed, where the CL values were the median loss values
Modelling losses in flood estimation 8-31
0100200300400500600700800900
1000
0.1 1 10 100 1000 10000ARI(years)
Q(m
3 /s)
Observed floods
Floods of CL loss of ARR
Derived flood of CL loss of 48catchments
Figure 8.11 Comparison of derived flood frequency curve with the losses of 48
selected catchments and ARR (I. E. Aust., 1998) for Bluewater Creek.
The probability distributed initial loss values were used for the Joint Probability
Approach flood estimation method, to construct the flood frequency curves. As ARR
(I. E. Aust., 1998) did not illustrate the probability distributed initial loss values,
hence comparison of the flood frequency curves in Figure 8.11 was performed using
only the median CL values for Bluewater Creek, the 48 catchments and ARR (I. E.
Aust., 1998). However, the difference between the flood frequency curves was not
remarkably high for Bluewater Creek.
In Figure 8.12 the comparison of the flood frequency curves was performed between
the observed floods of Black River and derived floods with the IL and CL loss results
of the 48 selected Queensland catchments. For both the flood frequency curves, the
generated data for the duration, temporal pattern and the intensity were kept same
value for all the flood results. Only the IL and CL values were changed, where the IL
values were probability distributed and the CL value was median loss value. The
result shows that the flood frequency curve from the derived loss values of 48
catchments fitted well with the flood frequency curve of Black River up to the range
of 1100 years ARI.
Modelling losses in flood estimation 8-32
0
1000
2000
3000
4000
5000
0.1 1 10 100 1000 10000ARI (years)
Q(m
3 /s)
Observed floods
Derived floods with IL and CL of 48catchments
Figure 8.12 Comparison of derived flood frequency curve with the results of 48
selected catchments and ARR (I. E. Aust., 1998) for Black River.
The results show that the derived flood frequency curves of all the three catchments
compare quite well with the derived loss values of this research analysis. From the
above findings it can be mentioned that the derived IL and CL loss values of this
research analysis, which were computed from the 48 selected catchments of the
entire of Queensland, can be used as the new recommended loss values for all the
Queensland catchments.
From this analysis it was found that the design flood peak did not change upto the
range of 350 years of ARI for Burnett River. For Bluewater Creek the design flood
peak remained unchanged upyo 65 years of ARI. Similarly for Black River the
design flood peak remain unchanged upto 460 years of ARI. Hence this analysis also
showed that the loss estimate from the rainfall does not change the probability of the
design flood peak for shorter range of ARI.
Modelling losses in flood estimation 9-1
CHAPTER 9
IMPLICATIONS AND IMPACTS OF THE RESEARCH
9.1. General
The implications and impacts from the outcome of this research work are discussed
in this chapter. These are categorised into two sections as follows:
1) Implications on model development
2) Socio-economic implications
9.2 Implications on model development
To compute the design IL and CL values, it is important to select a surface runoff
threshold value. ARR (I. E. Aust., 1998) recommends the IL-CL model to compute
design losses, but there is no surface runoff threshold given in ARR (I. E. Aust.,
1998) to use as a recommended threshold value. Hill et al., (1996a) used a surface
runoff threshold value to compute losses for Victorian catchments, which was
0.01mm/h. An investigation was performed on the 48 selected Queensland
catchments to select a suitable surface runoff threshold value that can be used as a
recommended threshold value for Queensland catchments. The surface runoff
threshold values used in this research were 0.005mm/h, 0.01mm/h and 0.02mm/h.
The result of the analysis showed that there is nothing called most suitable threshold
value which can be used as a recommended surface runoff threshold value for all
Queensland catchments. It depends, upon the catchment characteristics and the size
of the catchment. The result indicated that the median initial loss of a rainfall event is
quite sensitive to the selection of a surface runoff threshold value. However the
median continuing loss is not very sensitive to the selection of a surface runoff
threshold value. Some important information from this analysis was drawn.
Modelling losses in flood estimation 9-2
Information on surface runoff threshold value that will assist the designer to choose
the more accurate surface runoff threshold value are discussed below:
• For a candidate rainfall streamflow event the initial loss value increases by
increasing the surface runoff threshold value.
• The threshold value depends on the rainfall input and the size of the
catchment.
• For larger catchments the threshold value should be smaller, otherwise in loss
estimation a small increase in threshold value will convert a huge volume of
surface runoff into initial loss.
In ARR (I. E. Aust., 1998) it is mentioned that the continuing loss in a rainfall event
occurs for the remaining time period (t, discussed in section 8.2.2) of the storm, this
assumption is used for design storm. It is not possible to compute the duration of the
surface runoff from the design storm. In case of actual rainfall events this assumption
is not true, because the continuing loss of a rainfall event occurs until the end of the
surface runoff (t1, discussed in section 8.2.2). In this research analysis continuing loss
was computed using the pluviograph and streamflow data. It was possible to compute
the duration of the surface runoff for Queensland catchments. Continuing loss based
on both t and t1 were computed in this analysis. It was found that the median CL of
48 selected catchments based on t is 12.0% higher than the median CL1 (CL1
discussed in section 8.2.2) of the same 48 selected catchments based on t1. These
continuing loss values which were computed up to the end of the surface runoff are
more reliable than the ARR (I. E. Aust., 1998) recommended loss values. The
observed continuing loss value in this research analysis based on t1 would provide
better design flood estimation. Hence the computed continuing loss values which
were considered up to the end of the surface runoff can be used as recommended
continuing loss values for the future real time flood estimation.
ARR (I. E. Aust., 1998) recommended that continuing loss occurs at a constant rate
throughout the duration of the storm. An investigation was performed to examine
Modelling losses in flood estimation 9-3
whether continuing loss occurs at a constant rate for all storm durations or continuing
loss decays with duration of separate storm. Appendix F and Figures 8.1, 8.2, 8.3 and
8.4 shows that the continuing loss decreases with the duration of the rainfall event for
a catchment and for a region. From the Figures 8.1, 8.2, 8.3, 8.4 and Appendix F, it
can be noted that the fixed continuing loss values for Queensland catchments
recommended in ARR (I. E. Aust., 1998) is not acceptable. Hence continuing loss
function for Queensland catchments would provide probability distributed continuing
loss values, which will ultimately provide better design flood estimates.
9.3 Socio-economic implication
Rainfall runoff models are used to estimate design floods. These models require
several inputs such as rainfall duration, intensity, loss. Selection of an appropriate
loss value is very important for design flood estimation. A low loss value results in
over estimation of design floods (means large structures and higher capital cost). A
high loss value gives underestimation of design floods (means higher flood risk and
damage). Design of hydraulic structures that are subject to floods cost an average
$650 million per year in Australia (I. E. Aust., 1998). Since loss is an important input
for design flood estimation, an appropriate recommended loss value can save
millions of dollars per year in the design and construction of hydraulic structures
each year in Australia.
The steps followed in real time flood forecasting are similar to those used in the
design storm method of flood estimation. However, in real time forecasting, the
actual rainfall hyetograph over the catchment is combined with a model of losses to
determine the rainfall excess for input to a runoff-routing model. An investigation
was performed in this analysis, how conceptual loss models can provide more
accurate loss values by their more accurate loss functions. This improved conceptual
loss function can be used to provide better real time flood forecasting in the near
future. A more accurate flood forecasting will provide better socio-economic support
for the society.
Modelling losses in flood estimation 9-4
An accurate flood estimate is an important requirement for flood insurance. The
flood insurance is related to the anticipated flood damage or flood risk. The factors
which govern the flood risk are the magnitude, frequency and duration of the flood.
To estimate an accurate flood magnitude a suitable loss model is necessary. In this
research analysis an investigation was conducted on, how the existing IL-CL model
can compute better loss values by improving their loss functions. The outcome of
this research illustrated that an existing loss model by using an improved loss
function would provide more accurate flood estimation which would ultimately
reduce the flood insurance premium. Also the derived loss values from this research
would provide better flood estimation and ultimately would reduce the premium of
the insurance company.
Modelling losses in flood estimation 10-1
CHAPTER 10
CONCLUSIONS AND RECOMMENDATIONS
10.1 Introduction At present, there is inadequate information on design losses in most parts of
Australia including Queensland and this is one of the greatest weaknesses in
Australian flood hydrology. Currently recommended design losses in Queensland are
not compatible with design rainfall information in Australian Rainfall and Runoff
(ARR). These recommended losses in ARR are biased towards wet antecedent
conditions and are generally too low. Also design losses for observed storms show a
wide variability and it is always difficult to select a representative value of loss from
this wide range for a particular application. Despite the wide variability of loss
values, in the commonly used Design Event Approach, a single value of initial and
continuing losses is adopted. Because of the non-linearity in the rainfall-runoff
process, this is likely to introduce a high degree of uncertainty and possible bias in
the resulting flood estimates. In contrast, the Joint Probability Approach can consider
probability-distributed losses in flood estimation. For Queensland, no data on
probability-distributed losses are yet available.
This research examines the currently recommended design initial and continuing loss
data for flood estimation in Queensland. A total of 48 catchments from Queensland
ranging from 7 km2 to 635 km2 are selected. From these 48 catchments 969 rainfall
events that have the potential to produce significant runoff are selected and their
initial and continuing losses are computed from the observed pluviograph and
streamflow data. The observed initial losses are much greater than the currently
recommended values in the ARR (I. E. Aust., 1998) for eastern Queensland. The
observed median initial loss for all over Queensland is found to be 44.0% greater
than the ARR (I. E. Aust., 1998) recommended initial loss value. This finding has
important economic significance.
Modelling losses in flood estimation 10-2
The initial losses in eastern Queensland show a greater variability than that of
Victoria. The upper limit and median of initial losses in Queensland are 32.0% and
34.0% greater than those of Victoria.
It has also been found that a four-parameter Beta distribution can be used to
approximate the initial loss distributions in Eastern Queensland. Probability-
distributed initial losses and continuing losses were specified in 15 and 3 Queensland
catchments, and it is found that the probability distributed initial and continuing
losses can be used with the Joint Probability Approach to determine derived flood
frequency curves.
10.2 Overview of this study
The investigation involved in modelling losses in flood estimation is summarised in
the following steps:
10.2.1. Methods adopted for estimation of flood:
In ARR (I. E. Aust., 1998) the currently recommended method for design flood
estimation is based on Design Event Approach. The key assumption involved in this
approach is that the result in a flood output that has the same AEP as the rainfall
input. The success of this approach is crucially dependent on how well this
assumption is satisfied.
Significant developments have been made in the application of more advanced
techniques in flood estimation e.g. Joint Probability Approach (Rahman et al., 2001a,
b) and Continuous Simulation (Boughton, 2000). Thus it is clear that the Joint
Probability Approach, which considers the outcomes of events with all possible
combinations of input, values and if necessary their correlation structure is most
likely to lead to better estimates of design flows. Joint probability Approach is
theoretically superior to the Design Event Approach and regarded as an attractive
design method (I.E. Aust., 1998).
Modelling losses in flood estimation 10-3
It is difficult to use the continuous modelling for design flood estimation in rural
catchments because of the complexity involved in calibration of the model. The main
problems with the Continuous simulation arise from the difficulties in adequately
modeling the soil moisture balance, synthesising long records of rainfall and
evaporation at the appropriate temporal and spatial resolution, and accounting for
correlations between inputs. Hence Design Event Approach and Joint probability
were used in this research as the flood estimation methods.
10.2.2. Selection of study catchments to compute losses:
A total of 48 catchments were selected from the entire Queensland to compute loss
values. The electronic version of all the 48 catchment boundaries was collected from
the Department of Natural Resources and Mines. All the selected catchments were
small to medium size catchments, not large catchments. Since major regulation
affects the rainfall runoff relationship significantly, gauging stations subject to major
regulation (such as dams, back water effect etc) were not included in this research.
The selected catchments were mainly unregulated and rural and have reasonably long
rainfall and streamflow records.
Each catchment was represented by a stream gauging station, which was located at
the outlet of the catchment. The stream gauging record length for any catchment was
less than 11 years, that catchment was not selected as a candidate catchment. A total
of 82 pluviograph stations and 24 daily rainfall stations were used to compute the
loss values for all the selected catchments. If there is no pluviograph station inside
the catchment or within 50 km of the catchment boundary then this type of
catchment was not selected as study catchment. Some of the study catchments have
more than one pluviograph and daily rainfall stations inside the catchment boundary.
Most of these catchments were selected from coastal areas because there are only a
few catchments in western Queensland having reasonably long streamflow record.
The areas of the catchment were ranging from 7 km2 – 645km2. The median and
mean of the catchment areas were respectively 248 km2 and 273 km2. Only 8% of the
catchments were greater than 500 km2 and 15% were smaller than 100 km2.
Modelling losses in flood estimation 10-4
10.2.3. Procedure adopted to compile and analyse the data:
A total of 48 streamflow gauging stations were selected for this research work. A
total of 82 pluviograph stations and 24 daily rainfall stations were selected for this
research. The range of stream flow data is 11 years to 48 years. The range of the
pluviograph record lengths is 4 years to 48 years. The range of the daily rainfall data
record lengths is 7 years to 112 years.
The list of the selected stream gauge stations, their location, streamflow record
lengths, selected pluviograph stations and daily rainfall stations was developed. The
Basin ID is the unique identifier of stream gauging stations used by Queensland
Department of Natural resources and Mines (DNRM) for Queensland catchments.
For each catchment only one stream gauge station was selected but 5 pluviograph
stations and 5 daily rainfall stations were selected.
In some instances more than one pluviograph and daily rainfall stations were used to
obtain the representative rainfall of catchments. In that cases the weighted average
rainfall was used to obtain the representative rainfall of the catchment, rather than
using a single pluviograph data or the disintegrated daily rainfall data. All the three
stream flow, pluviograph and rainfall data were made concurrent before selecting the
rainfall and streamflow event, as to understand how streamflow varies with the
variation of rainfall.
10.2.4. Storm loss computation in this research:
Initial loss and continuing loss are two parameters simplified from the loss concept to
represent the average situation of storm over the catchment with time. This method is
widely used in Australian practice and hence adopted in this study. The most
appropriate design case would usually involve the use of a large storm from which
runoff was likely to occur from the whole catchment. Therefore the concept of runoff
produced by infiltration excess was adopted in this study.
The procedure which was adopted in this analysis to compute the continuing losses
was the same as the procedure adopted in ARR (I. E. Aust., 1998). The capacity rates
Modelling losses in flood estimation 10-5
of continuing loss is constant as assumed in ARR (I. E. Aust., 1998), it also could be
decreasing with the duration of the storm. In this analysis the loss values were
computed from rainfall and streamflow data. The proportional loss and the
volumetric runoff coefficient are computed in this analysis using the equations 6.4
and 6.5.
The rainfall excess and storm runoff that occur only on some portion of the
catchment, is called ‘partial area storm’ situations can more probably happen in
larger catchment and does not satisfy the criteria of IL-CL model for loss
computation. For that reason the loss model which was used for this research
analysis was suitable for small to medium sized catchments but not suitable for large
catchments. For large catchments the source area is far away from the rainfall
gauging station. Hence a considerable delay will occur between the beginning of
rainfall excess and the start of surface runoff at the catchment outlet, hence the
hydrographs tend to be broad crested.
Three different surface runoff threshold values were used in this research to examine
the effect of loss values with the change of threshold values. There was nothing
called most suitable surface runoff threshold value for the loss computation of
different catchments.
A computer program was developed by the author using FORTRAN PowerStation
4.0 to compute the initial loss, continuing loss, proportional loss and volumetric
runoff coefficient from all the selected events of 48 selected catchments is illustrated
in Appendix D.
10.2.5. Procedure adopted to separate the baseflow from streamflow:
The various methods of baseflow separation were reviewed to select an appropriate
method of baseflow for the study catchments. The two models of baseflow separation
were compared to select a suitable baseflow separation method for this research
work. The baseflow separation method Model 2 given in section 7.2 was selected for
this analysis and was applied to all the selected 48 study catchments to determine the
α value for each catchment and finally to compute their continuing losses. It was
Modelling losses in flood estimation 10-6
found that an acceptable baseflow separation coefficient α was selected for a
catchment by trial and error and sensitivity analysis, using 3 to 4 streamflow events
of that catchment. The rate of baseflow at any time was computed from the total
streamflow by using the α value in Equation 7.3.
The continuing loss is not very sensible with the change of α value. It was found that
a change in α by about 50% made less than 10% variation in the value of continuing
loss. For the remaining study catchments an acceptable baseflow separation
coefficient α was selected for each catchment to estimate baseflow separation from
streamflow hydrograph and ultimately to compute the continuing loss values from a
rainfall event.
10.2.6. Outcome of the research work:
Comparison was made between the computed median initial and median continuing
loss values with ARR (I. E. Aust., 1998) recommended median loss values. The
computed median IL of all over Queensland (Eastern and western catchments) is
44.0% higher than the ARR (I. E. Aust., 1998) recommended median initial loss
value. The computed median CL value is 33.0% less than that of ARR (I. E. Aust.,
1998) recommended median continuing loss value. The use of smaller initial losses,
according to ARR (I. E. Aust., 1998) recommendation, is likely to result in
significant overestimation of design floods.
It was found that the ARR (I. E. Aust., 1998) recommended median CL value which
is 2.5 mm/h is 49.7% higher than the computed median CL value for time t which is
1.67 mm/h. The computed median CL value for time t which is 1.67 mm/h is 12.1%
higher than the computed median CL1 value for time t1 which is 1.49 mm/h. This
result indicates that although the initial losses for the study catchments are relatively
higher than the ARR (I. E. Aust., 1998) recommended initial losses (Table 8.2), but
the CL and CL1 values are relatively smaller, which indicates that once the
catchments surface get saturated or near saturated, the subsequent loss rate decreases.
This is likely to produce relatively greater volume of runoff particularly, at the latter
part of the rainfall event.
Modelling losses in flood estimation 10-7
To select a surface runoff threshold value two things should be considered, the
rainfall input and size of the catchment. If the catchment size is big then the
threshold value should be small, because a small increase in threshold value would
result in a high initial loss. An assumption is made in ARR (I. E. Aust., 1998) that for
a rainfall event, the continuing loss occurs at a constant rate during the remainder of
the storm. An analysis was executed in this research to investigate the assumption in
ARR (I. E. Aust., 1998). This research illustrated that the assumption of ARR (I. E.
Aust., 1998) is not correct as Figures 8.1, 8.2, 8.3 and 8.4 shows that the continuing
loss decreases throughout the duration of the storm.
It was observed from the analysis of 15 catchments that the initial losses in
Queensland can be approximated by a four-parameter Beta distribution and from the
analysis of 3 catchments it is found that the CL distribution can be approximated by
an exponential distribution. Stochastic IL and CL values can be generated from the
fitted distribution, which can be used to obtain derived flood frequency curves using
a Joint Probability Approach. The computed IL and CL value which was derived
from the 48 selected catchments of the whole of Queensland can be used as the
recommended IL and CL values for all other ungauged Queensland catchments. To
examine the flood frequency analysis, three catchments were randomly selected and
their derived flood frequency curves were compared with the observed flood
frequency curves, where the computed IL and CL values of 48 catchments were used.
The result shows that the flood frequency curve fits quite well with each other. Only
in the case of Bluewater creek for greater ARI, the range from 17 to 470 years the
flood magnitudes increase by about 0.0-20.0%.
10.3 Conclusions
This research analysis examined how to improve the design initial and continuing
loss data for flood estimation in Queensland. The finding has important significance
for design flood estimation. The following conclusions can be drawn from the
analysis:
Modelling losses in flood estimation 10-8
• The average of the median IL for eastern catchments was 52.0% higher than the
ARR (I. E. Aust., 1998) recommended median initial loss value. The average of
the median CL value for eastern catchments is 39.2% less than that of ARR (I. E.
Aust., 1998) recommended median continuing loss value. The use of smaller
initial losses, according to ARR (I. E. Aust., 1998) recommendation, is likely to
result in significant overestimation of design floods.
• The computed median initial loss and continuing loss rate values obtained from
this research analysis can be used as the recommended design loss values for
Queensland catchments to obtain more accurate flood estimation for Queensland.
• The average of the median IL which was computed from the western catchments
was 44.8mm and the ARR (I. E. Aust., 1998) recommended median initial loss
value for western catchments is 0.0mm. It was observed that the computed
median CL value for western catchments was 12.86% higher than that of ARR (I.
E. Aust., 1998) recommended median continuing loss value
• To examine the effect of surface runoff threshold value on storm loss, an analysis
was performed on the 48 selected Queensland catchments. The result indicated
that the median initial loss of a rainfall event is quite sensitive to the selection of
a surface runoff threshold value i.e. if the threshold value increases 100.0% the
median initial loss increases in the range of 12.0-15.0%. But the median
continuing loss for both the cases CL and CL1 is not very sensitive to the
selection of a surface runoff threshold value.
• The continuing loss is not very sensitive to the selection of α value in the
baseflow separation. It was found that a change in α by about 50.0% makes less
than 10% variation in the value of continuing loss.
• The initial loss for Queensland catchments can be approximated by a four-
parameter Beta distribution. The generated initial loss preserves the statistics very
well with respect to the lower limit, upper limit, mean and standard deviation of
the observed losses. The lower limits of both the generated and observed data
Modelling losses in flood estimation 10-9
varies in the range of 0.0-2.73mm (which is close to 0.0mm) with an exception of
two values 4.59mm for Black River and 8.99mm for Cadamine River. For upper
limit the range of variation of generated and observed data is 0.0-0.80%, for
mean the variation is in the range of 0.17-3.9% and for standard deviation the
variation is 0-1.62%.
• In ARR (I. E. Aust., 1998) it is mentioned that the continuing loss in a rainfall
event occurs till the remaining time period (t, discussed in section 8.2.2) of the
storm. But this assumption is applied for design storm, in actual rainfall event the
continuing loss occurs till the end of the surface runoff (t1, discussed in section
8.2.2). Continuing loss based on both t and t1 were computed in this analysis. It
was found that the median CL of 48 selected catchments based on t is 12.0%
higher than the median CL1 (CL1 discussed in section 8.2.2) of the same 48
selected catchments based on t1. Also by taking the average of the range of both
the continuing loss values CL and CL1, it was found that the CL value was
76.15% higher than the CL1 values.
• It is recommended in ARR (I. E. Aust., 1998) that the continuing loss that occurs
for a rainfall event is at a constant rate during the remainder of the storm. But this
recommendation is not correct as Appendix F and Figures 8.1, 8.2, 8.3 and 8.4
show that the continuing loss decreases throughout the duration of the storm i.e.
it was not a single fixed value during the remainder of the storm. Hence the
continuing loss for the Queensland catchments were probability distributed
losses.
• The CL distribution can be approximated by an exponential distribution.
Stochastic CL values can be generated from the fitted distribution. This finding is
being confirmed with a larger data set. However a large data set is required to
derive stochastic continuing losses for application with Joint Probability
Approach.
Modelling losses in flood estimation 10-10
10.4 Suggestions for future research
Based on the findings of this study, following research tasks are identified as being
worth further research. These are:
• Derive probability-distributed initial and continuing losses for all the selected 48
catchments that can be used with the recently developed Joint Probability
Approach to flood estimation to obtain more accurate design flood estimates.
• Formulate a more realistic loss function that considers a fixed initial loss at the
beginning followed by continuing loss as a decaying function with respect to
rainfall duration. This loss function will be more suited to long duration storms
particularly on large catchments.
• This study did not address the issue of losses for extreme events. A detailed
investigation can be done on the rainfall events which have an initial loss of more
than 200 mm and a continuing loss of more than 20 mm/h.
• Regionalisation of the stochastic losses for use in ungauged catchments.
Modelling losses in flood estimation R-1
REFERENCES Ahern, P. A., & Weinmann, P. E. (1982). Considerations for design flood estimation using catchment modelling. Hydrol. and Water Resour. Symp., Melbourne, I.E. Aust., 44 - 48. Akan, A. O. (1992). Horton infiltration equation revisited. Journal of Irrigation and Drainage Engineering, 118(5), 828-830. Aston, A. R., & Dunin, F. X. (1979). Coupled infiltration and surface runoff on a 5 ha experimental catchment, Krawaree, NSW. Aust. Jour. Soil Res., 17, 53-64. Avery, D. R. (1986). Master of engineering science project report. School of Civil Engineering, The University of New South Wales. Avery, D. R. (1989). Application of RORB model to a flood warning system in the tweed valey, New South Wales. Hydrol. and Water Resour. Symp., I.E. Aust., 89/19, 90-94. Bates, B. C., Sumner, N. R., & Boyd, M. J. (1993). Non- linearity of flood runoff: what can be gleaned from calibrated runoff routing models? Civil Eng. Trans., I. E. Aust., CE35(2), 151-164. Bauer, S. W. (1974). A modified Horton equation for infiltration during intermittent rainfall. Hydrology Science Bulletin, 19(2), 219-225. Beaver, R. D., Hartman, J. P., & Wanielista, M. P. (1977). Infiltration and stormwater retention/ detention ponds.: Florida Technical University, Orland, FL. Bethlahmy, N. (1971). Hydrograph analysis: a computerised separation technique. Roneod report, USDA forest services, Intermountain Forest and Range Experiment Station, Moscow, Idaho, 14 pp. Beran, M. A. (1973). Estimation of design floods and the problem of equating the probability of Rainfall and Runoff. Paper presented at the Symposium on the Design of Water Resources Projects with Inadequate Data., Madrid, Spain. Beran, M. A., & Sutcliffe, J. V. (1972). An index of flood producing rainfall based on rainfall and soil moisture deficit. Jour. Hydrol., 17, 229-236. Betts, H. (1999). Floodplain planning - risk management from a local government perspective. Hydrol. And Water Resource. Symp. I.E. Aust. Nat. Conf. Brisbane, 99(25), 679 - 684. Boughton, W. C. (1987). Hydrograph analysis as a basis for water balance modelling. The institution of Engineers, Australia, Civil Engineering Transactions, Vol. CE 29, no.1, pp. 28-33.
Modelling losses in flood estimation R-2
Boughton, W. C. (1988). Partitioning streamflow by computer. Civil Engineering Transcation, Institution of Engineers Australia, pp. 285-291. Boughton, W. C. (1989). A review of the USDA SCS curve number method. Aust. Jour. Soil Reaser, Soil and Water Management and Conservation, Vol. 27, pp 511-523. Boughton, W. C. (1993). A hydrographed-based model for estimating the water yield of ungauged catchments. Hydrol. Water Resour. Symp., I.E. Aust., 317-324. Boughton, W. C., & Carrol, D. G. (1993). A simple combined water balance/ flood hydrograph model. Hydrol. and Water Resour. Symp., 93/14, 299-304. Boughton, W. C., Srikanthan, S., & Weinmann, E. (2000). Benchmarking a new design flood estimation system. Hydrol. and Water Resour. Symp., 570-575. Boughton, W. C. (2001). Continuous Simulation model for flood estimation, User Manual Computer Version. Bouwer, H. (1966). Rapid field measurement of air-entry value and hydraulic conductivity of soil as significant parameters in flow system analysis. Water Resour. Resear. 2(4), 729-738 Bouwer, H. (1969). Planning and interpreting soil permeability measurements. Jour. Irrig. Drain. Engg,, 95(1R3), 391-402. Boyd, M. J., Bates, B. C., Pilgrim, D. H., & Cordery, I. (1987). WBNM a general runoff routing model computer programs and user guide (Water Research Lab. 170): The University of New South Wales. Boyd, M. J., Rigby, E., & Vandrie, R. (2001). WBNM2001 version 3.00 runoff routing model for floods on natural, urban and part urban catchments.: University of Wollongong. Bradley, A. A., & Potter, K. W. (1992). Flood frequency analysis of simulated flows. Water. Resour. Res., 28(9), 2375-2385. Brakensiek, D. L. (1970). Infiltration of water into non-uniform soil, discussion. Jour. Irrig. Drain. Div., American Society of Civil Engineers. New York, NY. Bras, R. L., Gaboury, D. R., Grossman, D. S., & Vicens, G. L. (1985). Spatially varying rainfall and flood risk analysis. J. Hydraul. Eng. ASCE, 111(5), 754-773. Broadbridge, P., & White, I. (1988). Constant rate rainfall infiltration: A versatile nonlinear model 1. Analytic solution. Water Resour. Res., 24, 145 -154. Bureau of Meteorology (1963). Report on development of a flood forecasting system for the Lower Macleay River Valley, NEW. Cadavid, L. J., Obeysekera, T., & Shen, H. W. (1991). Flood Frequency Derivation from Kinematic Wave. Jour. Hydraul. Eng. ASCE, 117(4), 489-510.
Modelling losses in flood estimation R-3
Campbell, G. S. (1985). Soil Physics with BASIC. Elsevier, New York, 73-97. Carroll, D. G. (1991). Flood warning and monitoring in the Brisbane City area using event based radio telemetry systems. Unpublished MEngSc Thesis, QUT. Carroll, D. G. (1992). URBS, The urbanised catchment runoff model, Brisbane City Council. Carrol, D. G. (1994). The BCC catchment runoff routing model manual, version 3.3.: Brisbane City Council, Australia. Chander, S., & Shanker, H. (1984). Unit hydrograph based forecast model. Hydrol. Scien. Jour., 29(3), 279-291. Chapman, T. G. (1968). Catchment parameters for a deterministic rainfall-runoff model. Melbourne: In 'Land Evaluation', ed. G.A. Stewart, McMillan. Chapman, T. G. (1987). Unitgraph identification using only streamflow data. Civil Engineering Transaction, Institution of Engineers Australia, Vol. CE29, pp 187-191. Chapman, T. G. (1999). A comparison of algorithms for streamflow recession and baseflow separation, Hydrol. Process. 13, 701-714. Childs, E. C., (1969). An introduction to the physical basis of soil and water phenomena, John Wiley, New york, 210-212. Chu, S. T. (1978). Infiltration during a unsteady rain. Water Resour. Res., 14(3), 461-466. Clark, C. O. (1945). Storage and unit hydrograph. ASCE Trans., 110(1419-1446). Clark, R. D. S. (1980). Rainfall stream flow analysis to investigate spatial and temporal variability of rainfall excess generation. Jour. Hydrol., 47, 91-101. Cordery, I. (1968). Improved techniques in flood estimation. Intern. Assn. Hydrol. Sc., 140, 187 -195. Cordery, I. (1968). Initial loss for design flood estimation and flood forecasting. Flood Symposium, Sydney Division I.E. Aust and University of New South Wales, Sydney,. Cordery, I. (1970a). Antecedent wetness for design flood estimation. Civil Engineering Transactions, Institute of Engineers, Australia,, 12(2), 181 - 185. Cordery, I. (1970b). Initial loss for flood estimation and forecasting. Amer. Soc. Civil Engrs, Jour. Hydraulics Div (HY12), 96, 2447--2466.
Modelling losses in flood estimation R-4
Cordery, I., &Webb, S.N. (1974). Flood estimation in Eastern New South wales – a design method, Civil Engg. Trans., Institution of Engineers, Australia., Vol CE 16, pp 87 – 93. Cordery, I., & Pilgrim, D. H. (1983). On the lack of dependence of losses from flood runoff on soil and cover characteristics. IAHS, 140, 187 - 195. Cordery, I., Pilgrim, D. H., & Doran, D.G. (1983) Some hydrological characteristics of arid western New South Wales . Hydrol. And Water Resour. Symp., Institution of Engineers, Australia, Natl. Confer. Publ. no. 86/13, pp 203 – 208. Corradini, C., Melone, F., & Ubertini, L. (1986). A semi-distributed adaptive model for real-time flood forecasting. Water Resour. Bull., 22(6), 1031-1038. Corraini, C., & Melone, F. (1986). An adaptive model for on-line flood predictions using a piecewise uniformity framework. Jour. Hydrol., 88, 365-382. Crapper, G. C. A. (1989). Development of areal-time reservoir routing flood warning procedure using the RORB model. Hydrol. and Water Resour. Symp. I.E. Aust, 89/19, 86-89. Crawford, N. H., & Linsley, R. K. (1966). Digital simulation in Hydrology: Stanford Watershed model IV, Technical Report 39. California: Stanford University. CSIRO, (1992). TOPOG user guide, version 4.0. CSIRO Division of Water Resources. Canberra. Dickinson, W.T., Holland, M.E., Smith, G.L. (1967). An experimental rainfall-runoff facility, Hydrology Paper no. 25, Colorado State University, 78 pp. DNRM, (2000). Department of Natural Resources and Mines. An index for stream gauging stations in Queensland. Dyer, B. G., Nathan, R. J., McMahon, T. A., & O'Neill, I. C. (1994). Development of regional prediction equations for the RORB runoff routing model.: CRC Report : 94/1. CRC for Catchment Hydrology, Monash University, Australia. Eagleson, P. S. (1972). Dynamics of flood frequency. Water Resour. Res., 8(4), 878-898. Flavell, D. J., & Belstead, B. S. (1986). Losses for design flood estimation in Western Australia. Hydrol. Water Resour. Symp., I. E. Aust., 86(13), 203 - 208. Gieseman, M. (1986). Melbourne's flood warning system. Flood warning in Australia. Centre for Resources and Environmental Studies, Canberra, 153-174. Goyen, A. G. (1983). A model to statistically derive design rainfall losses. Hydrol. and Water Resour. Symp. Hobart, 220-225.
Modelling losses in flood estimation R-5
Green, W. H., & Ampt, C. A. (1911). Studies on soil physics; flow of air and water through soils. Jour.. Agricul. Sci., 4, 1-24. Hann, C. T. (1977). Statistical methods in Hydrology. Iowa State: The Iowa State University Press. Haan, C. T., & Schulze, R. E. (1987). Return period flow prediction with uncertain parameters. ASCE Trans., 30(3), 665-669. Hall, A.J. (1971). Baseflow recessions and the baseflow hydrograph separation problem . Hydrology Papers , The Institution of Engineers, Australia, pp. 159-170 Harvey, R. A. (1982). Estimation of the probable maximum flood - Western Australia. Proc. of the Workshop on spillway Design, Department of National Development and Energy. AWRC Conference series, 6, 157-176. Hill, P.I. (1993). Extreme flood estimation for the Onkaparinga River catchment, M. Eng. Science Thesis, Department of Civil and Environmental Engineering, University of Adelaide Hill, P. I., & Mein, R. G. (1996). Incompatibilities between storm temporal patterns and losses for design flood estimation. Hydrol. and Water Resour. Symp., Hobart, I.E. Aust., 2, 445 - 451. Hill, P. I., Maheepala, U. K., Mein, R. G., (1996). Empirical analysis of data to derive losses: methodology, programs and results, working document, CRC Report 96/5.CRC for Catchment Hydrology, Monash University, Australia. Hill, P. I., Mein, R. G., & Weinmann, P. E. (1996a). Testing of improved inputs for design flood estimation in South-Eastern Australia. CRC Report 96/6, CRC for Catchment Hydrology , Monash University, Australia. Hill, P. I., Maheepala, U. K., Mein, R. G., & Weinmann, P. E. (1996b). Empirical analysis of data to derive losses for design flood estimation in South-Eastern Australia.: CRC Report 96/5.CRC for Catchment Hydrology, Monash University, Australia. Hill, P. I., Mein, R. G., & Weinmann, P. E. (1997). Development and testing of new design losses For South-Eastern Australia. Hydrol. and Water Resour. Symp., I.E.Aust. Auckland:, 71 – 76. Hill, P. I., Mein, R. G., Siriwardena. (1998). How much rainfall becomes runoff? loss modelling for flood estimation. CRC Report 98/5, CRC for catchment hydrology, Monash University, Australia. Hino, M., Hasebe, M. (1985) Separation of a storm hydrograph into a runoff component by both filter separation AR method and environmental isotope tracers. Jour. of Hydrol., Vol. 85, pp. 251-264.
Modelling losses in flood estimation R-6
Hoang, T. M. T., Rahman, A., Weinmann, P. E., Laurenson, E. M., & Nathan, R. J. (1999). Joint Probability description of design rainfalls. Hydrol. and Water Resour. Symp., Brisbane, I.E. Aust., 99(25), 379 - 384. Horton, R. E. (1940). An approach toward a physical interpretation of infiltration capacity. Proc. Soil. Sci. Soc., 5, 399 - 417. Hromadka, T. V. (1996a). Rainfall – Runoff probabilistic simulation program :1 synthetic data generation. Environmental Software, Elsevier Science Ltd, Great Britain, 11(4), 235 - 242. Hromadka, T. V. (1996b). A Rainfall - Runoff probabilistic simulation program :2 Synthetic data generation. Environmental Software, Elsevier Science Ltd, Great Britain, 11(4), 243 - 249. Hubber, W. C., Heaney, J. P., Nix, S. J., Dickenson, R. E., & Polman, D. J. (1982). Storm water management model user's manual.: Department of Environmental Engineering Sciences, University of Florida. Florida. Hubber, W. C., Cunningham, B. A., & Cavendor, K. A. (1986). Continuous SWMM modelling for selection of design events. Comparison of urban drainage models with real catchment data (C. Maksimocic & M. Radojkovic Ed. ). Yugoslavia: DubrovniK. Hughes, W. C. (1977). Peak discharge frequency from rainfall information. Jour. Hydraul. Div. ASCE, 103(1), 39-50. Hydsys (2002). Hydsys version 8.0, pluviograph data extraction software for PC, Hydsys Pty Ltd. Institution of Engineers Australia. (1958). Australian Rainfall and Runoff, Institution of Engineers, Australia. Institution of Engineers Australia. (1977). Australian Rainfall and Runoff, Institution of Engineers Australia. Institution of Engineers Australia. (1987). Australian Rainfall and Runoff, Institution of Engineers Australia. Institution of Engineers Australia. (1998). Australian Rainfall and Runoff, Vol, 1and 2, Institution of Engineers Australia. Institution of Engineers. Australia. (2001). Australian rainfall and Runoff, Institution of Engineers Australia. James, W., & Robinson, M. A. (1982). Continuous Models Essential for Detention Design. Engineering Foundation Conference, Amer. Soc. of Civil Eng., 163-175.
Modelling losses in flood estimation R-7
James, W., & Robinson, M. A. (1986). Continuous deterministic urban runoff modelling. Comparison of urban drainage models with real catchment data (C. Maksimovic & M. Radojkovic Ed. ). Yugoslavia: Dubrovnik. Knee, R. (1986). Derived loss parameters in the A.C.T. First report. Hydrol. And Water Resour. Unit, Department of Territories, H.W.R. no. 86/15, September. Kobayashi, D. (1985). Separation of the snowmelt hydrograph by stream temperatures. Jour. of Hydrol., Vol.76, pp 155-162. Kobayashi, D. (1986). Separation of a snowmelt hydrograph by stream conductance. Jour. of Hydrol., Vol.84, pp 157-164. Kool, J. B., Parker, J. C., & Genuchten, M. T. (1985). Determining soil hydraulic properties from one-step outflow experiments by parameter estimation: 1. Theory of numerical studies. Soil Scien. Soc. Amer. Jour.., 49, 1348-1353. Kumar, S., & Jain, S. (1982). Application of SCS infiltration model. Water Resour. Bull., 18(3), 503-507. Laurenson, E., M., & Pilgrim, D., H. (1963). Loss rates of Australian catchments and their significance. I.E Aust., 35(1-2), 9 - 24. Laurenson, E. M. (1974). Modelling for stochastic-deterministic hydrologic systems. Water Resour. Res., 10/5, 955-961. Laurenson, E. M., & Mein, R. G. (1997). RORB version 4 runoff routing program user manual, Monash University, Australia. Linsley, R., & Crawford, N. (1974). Continuous simulation models in Hydrology. Geophysical Research Letters, 1(1), 59-62. Linsley, R. K., Kohler, M. A., & Paulhus, J. L. (1982). Hydrology for engineers. McGraw-Hill Co. ( Vol. 3). New York. Linsley, R. K., Kohler, M. A., & Paulhus, J. L. H. (1988). Hydrology for Engineers.: McGraw-Hill Book Company. Loy, A., Ayre, R. A., & Ruffini, J. L. (1996). Regional loss model relationships for catchments in South East Queensland. Hydrol. Water Resour. Symp. I.E. Aust. Hobart, 133-140. Lumb, A. M., & James, L. D. (1976). Runoff files for flood hydrograph simulation. Jour. Hydraul. Div. ASCE, 1515-1531. Lyne, V.D., (1979). Recursive modelling of sluggish and time varying streamflow responses, M. Eng., Science Thesis, Department of Civil Engineering, University of Western Australia.
Modelling losses in flood estimation R-8
Lyne, V.D., and Hollick, M. (1979). Stochastic time variable rainfall runoff modelling. The Institution of Engineers, Australia, Hydrology and Water Resources Symposium, Perth, pp. 89-92 Maller, R. A., & Sharma, M. L. (1980). Time of ponding and areal infiltration from spatially varying parameters. Hydrol. Water Resour. Symp., I. E. Aust., 80(9), 162 - 165. Malone, T. (1999). Using URBS for real time flood modelling. Water 99 Joint Congress, Brisbane, 603-608. Manly, R. E. (1974). Catchment Models for River Management. Unpublished MSc Thesis, University of Birmingham. Mapinfo (1998). Mapinfo prefessional version 5.0, GIS software, Mapinfo corporation. Marivoet, J. L., & Vandewiele, G. L. (1980). A real-time rainfall runoff model. In Hydrological Forecasting. IAHS, 129, 409-418. Mein, R. (1999). Overview of the Flood Hydrology Program of the CRC for Catchment Hydrology. Hydrol. And Water Resour. Symp. I.E. Aust. Brisbane, 99(25), 173 - 177. Mein, R. G. (1980). Recent development in modelling of infiltration: Extension of the Green - Ampt model. Hydrol. Water Resour. Symp., I. E. Aust., 80(9), 23 -28. Mein, R. G., & Larson, C. L. (1971). Modelling infiltration during a steady rain. Water Resour. Res., 9(2), 384 - 394. Mein, R. G., Laurenson, E.M., & McMahon, T.A. (1974). “Simple nonlinear method for flood estimation. J. Hyd. Div. ASCE, Vol 100, No. HY11, pp 1507 – 1518. Moore, I. D., Larson, C. L., Slack, D. C., Wilson, B. N., & Idike, F. (1981). Modelling infiltration: A measurable parameter approach. Jour. Agri. Engg. Res., 25, 21-32. Moore, R. J. (1985). The probability-distributed principle and runoff production at point and basin scales. Hydrol. Scien., 30, 273-297. Nandakumar, N., Mein, R. G., & Siriwardena, L. (1994). Loss modelling for flood estimation - A Review. CRC Report 94/4, CRC for Catchment Hydrology, Monash University, Australia. Nathan, R. J. and McMahon, T.A.(1990), Evaluation of automated techniques for
baseflow and recession analysis, Water Resour. Res., 26 (7), 1465 – 1473. Natural Environment Research Council. (1975). Flood studies report. Hydrological Studies, London, 1.
Modelling losses in flood estimation R-9
O'Loughlin, E.M., Cheny, N.P., Burns, J., (1982) The bushranger experiment; Hydrological response of a eucalypt catchment to fire. The First National Symposium of Forest Hydrology, Melbourne, I.E. Aust. Nat. Conf, Pub. No. 82/6, pp 132-138 O'Loughlin, G. (1988). The ILSAX Program for urban stormwater drainage design and analysis (Users Manual for microcomputer version V2.0): Sydney University of Technology. Onstad, C. A., Olson, T. C., & Stone, L. R. (1973). An infiltration model tested with monolith moisture measurements. Soil Science, 116(1), 13-17. Overney O. et al. (1995). Estimating expected flood values with Monte-Carlo Methods. Paper presented at the Bernier Conference, Paris. Pearse, M. A. (1997). A Joint Probability Approach to estimating the frequency of extreme precipitation for flood estimation. (PhD Thesis ): Monash University. Peck , E. L., Johnson, E. R., Krouse, K. M., & Carroll, T. R. (1980). Hydrological update techniques used by the US National Weather Service. Hydrological Forecasting, IAHS, 129, 419-424. Philip, J. R. (1957). The theory of infiltration, 1: The infiltration equation and its solution. Soil. Science, 77(2), 153 - 157. Philip, J. R. (1957a). The theory of infiltration, 4, sorptivity and algebraic infiltration equations, Soil. Science, 84(3), 257 – 264 Pilgrim, D., Huff, D., Steels, T (1979). Use of specific conductance and contact time relations for separating flow components in storm runoff. Water. Resour. Resear., Vol. 15, no.2, pp. 329-339 Pilgrim, D. H., & Robinson, D. K. (1988). Flood estimation in Australia - progress to the present possibilities for the future. Civil Eng. Trans., I.E. Aust., CE30(4), 187 - 206. Porter, J. W., & McMahon, T. A. (1971). A model for the simulation of stream flow data from climatic records. Jour. Hydrol., 13, 297-324. Queensland Water Resources Commission (1982). Private communication, W.D. Weeks, 17 December, 1982. Rahman, A. (1999). FORTRAN routines for determining derived flood frequency curve by Monte Carlo Simulation, Working Document. CRC for Catchment Hydrology, Monash University, Australia. Rahman, A., Hoang, T., Weinmann, P., & Laurenson, E. M. (1998). Joint Probability Approaches to design flood estimation. CRC Report 98/8, CRC for Catchment Hydrology, Monash University, Australia.
Modelling losses in flood estimation R-10
Rahman, A., Weinmann, P. E., & Mein, R. G. (2000). The probabilistic nature of initial losses and its impacts on design flood estimates. Hydro. and Water Resour. Symp., Perth, I. E. Aust., 1, 71 -75. Rahman, A., Weinmann, P. E., Hoang, T., Laurenson, E., M, & Nathan, R. J. (2001a). Monte Carlo Simulation of flood frequency curves from rainfall. CRC Report 01/4. CRC for Catchment Hydrology, Monash University, Australia. Rahman, A., Weinmann, P. E., Hoang, T., Laurenson, E., M, & Nathan, R. J. (2001b). Monte Carlo Simulation of flood frequency curves from rainfall. Jour. Hydrol. Rajendran, R., & Mein, R. G. (1986). Determination of rainfall excess on spatially variable catchments. Jour. Hydraul., 83(1/2), 67 - 89. Renard, K. G., (1977) past, present and future water resources research in arid and semi-arid areas of the south western United States. Institution of Engineers, Australia, National conference, Pub., 77/5: 1 – 29. Richards, L. A. (1931). Capillary conduction through porous medium. physics, 1, 313-318. Ross, P. J. (1990a). Feeicient numerical methods for infiltration using Richards equation. Water Resour. Res., 26(2), 279-290. Ross, P. J. (1990b). SWIM - a simulation model for soil water infiltration and movement, . Queensland: CSIRO Divisions of Soils. Russell, S. O. (1977). A new method for computing design flows for urban drainage. Hydrol. Sym. Edmonton, Canada, 11, 188-197. Russel, S. O., Kenning, B. F. I., & Sunnell, G. J. (1979). Estimating design flows for urban drainage. Jour. Hydrau. Div., 105, 43-52. Sharma, M. L., Gander, G. A., & Hunt, C. G. (1980). Spatial variability of infiltration in a watershed. Jour. Hydrol., 45, 101 - 122. Sharma, M. L., & Seely, E. (1979). Spatial variability and its effect on areal infiltration. Hydrol. and Water Resour. Symp. I.E. Aust., 79(10), 69 - 73. Shirmohammadi, A., Knisel, W.G., and Sheridan, J.M. (1984) An approximate method for partitioning daily streamflow data. Jour. of Hydrol.Vol. 74, pp. 335-354. Simpson, R. J., Wood, T. R., & Hamlin, M. J. (1980). Simple self-correcting models for forecasting flow on small basins in real-time. Hydrological Forecasting, IAHS, 129, 433-443. Singh, V.P. (1989). Hydrological systems watershed modelling. (Prentice-Hall Inc. : New Jersey)
Modelling losses in flood estimation R-11
Siriwardena, L., & Weinmann, P. E. (1996). Derivation of areal reduction factors for design rainfalls in Victoria. CRC Report 96/4, CRC for Catchment Hydrology , Monash University, Australia, 96(4), 60. Skaggs, R. W., & Khaleel, R. (1982). Infiltration in hydrologic modelling of small watershed. C.T. Haan, Amer. Soc. of Agricul. Eng., St. Joseph, MI, 121-166. Smith, R. E., & Hebbert, H. B. (1979). A Monte-Carlo analysis of the hydrologic effects of spatial variability of infiltration. Water Resour. Res., 15(2), 419-428. Srikanthan, R., & Kennedy, M. R. (1991). Rainfall antecedent to storm burst from which temporal patterns were derived for Australian Rainfall and Runoff. Hydro. and Water Resour. Symp., Perth, I. E. Aust., 19, 280 - 282. Srikanthan, R., Sooriyakumaran, P., & Elliott, J. F. (1997). Comparison of four real-time flood forecasting models. Hydrol. Water Resour. Symp. Auckland, 185-190. Terstriep, M. L., & Stall, J. B. (1974). The Illinios urban area simulator, ILLUDAS ( 58). Urbana: Illinios State Water Survey. Titmarsh, G. W., Pilgrim, D. H., Cordery, I., & Hoesein, A. A. (1989). An examination of design flood estimation using the US Soil Conservation Services Method. Hydrol.and Water Resour. Symp., 89/19, 241-247. Todd, D. K. (1980). Ground water hydrology, Wiley. New York. Tucci, C. M. E., & Clarke, R. T. (1980). Adaptive forecasting with a conceptual rainfall-runoff model. Hydrological Forecasting, IAHS, 129, 445-454. U.S. Soil Conservation Services. (1985). National Engineering Handbook, Section 4, Hydrology. Washington, D.C.: U.S. Department of Agriculture. Van Mullem, J. A. (1991). Runoff and peak discharges using Green-Ampt infiltration model. Jour. Hydraul. Engg., 117(3), 354-370. Viessman, W. J., Lewis, G. L., & Knapp, J. W. (1989). Introduction to Hydrology.: Harper Collins Publisher. Walsh, M. A., & Pilgrim, D. H. (1993). Re-assessment of some design parameters for flood estimation in New South Wales. Hydro. Water Resour symp., I.E. Aust., 93(14), 251-256. Walsh, M. A., Pilgrim, D. H., & Cordery, I. (1991). Initial losses for design flood estimation in New South Wales. Hydro. and Water Resour. Symp, Perth, I.E Aust., 19, 283 - 288. Waugh, A. S. (1991). Design losses in flood estimation. Hydrol. Water Resour. Symp., I.E. Aust., 12, 629 - 630.
Modelling losses in flood estimation R-12
Weinmann, E., Rahman, A., Hoang, T., Laurenson, E. M., & Nathan, R. J. (1998). A new modelling framework for design flood estimation. International Conference on Hydraulics in Civil Engineering, Adelaide, Australia, 393-398. Weinmann, P. E., Rahman, A., Hoang, T. M. T., Laurenson, E. M., & Nathan, R. J. (2000). Monte Carlo Simulation of flood frequency curves from rainfall - the way ahead. Hydrol. and water Resour. Symp., I. E. Aust., Perth, 1, 564-569. Weeks, W.D., Stewart, B.J. (1978). Linear and nonlinear runoff routing for ungauged catchments, Hydrology Symposium, I.E. Aust., National Conference Publication78/9, pp 124-128.
APPENDIX A program sffor1 c ****************************************************** c This program is written by Mr.Mahbub Ilahee to format c the stream flow data received from Department of c Natural resources. So that this stream flow data can be c be used as a input file for loss computation in Design c Flood Estimation. Date 01/02/2002 c c ****************************************************** double precision x(500000),y(50000),qual(500000),sf(500000) integer jd(500000),jm(500000),jyr(500000),jh(500000) integer jq(500000) character*50 fn1,fn2 write (*,*)'Enter filename containing unformatted flow data' read(*,*)fn1 open(3, file=fn1,status='unknown') write (*,*)'Enter output filename containing formatted flow data' read(*,*)fn2 open(7, file=fn2,status='unknown') open(5,file='floint.out',status='unknown') i=1 5 read(3,*,end=10)x(i),y(i),qual(i) c write(*,*)x(i) c write(*,*)y(i) c write(*,*)qual(i) c write(*,555)x(i),y(i),qual(i) write(5,555)x(i),y(i),qual(i) i=i+1 goto 5 10 continue rewind(5) i=1 11 read(5,560,end=15)jd(i),jm(i),jyr(i),jh(i),sf(i),jq(i)
Modelling losses in flood estimation A-1
c write(*,*)jq(i) c write(*,*)jm(i) c pause write(7,655)jyr(i),jm(i),jd(i),jh(i),sf(i),jq(i) i=i+1 goto 11 15 continue 555 Format(1x,f14.1,2x,f11.3,10x,f5.1) 560 Format(1x,i2,i2,i4,i2,6x,f9.3,5x,i3) 655 Format(14x,i4,i2,i2,i2,4x,f12.3,i4) close(3) close(5) close(7) stop end
Modelling losses in flood estimation A-2
APPENDIX B program sffor2 c ****************************************************** c This program is written by Mr.Mahbub Ilahee to format c the stream flow data received from Department of c Natural resources. So that this formatted stream flow data c can be be used as a input file for loss computation in Design c Flood Estimation. Date 01/03/2002 c c ****************************************************** double precision x(500000),y(50000),qual(500000),sf(500000) integer jd(500000),jm(500000),jyr(500000),jh(500000) integer jq(500000) character*50 fn1,fn2, fn(50000) write (*,*)'Enter filename containing unformatted flow data' read(*,*)fn1 open(3, file=fn1,status='unknown') write (*,*)'Enter output filename containing formatted flow data' read(*,*)fn2 open(7, file=fn2,status='unknown') open(5,file='floint.out',status='unknown') i=1 5 read(3,*,end=10)x(i),y(i),fn(i),qual(i) c write(*,*)x(i) c write(*,*)y(i) c write(*,*)qual(i) c write(*,555)x(i),y(i),qual(i) write(5,555)x(i),y(i),qual(i) i=i+1 goto 5 10 continue rewind(5) i=1
Modelling losses in flood estimation B-1
11 read(5,560,end=15)jd(i),jm(i),jyr(i),jh(i),sf(i),jq(i) c write(*,*)jq(i) c write(*,*)jm(i) c pause write(7,655)jyr(i),jm(i),jd(i),jh(i),sf(i),jq(i) i=i+1 goto 11 15 continue 555 Format(1x,f14.1,2x,f11.3,10x,f5.1) 560 Format(1x,i2,i2,i4,i2,6x,f9.3,12x,i3) 655 Format(14x,i4,i2,i2,i2,4x,f12.3,i4) close(3) close(5) close(7) stop end
Modelling losses in flood estimation B-2
APPENDIX C program RAINDATA3 c ***************************************************** c Changing the rainfall data from horizontal form c to vertical form.This program is written by Mr.Mahbub c ilahee on 30/03/2002 c ***************************************************** integer m,n,jsn,jy,jm,jd,jh,jav,d1 real x(5000),Tr,c character*15 fn1,fn2 write (*,*)'Enetr input file:' read(*,*)fn1 Write(*,*)'Enetr output file' read(*,*)fn2 open(5, file=fn1, status='unknown') open(6, file=fn2, status='unknown') c write(6,*)'Daily Rainfall data' c write(6,*)'STN NO, YY, MM,DD,RAINFALL(mm)' c write(6,*) Do m=1,85 Do n=1,12 read(5,60,end=10)jsn,jy,jm,jd,jh,Tr,jav,(x(i),i=1,31) c write(*,100) jsn,jy,jm,jd,jh,Tr,jav,(x(i),i=1,31) c pause c write(*,*)'Daily Rainfall data' c write(*,*) if(jm .eq.4 .or. jm .eq. 6 .or. jm .eq. 9 .or. jm .eq. 11)then do i=1,30 write (6,200)jsn,jy,jm,i,x(i) enddo endif if(jm .eq.1 .or. jm .eq. 3 .or. jm .eq. 5 .or. jm .eq. 7 +.or. jm .eq. 8 .or. jm .eq. 10 .or. jm .eq. 12 )then
Modelling losses in flood estimation C-1
do i=1,31 write (6,200)jsn,jy,jm,i,x(i) enddo endif c=MOD(jy,4) if (c .eq. 0) then d1=29 endif if (c .ne. 0) then d1=28 endif if (jm .eq. 2) then do i=1,d1 write (6,200)jsn,jy,jm,i,x(i) enddo endif c write(*,*)'STN NO, YY, MM, RAINFALL(mm)',jsn,jy,jm, c + (x(i), i=1,31) enddo enddo 10 write(*,*)'no data' close(5) close(6) 60 format(7x,i6,1x,i4,1x,i2,1x,i1,1x,i1,2x,f5.1,1x,i2, + 3x,f4.1,30(8x,f5.1)) 200 format(i6,x,i4,x,i2,x,i2,x,f5.1) c100 format(I6,4x,I4,2x,I2, 31(3x,/f5.1)) stop end
Modelling losses in flood estimation C-2
APPENDIX D
program Icloss c A program to compute continuing loss, proportional loss c and volumetric runoff coefficient c written by Mahbub Ilahee at QUT 28/05/03 character*50 file2,file3,file5 integer iyr(300),im(300),id(300),ih(300) real x(300),y(300),u(300),v(300),z(300) real xsum,il,xsum1,A,PL,ROC,CL,CL1 real sf(300),hrf(300),w1 write(*,*)'Enter inputfile contains streamflow and pluvio data' read(*,*)file2 open(2,file=file2,status='old') write(*,*)'Put the value of fraction of baseflow' read(*,*)w1 write(*,*)'put the value of area, (A)' read(*,*)A Write(*,*)'Enter output filename containing baseflow data:' read(*,*)file3 open(3,file=file3,status='unknown') Write(*,*)'Enter output file contains IL,CL,CL1,PL,ROC,d1,d2' read(*,*)file5 open(5,file=file5,status='unknown') write (5,202)'yr','m','d','h','il(mm)','CL(mm/h)','PL','ROC', + 'j-1','j1-1','CL1(mm/h)','thresholdvalue=.01' write (3,201)'yr','m','d','h','sf','hrf','x','y', + 'z','u','v' c x(i)=the difference between the consecutive two c streamflow values c y(i)= x*fraction of the surface runoff (a) c z(i)=y(i)+y(i-1)(one timestep before) c u(i)= sf(i) if u(i)<0, otherwise u(i)=z(i) c v(i)=sf(i)-u(i)(surface runoff) c xsum = total rainfall c il = initial loss c reading streamflow and pluvio data xsum=0 il=0 xsum1=0
Modelling losses in flood estimation D-1
i=1 k=1 30 read (2,*,end=21)iyr(i),im(i),id(i),ih(i),sf(i),hrf(i) if(hrf(i) .eq. -99.000)then n=i-1 write (3,*)'no of line=',n go to 21 else continue endif if(hrf(i) .eq. -1111.000)then n=i-1 c write (3,*)'no of line=',n c write(3,*) 'endfil' endfil=-1111.000 go to 21 else continue endif x(i)= (sf(i)-z(i-1)) y(i)=(sf(i)-z(i-1))*w1 z(i)=(z(i-1)+(sf(i)-z(i-1))*w1) if(y(i) .lt. 0)then u(i)=sf(i) else u(i)=z(i) endif v(i)=(sf(i)-u(i)) if (z(i-1) .gt. sf(i))then x(i)=0 y(i)=0 endif if (k .gt. 3)then go to 28 else if(sf(i)-sf(i-1) .lt. (0.00277*A)) then c if (sf(i) .lt. (.00277*A))then x(i)=0 y(i)=0 z(i)=sf(i) u(i)=sf(i) v(i)=0 else
Modelling losses in flood estimation D-2
k=i if (i .eq. 1)then v(i)=0 x(i)=0 else continue endif c write(*,*)'start of surface runoff from line no =',k endif endif 28 xsum=xsum+hrf(i) write(3,100) iyr(i),im(i),id(i),ih(i),sf(i),hrf(i) +,x(i),y(i),z(i),u(i),v(i) i=i+1 go to 30 21 continue c q(i)=LOG10(ABS(sf(i))) c bf(i)=LOG10(ABS(u(i))) i=1 33 if (sf(i+1)-sf(i) .lt. (0.00277*A)) then v(i+1)=0 i=i+1 go to 33 else 34 if( v(i+1) .gt. 0 .or. v(i+2) .gt. 0 .or. v(i+3) .gt. 0 +.or. v(i+4) .gt. 0 .or. v(i+5) .gt. 0 + .or. v(i+6) .gt. 0) then xsum1=xsum1+v(i+1) i=i+1 go to 34 else xsum1=xsum1+v(i+1) endif endif i=1 45 if (sf(i+1)-sf(i) .lt. (0.00277*A))then i=i+1 go to 45 else j=1 55 if (hrf(i+1) .eq. -99.000)then hrf(i)=j else if(hrf(i+1) .gt. 0 .or. hrf(i+2) .gt. 0 .or. hrf(i+3) .gt. 0 +.or. hrf(i+4) .gt. 0 .or. hrf(i+5) .gt. 0 + .or. hrf(i+6) .gt. 0) then hrf(i+1)=j
Modelling losses in flood estimation D-3
j=j+1 i=i+1 go to 55 else hrf(i+1)=j c continue endif endif endif i=1 11 if (sf(i+1)-sf(i) .lt. (0.00277*A))then i=i+1 go to 11 else j1=1 12 if (hrf(i+1) .eq. -99.000)then hrf(i)=j1 else if(v(i+1) .gt. 0 .or. v(i+2) .gt. 0 .or. v(i+3) .gt. 0 +.or. v(i+4) .gt. 0 .or. v(i+5) .gt. 0 + .or. v(i+6) .gt. 0) then v(i+1)=j1 j1=j1+1 i=i+1 go to 12 else v(i+1)=j1 c continue endif endif endif i=1 50 if (sf(i+1)-sf(i) .lt. (0.00277*A))then if (hrf(i+1) .eq. -99.)then il=il+hrf(i) go to 52 else il=il+hrf(i) i=i+1 go to 50 endif else il=il+hrf(i) endif c if there is no surface runoff then to calculate only Total c rainfall and initial loss
Modelling losses in flood estimation D-4
52 if (PL .lt. 0)then PL=0 else if (CL .lt. 0)then CL=0 else if (CL1 .lt. 0)then CL1=0 else continue endif endif endif if(xsum .eq. il .or. xsum .eq. 0 .or. (j-1) .eq. 0 + .or. (j1-1) .eq. 0)then write(3,*)'Total rainfall =',xsum c Total volume of rainfall = xsum*1000*A write(3,*)'initial loss =',il c Total volume of il = il*1000*A PL=0.0 CL=0.0 CL1=0.0 write(3,*)'continuing loss =',CL write(3,*)'proportional loss =',PL write(3,*)'Contining loos upto surface runoff =',CL1 c write (5,203)iyr(i),im(i),id(i),ih(i),il,CL,PL,ROC,(j-1),(j1-1), c + CL1 go to 97 else continue endif CL= ((xsum*1000*A)-(il*1000*A)-(xsum1*60*60))/(A*1000* (j-1)) PL= 1-(xsum1*60*60)/(xsum*1000*A-il*1000*A) ROC= xsum1*60*60/(xsum*1000*A) CL1=((xsum*1000*A)-(il*1000*A)-(xsum1*60*60))/(A*1000* (j1-1)) if (PL .lt. 0 .or. CL .lt. 0 .or. CL1 .lt. 0)then c write (5,203)iyr(i),im(i),id(i),ih(i),il,CL,PL,ROC,(j-1),(j1-1), c + CL1 go to 97 endif write(3,*)'Total rainfall is =',xsum c Total volume of rainfall = xsum*1000*A write(3,*)'Initial loss=',il
Modelling losses in flood estimation D-5
c Total volume of il = il*1000*A write(3,*)'Surface runoff=',xsum1*60*60 c Total volume of surface runoff = xsum1*60*60 m3 write(3,*)'remainder of the rain rr=',j-1 write(3,*)'Continuing loss CL=',CL write(3,*)'surface runoff finished after =',j1-1 write(3,*)'Continuing loss CL1=',CL1 write(3,*)'Proportional loss PL=',PL write(3,*)'Volumetric r.c. =',ROC c pause go to 137 97 if(endfil .eq. -1111.000)then go to 200 else xsum=0 xsum1=0 il=0 i=i+2 i=1 k=1 go to 30 endif 137 write (5,203)iyr(i),im(i),id(i),ih(i),il,CL,PL,ROC,(j-1),(j1-1), + CL1 go to 97 200 continue close(2) close(3) close(5) 100 format(i4,i3,i3,i3,2f8.3,5f8.3) 201 format(a4,3a3,2a8,5a8) 202 format(a4,a3,a3,a3,a8,a8,a8,a8,a6,a6,a10,a20) 203 format(i4,3i3,4f8.2,2i6,f8.2) stop end
Modelling losses in flood estimation D-6
APPENDIX E PL = Proportional lossVRC = Volumetric runoff coefficient
Year month date hour IL(mm) CL(mm/h) PL VRC1968 2 6 6 84.47 0.08 0.46 0.09 PASCOE RIVER1970 3 28 12 75.26 0.39 0.87 0.021970 11 27 7 169.64 0.49 0.86 0.011971 4 2 9 114.15 3.12 0.96 0.011971 4 21 12 98.46 10.63 0.86 0.021972 4 12 18 84.63 3.45 0.77 0.171973 2 25 17 84.78 2.53 0.88 0.051973 3 13 2 24.35 0.84 0.83 0.111973 3 17 20 33.9 1.62 0.58 0.31973 3 20 23 1.27 5.22 0.7 0.31974 2 15 14 14.51 2.25 0.48 0.361975 2 11 18 10.71 12.85 0.9 0.091975 4 3 9 25.63 13.67 0.59 0.261975 4 4 15 38.34 2.83 0.7 0.231975 4 7 15 1.53 2.56 0.71 0.291975 12 21 17 37.61 0.4 0.97 01976 1 2 16 37.54 1.65 0.61 0.031976 1 5 1 4.54 2.01 0.9 0.091976 1 18 15 30.65 1.24 0.14 0.21976 2 18 6 28.69 0.87 0.46 0.291976 2 19 18 2.73 0.81 0.2 0.761976 4 8 16 0.63 8.35 0.81 0.191976 4 9 9 19.37 3.81 0.58 0.341976 4 10 9 8.14 1.43 0.24 0.611977 1 25 22 60.67 1.91 0.94 0.031977 3 15 19 38.34 0.37 0.17 0.321977 4 7 3 63.22 2.74 0.91 0.041977 4 13 20 50.83 8.95 0.97 0.031977 5 15 23 60.01 0.05 0.91 0.021978 2 20 14 50.33 1.86 0.84 0.11978 3 11 16 79.54 4.25 0.87 0.041978 4 1 15 48.51 1.67 0.94 0.041978 4 5 19 5.8 6.77 0.8 0.181978 4 7 2 86.84 0.5 0.37 0.181978 5 5 8 112.62 0.08 0.65 0.021979 1 7 16 28.8 2.72 0.91 0.071979 1 9 22 48.13 1.61 0.55 0.361979 2 17 8 64.63 0.47 0.88 0.021980 2 14 14 0.83 1.53 0.97 0.031980 3 15 1 11.88 8.73 0.85 0.131980 3 16 9 8.44 1.18 0.77 0.21980 3 20 14 50.79 0.08 0.08 0.241981 1 26 10 0.2 2.8 0.84 0.161981 3 4 15 1.1 7.46 0.89 0.11981 3 31 7 96.3 0.67 0.91 0.021982 1 10 20 43.86 1.4 0.97 0.011982 2 3 2 66.56 2.44 0.47 0.391982 3 20 14 3.29 3.71 0.71 0.281983 5 5 9 112.62 0.23 0.87 0.021984 2 11 19 44.15 0.95 0.9 0.021984 2 15 0 22.95 1.11 0.84 0.121984 2 23 16 30.85 1 0.31 0.261984 2 24 15 1.14 3.88 0.83 0.171984 3 2 15 23.59 4 0.8 0.121984 3 27 21 29.12 2.6 0.78 0.131985 3 20 23 17.87 3.15 0.91 0.061986 1 8 17 96.43 0.6 0.99 01986 1 14 8 52.89 4.03 0.94 0.011986 1 26 15 80.72 2.14 0.63 0.011986 3 20 3 94.02 0.11 0.9 0.01
Modelling losses in flood estimation E-1
Year month date hour IL(mm) CL(mm/h) PL VRC1987 1 31 18 63.25 0.58 0.84 0.011988 3 9 4 62.02 5.56 0.95 0.041988 12 26 20 22.47 1.67 0.93 0.041989 4 6 16 14.22 2.91 0.86 0.111989 4 21 0 50.57 1.95 0.73 0.11990 1 23 9 174.88 0.52 0.31 0.11990 3 3 10 6.78 12.48 0.95 0.041990 5 3 0 49.12 0.05 0.15 0.191992 1 10 19 39 0.51 0.26 0.451992 3 2 0 52.05 1.71 0.41 0.041992 4 6 5 18.52 3.56 0.96 0.031992 12 22 0 36.77 2.41 0.89 0.11993 1 24 6 68.45 2.82 0.8 0.141994 1 30 6 123.24 0.59 0.9 0.011994 2 17 21 19.17 7.12 0.83 0.131994 2 23 20 39.29 1.15 0.64 0.211995 2 24 22 58.87 6.95 0.98 0.011995 3 16 21 12.83 7.73 0.96 0.031996 1 16 19 3.4 1.37 0.45 0.511996 2 14 14 24.88 3.33 0.8 0.151996 2 15 17 16.57 7.69 0.56 0.41996 3 10 11 6.06 2.37 0.69 0.311996 4 5 10 14.32 1.2 0.54 0.351997 2 14 1 0.64 1.18 0.94 0.051997 2 21 22 5.81 6.38 0.93 0.071997 2 23 1 0.86 0.14 0.09 0.91998 1 17 20 74.93 0.19 0.64 0.021998 2 2 8 71.63 0.17 0.83 01998 3 14 7 71.99 0.66 0.99 0.011998 3 19 3 118.49 0.67 0.37 0.091998 3 20 15 37.13 5.27 0.83 0.121999 11 25 16 87.77 8.29 1 02000 3 2 18 2.39 5.62 0.72 0.271971 4 10 20 78.3 5.88 0.59 0.2 STEWART RIVER1972 4 13 20 73.71 0.94 0.22 0.461973 2 22 18 3.89 6.59 0.96 0.031973 3 5 20 66.46 1.22 0.25 0.31973 3 8 13 13.23 1.42 0.42 0.481973 3 23 0 34.87 2.06 0.57 0.31973 11 21 20 42.17 0.9 0.24 0.271997 2 1 16 10.92 0.92 0.99 0.01 EAST NORMAN RIVER1997 3 3 17 18.39 1.43 0.95 0.041997 4 30 15 11.92 1.3 1 01998 3 17 21 16.35 3.07 0.94 0.06 ENDEAVOUR RIVER1998 12 31 15 114.03 0.57 0.9 0.021999 2 1 18 71.77 0.71 0.95 0.031997 2 1 21 12.24 0.94 1 0 ANNA RIVER1997 3 3 16 14 1.49 0.97 0.021997 3 21 23 36.46 2.63 0.9 0.091964 10 12 12 12.6 3.1 0.98 0.02 NORTH JOHNSTON RIVER1971 2 16 18 36.15 2.64 0.92 0.061972 1 8 5 108.55 1.7 0.98 0.011972 3 17 16 83.31 1.91 0.76 0.121973 2 9 16 63.49 0.3 0.59 0.081973 3 4 0 26.92 6.07 0.88 0.111973 12 12 8 36.21 2.69 0.6 0.271974 1 2 17 10.73 6.62 0.58 0.341974 2 10 14 21.75 2.15 0.66 0.241974 3 5 21 34.04 7.79 0.76 0.211974 3 22 4 26.08 5.08 0.79 0.171974 3 24 14 3.3 2.72 0.52 0.451974 3 27 23 11.52 7.29 0.77 0.21975 2 19 23 55.93 0.44 0.09 0.141975 3 6 2 59.66 1.65 0.85 0.071982 1 29 0 41.66 4.48 0.86 0.09 SOUTH JOHNSTON RIVER
Modelling losses in flood estimation E-2
Year month date hour IL(mm) CL(mm/h) PL VRC1985 1 16 4 112.72 3.34 0.77 0.111985 2 13 19 31.52 2.68 0.88 0.111973 5 4 22 38.15 8.14 0.74 0.19 MURRAY RIVER1973 12 9 17 70.95 4.83 0.95 0.041975 1 15 23 51.45 5.66 0.95 0.041975 1 31 0 110.27 2.92 0.98 0.011975 2 8 22 38.51 5.71 0.88 0.071976 12 19 23 95.42 6.77 0.99 0.011977 1 31 3 42.94 8.44 0.84 0.131977 5 15 20 159.22 1.67 0.71 0.021978 1 31 5 104.1 0.92 0.86 0.031978 4 8 4 72.39 5.53 0.96 0.031979 1 25 9 18.85 5.14 0.9 0.091979 2 12 22 130.63 3.84 0.95 0.021979 12 29 1 69.94 2.52 1 01980 1 5 1 79.22 3.16 0.99 0.011984 2 8 22 97.19 4.74 0.88 0.081985 2 25 19 1.6 7.92 0.97 0.031988 12 27 20 18.02 2.23 0.88 0.111989 3 20 23 71.75 0.06 0.04 0.551992 2 25 13 25.08 7.93 0.93 0.061995 11 21 21 35.6 0.05 0.66 0.011997 3 3 14 2.07 5.13 0.62 0.371997 3 6 11 63.86 2.54 0.79 0.11998 12 30 18 65.75 0.16 0.88 01973 3 4 15 8.19 4.53 0.8 0.19 GOWRIE CREEK1973 12 9 8 14.86 2.63 0.9 0.091974 1 19 6 6.35 2.37 0.8 0.191974 2 3 21 0.5 2.69 0.21 0.791974 2 15 8 15.2 0.6 0.23 0.681974 3 3 1 73.6 1.99 0.64 0.181975 1 16 8 87.82 1.96 0.31 0.271975 12 11 3 0.94 1.76 0.4 0.591975 12 27 3 13.95 4.61 0.73 0.261976 2 6 22 4.97 6.07 0.83 0.161976 3 11 22 4.08 4.75 0.85 0.141976 4 10 8 2.6 2.89 0.72 0.271977 1 31 8 42.08 4.35 0.44 0.421977 2 5 21 7.65 2.5 0.55 0.441977 4 9 8 2.1 1.71 0.53 0.471978 2 27 9 1.03 4.17 0.42 0.571979 1 25 8 32.5 4.02 0.57 0.371979 2 1 16 7.71 1.6 0.29 0.651979 3 10 2 8.86 3.25 0.74 0.251979 12 29 2 66.29 2.22 0.79 0.131980 1 4 4 2.98 6.22 0.92 0.081980 3 18 6 67.35 2.18 0.34 0.451981 2 24 13 0.68 0.3 0.14 0.861981 5 20 0 7.26 1.83 0.62 0.351981 11 29 13 1.54 4.79 0.3 0.671982 1 28 8 104.23 2.92 0.75 0.11982 3 17 2 71.21 1.32 0.72 0.21982 4 14 18 40.7 1.76 0.38 0.491983 3 6 7 122.03 3.79 0.8 0.091983 3 8 6 47.11 8.82 0.86 0.131983 4 26 20 34.04 1.42 0.8 0.141985 2 5 11 21.74 2.61 0.95 0.041985 2 15 8 94.74 1.82 0.83 0.071985 3 24 17 24.62 1.56 0.62 0.251985 4 21 1 40.97 3.99 0.98 0.021987 3 26 11 107.46 1.62 0.29 0.091988 12 12 23 18.54 2.89 0.76 0.211988 12 28 22 6.78 3.97 0.7 0.291989 4 5 18 53.31 6.33 0.91 0.081989 11 22 12 6.44 3.04 0.5 0.47
Modelling losses in flood estimation E-3
Year month date hour IL(mm) CL(mm/h) PL VRC1991 1 31 1 28.69 1.33 0.51 0.411991 2 11 1 0.3 3.08 0.19 0.811991 2 16 12 10.41 0.25 0.03 0.941992 12 19 18 83.24 2.32 0.98 0.011993 1 17 0 76.82 1.33 0.79 0.081993 1 30 7 155.01 1.41 0.64 0.051995 2 28 18 0.33 3.24 0.88 0.111996 1 7 5 54 0.41 0.95 0.011997 3 3 16 4.32 6.33 0.68 0.311997 12 29 14 45.71 4.22 0.73 0.181998 1 12 18 0.29 0.57 0.19 0.811998 1 26 19 46.88 3.7 0.53 0.181998 2 26 9 24.64 2.18 0.93 0.051998 12 10 19 29.17 9.81 0.99 0.011999 1 19 23 24.54 8.24 0.91 0.071999 2 11 10 105.59 2.7 0.42 0.481999 2 27 10 115.19 3.03 0.88 0.061999 11 18 22 0.81 10.29 0.78 0.222000 3 10 1 73.93 0.01 0 0.262000 3 16 13 3.71 2.5 0.63 0.372000 4 24 16 2.64 2.05 0.74 0.261968 2 12 15 15.56 0.11 0.06 0.75 BLUNDER CREEK1968 3 28 12 115.84 0.42 0.74 0.031969 2 19 12 71.87 2.66 0.69 0.211969 3 23 12 87.5 0.74 0.82 0.11970 4 19 1 78.25 0.68 0.84 0.041971 2 16 9 48.48 2 0.83 0.121971 3 10 22 116.25 1.34 0.81 0.11971 3 16 18 68.74 11.3 0.9 0.061971 3 30 9 50.87 1.87 0.74 0.21971 4 10 3 33.72 1.5 0.84 0.151972 3 17 15 13.26 3.78 0.78 0.21972 3 27 3 69.26 0.23 0.26 0.121972 5 5 12 68.54 1.75 0.91 0.051973 2 10 13 20.07 8.77 0.99 01973 2 21 9 35.06 0.07 0.44 0.041973 3 4 7 128.08 1.41 0.81 0.131973 4 6 6 73.34 0.59 0.87 0.061973 4 13 3 107.22 1.25 0.51 0.31973 12 10 14 124.7 1.73 0.83 0.121974 1 30 14 24.81 1.41 0.69 0.221974 2 5 16 1.92 3.26 0.61 0.381974 2 10 10 35.18 2.52 0.69 0.21974 3 17 4 90.83 1.37 0.65 0.231975 2 8 23 75.78 5.1 0.77 0.091975 3 6 9 18.16 1.29 0.41 0.541975 7 30 17 157.61 0.63 0.99 01975 9 8 6 68.66 0.81 0.96 0.021976 2 16 13 38.16 1.9 0.51 0.291976 2 24 16 49.43 1.57 0.64 0.211976 11 19 17 42.36 9.38 0.97 0.011977 1 31 14 137.98 2.67 0.85 0.021977 4 6 18 121.68 4.49 0.87 0.041977 4 13 14 159 0.5 0.72 0.051979 1 2 5 189.37 8.09 0.83 0.141979 1 11 6 60.96 1.91 0.67 0.241979 1 26 2 71.8 1.28 0.74 0.011979 1 29 1 17.78 2.21 0.83 0.141979 2 22 1 113.72 0.26 0.17 0.341979 3 5 17 19.7 6.9 0.98 0.011979 3 6 18 44.31 0.55 0.69 0.031979 7 17 10 62.51 0.74 0.81 0.011981 1 7 16 99.27 2.03 0.91 0.071981 2 6 17 34.27 1.25 0.73 0.111981 7 20 17 107.15 0.26 0.96 0.01
Modelling losses in flood estimation E-4
Year month date hour IL(mm) CL(mm/h) PL VRC1981 9 15 7 78.28 0.66 0.85 0.031982 3 18 10 84.05 0.82 0.94 0.041982 4 1 15 90.19 1.63 0.82 0.121982 4 14 14 93.68 1.56 0.8 0.141973 2 6 3 51.09 1.99 0.89 0.04 STONE RIVER 1973 3 4 20 11.59 3.55 0.61 0.361973 12 12 10 114.68 0.54 0.93 0.021974 1 3 12 2.72 13.79 0.92 0.081974 3 2 23 23.77 1.89 0.83 0.141975 1 31 19 33.31 9.2 0.95 0.031975 3 29 10 31.21 4.49 0.92 0.051976 1 14 15 43.37 2.51 0.88 0.041977 1 31 0 18.32 3.9 0.89 0.091977 2 4 14 19.3 2.25 0.49 0.371977 3 3 22 7.73 3.53 0.72 0.271977 4 5 0 71.86 5.69 0.87 0.011977 5 15 3 42.84 2.64 0.74 0.181977 12 30 9 68.05 0.18 0.34 0.031978 2 27 18 80.33 4.34 0.87 0.081978 4 4 21 44.93 4.98 0.8 0.091979 2 1 16 0.26 6.34 0.95 0.051980 1 4 14 24.27 2.44 0.93 0.051980 3 18 13 53.62 6.88 0.87 0.081981 1 9 16 90.36 1.93 0.95 0.021981 2 3 16 1.61 0.39 0.17 0.811981 2 10 14 16.92 0.15 0.03 0.821981 2 24 10 11.86 1.65 0.74 0.231981 5 20 21 67.64 0.87 0.3 0.31982 2 21 10 0.51 3.41 0.97 0.021986 1 22 7 48 2.6 0.54 0.31986 2 27 22 1.14 2.89 0.64 0.351988 2 13 21 50.61 0.59 0.98 01989 3 6 3 62.01 0.78 0.87 0.011989 3 12 1 56.48 2.42 0.93 0.041989 3 20 23 71.36 2.61 0.65 0.21989 4 5 21 9.79 1.26 0.53 0.441990 3 20 19 85.39 2.03 0.67 0.191990 4 9 4 32.51 1 0.48 0.341990 4 17 13 23.22 2.8 0.63 0.321990 6 5 15 43.46 1.22 0.6 0.231991 2 1 2 7.03 0.09 0.03 0.911991 2 18 21 11.36 2.75 0.95 0.051992 2 26 18 31.89 4.79 0.84 0.141993 1 29 19 46.06 2.31 0.84 0.111994 2 21 2 30.25 14.52 0.97 0.031996 3 4 16 122.99 0.79 0.99 01997 3 22 18 13.29 4.12 0.53 0.451997 12 29 16 58.44 1.46 0.82 0.051998 1 8 11 19.19 6.1 0.77 0.221998 5 14 17 51.28 2.32 0.9 0.061998 11 24 18 33.23 3.39 0.29 0.291999 1 5 20 24.13 4.51 0.93 0.061999 11 4 6 9.72 2.9 0.99 0.011999 11 20 8 61.39 1.17 0.9 0.021999 11 27 1 69.98 1.95 0.84 0.081999 12 26 16 61.31 1.29 0.61 0.12000 2 6 0 22.04 5.28 0.86 0.132000 3 16 20 15.74 6.04 0.75 0.242000 4 3 12 46.03 2.54 0.7 0.181986 1 22 5 42.35 3.25 0.9 0.07 BOHLE RIVER1987 1 17 14 28.25 0.66 0.88 0.011987 12 25 16 27.12 0.7 0.96 01988 2 28 1 37.71 1.46 0.89 0.071988 12 10 3 0.2 4.17 0.42 0.581988 12 11 19 0.63 7.63 0.48 0.51
Modelling losses in flood estimation E-5
Year month date hour IL(mm) CL(mm/h) PL VRC1988 12 27 21 0.7 3.35 0.53 0.471989 2 3 20 41.14 2.64 0.34 0.371989 3 20 18 17.53 2.76 0.7 0.261989 12 15 18 29.16 2.1 0.88 0.091990 3 21 19 38.58 4.54 0.64 0.191990 4 18 9 46.43 4.32 0.56 0.271990 6 5 14 31.99 2.37 0.78 0.151990 12 30 13 0.11 2.05 0.27 0.721991 2 2 16 10.59 1.2 0.3 0.671991 2 16 0 10.69 1.19 0.36 0.61991 2 18 22 46.01 2.34 0.24 0.631992 2 28 20 1.47 1.75 0.14 0.851994 1 30 23 28.44 1.99 0.68 0.261995 2 9 17 18.55 3.23 0.6 0.331996 1 6 7 93.2 2.17 0.68 0.21997 12 22 4 61.88 5.09 0.82 0.131998 2 25 9 34.41 1.44 0.96 0.031999 2 27 17 39.57 1.64 0.43 0.371981 1 16 1 10.27 1.16 0.7 0.27 MAJOR CREEK 1981 1 20 3 15.45 1.2 0.49 0.441981 2 22 22 55.24 1.13 0.72 0.021981 5 21 2 79.87 0.33 0.17 0.191974 1 1 21 21.47 2.07 0.98 0.01 BROUGHTON RIVER1974 1 13 19 11.44 1.63 0.77 0.151974 1 14 14 8.81 2.09 0.72 0.221974 1 15 17 3.86 2.06 0.81 0.181974 4 26 2 23.9 5.17 0.6 0.211975 1 10 10 15.42 0.48 0.08 0.391975 1 12 16 2 2.16 0.65 0.331975 10 23 14 36.49 1.14 0.99 01975 11 8 23 21.24 2.7 0.7 0.111977 5 15 13 3.52 3.26 0.88 0.111978 1 31 13 71 0.29 0.48 01979 2 1 18 14.1 8.39 0.96 0.031980 2 24 16 18.42 2.92 0.97 0.021980 2 29 17 32.44 1.41 0.94 0.011980 3 3 21 12.76 1.86 0.88 0.091980 3 5 15 10.57 2.48 0.92 0.051980 4 23 21 36.11 0.16 0.73 0.011981 4 9 2 58.28 1.54 0.98 01981 5 21 2 67.27 1.35 0.88 0.021993 1 11 6 123.37 2.38 0.95 0.02 BROKEN RIVER1993 2 15 17 99.8 1.08 0.8 0.071994 2 1 17 85.73 0.86 0.37 0.221994 3 4 0 40.4 0.67 0.87 0.041998 8 28 21 82.79 9.11 0.65 0.291999 1 14 14 31.87 0.56 0.42 0.121999 2 10 19 69.1 0.74 0.78 0.071999 2 14 15 32.4 1.7 0.71 0.231999 12 25 17 39.44 2.71 0.84 0.132000 4 2 6 64.26 4.37 0.55 0.362000 5 3 7 29.35 2.78 0.87 0.111973 12 18 6 53.71 4.72 0.62 0.34 ST. HELENS CREEK1974 1 1 3 28.17 0.7 0.22 0.691974 1 17 11 11.56 3.43 0.68 0.311974 11 15 22 113.48 1.51 0.93 0.021975 1 14 23 14.21 6.04 0.71 0.281975 3 5 15 66.52 4.38 0.8 0.151975 9 10 5 154.62 0.47 0.82 01975 12 10 11 122.32 3.98 0.64 0.271976 1 28 18 46.75 1.62 0.59 0.31976 5 14 16 113.87 1.21 0.97 0.011977 1 19 22 18.54 0.33 0.18 0.551970 1 18 14 99.56 13.11 0.9 0.08 BLACKS CREEK1970 2 4 15 70.42 1.84 0.9 0.06
Modelling losses in flood estimation E-6
Year month date hour IL(mm) CL(mm/h) PL VRC1971 2 20 16 68.73 8.49 0.89 0.071971 3 6 18 64.66 2.95 0.86 0.091971 12 24 3 143.57 4.78 0.87 0.071972 1 1 22 29.86 12.33 0.99 01972 1 8 4 7.2 2.43 0.79 0.191973 12 18 14 8.35 4.02 0.94 0.061974 1 3 0 53.63 3.65 0.99 0.011974 1 5 16 0.8 2.7 0.95 0.051974 1 14 9 4.78 3.55 0.96 0.041974 1 17 17 2.27 13.54 0.93 0.071974 1 29 17 38.29 3.75 0.95 0.031974 2 16 18 20.7 2.49 0.99 0.011974 2 28 16 32.45 3.15 0.87 0.121975 12 10 20 138.58 4.67 0.97 0.021975 12 22 1 110.46 0.46 0.44 0.051976 2 6 5 42.99 1.34 0.97 0.021976 3 3 19 36.79 0.82 0.5 0.091976 3 4 22 8.07 5.42 0.62 0.361977 2 1 0 76.99 4.11 0.88 0.071977 3 7 17 11.29 4.56 0.93 0.061978 1 30 20 144.39 3.94 0.96 0.031979 1 26 8 57.63 3.08 0.98 0.021979 2 13 5 36.7 4.37 0.85 0.111979 3 8 19 81.3 1.71 0.97 0.021980 1 6 10 62.29 15.39 0.98 0.021980 2 10 14 117.95 1.65 0.86 0.021981 1 19 23 66.13 2.23 0.91 0.051981 5 21 3 77.85 5.05 0.97 0.011983 3 5 22 125.2 0.22 0.98 01983 5 20 5 39.7 1.75 0.95 0.041983 11 17 3 70.6 0.23 0.97 01984 1 14 23 40.07 1.73 0.96 0.031985 3 12 15 80.14 2.19 0.96 0.031972 5 10 23 31.85 0.56 0.47 0.11 SANDE CREEK1973 1 14 0 46.72 0.18 0.18 0.021973 12 19 12 19.65 8.97 0.89 0.11974 1 8 21 3.84 3.64 0.87 0.111974 12 28 16 26.46 4.67 0.96 0.031975 1 2 14 13.52 2.51 0.82 0.161975 2 25 9 23 2.66 0.82 0.131975 3 18 15 29.67 2.71 0.39 0.331975 12 21 14 39.53 2.46 0.87 0.091976 2 2 15 12.78 1.57 0.35 0.51976 10 31 20 6.12 8.99 0.75 0.211977 2 24 23 29.02 3.41 0.93 0.031978 1 31 16 97.04 2.69 0.7 0.21979 3 10 17 85.67 1.56 0.92 0.031967 6 21 14 48.2 0.24 0.92 0.01 MONAL CREEK1968 1 13 10 3.25 1.21 0.93 0.071969 12 7 18 26.15 0.85 0.88 01970 12 9 9 3.03 9.12 1 01971 1 22 18 2.71 6.35 0.86 0.131971 1 30 8 46.28 0.45 0.24 0.361971 2 3 16 10.79 2.57 0.27 0.531971 2 4 15 7.56 1.2 0.13 0.681971 2 7 19 15.36 1.27 0.34 0.491971 2 8 18 4.34 1.24 0.41 0.511971 2 21 11 25.66 1.03 0.65 0.261971 11 29 16 39.24 0.18 0.76 01973 7 8 9 40.58 4.6 0.86 0.1 GREGORY RIVER1973 12 20 14 123.05 5.27 0.83 0.051974 1 26 20 33.42 1.1 0.33 0.571974 11 24 22 50.94 0.12 0.96 01976 1 19 8 19.81 2.48 0.63 0.291976 2 22 5 26.19 1.86 0.59 0.27
Modelling losses in flood estimation E-7
Year month date hour IL(mm) CL(mm/h) PL VRC1976 3 5 7 3.57 2.64 0.54 0.441977 11 18 15 26.07 28.74 0.98 0.011967 1 28 21 37.51 1.02 0.38 0.44 MARY RIVER1967 6 26 6 15.13 0.42 0.15 0.681968 1 9 0 126.09 1.68 0.38 0.521971 2 18 6 57.86 0.67 0.19 0.651971 11 29 7 72.09 0.53 0.99 01971 12 27 17 112.21 0.77 0.56 0.151972 1 25 19 24.86 3.06 0.76 0.061972 2 18 23 16.88 0.48 0.21 0.61972 4 2 5 1.13 1.12 0.13 0.871972 11 11 14 35.73 2.62 0.41 0.281973 2 11 23 52.36 3.72 0.49 0.121973 2 23 17 32.33 2.01 0.53 0.151973 7 4 10 0.6 1.63 0.39 0.611974 1 12 12 8.41 0.89 0.77 0.21974 1 24 12 49.04 3.3 0.61 0.351974 3 9 8 23.66 0.81 0.42 0.451974 11 18 9 52.51 0.84 0.51 0.31976 2 11 18 21.42 4.01 0.32 0.531976 5 22 9 2.74 0.27 0.14 0.831978 3 31 22 21.56 2.1 0.88 0.081978 11 10 16 29.95 0.1 0.42 0.071981 2 14 23 56.45 0.63 0.24 0.51981 5 21 16 27 2.21 0.69 0.211983 6 21 9 1.98 0.24 0.04 0.95 NORTH MAROOCHY RIVER1983 11 21 4 57.23 0.62 0.29 0.451986 11 30 0 17.82 1.02 0.44 0.481987 10 16 15 68.8 3.71 0.68 0.161987 12 1 3 58.14 0.52 0.19 0.381988 4 4 19 13.47 2.37 0.71 0.271988 7 2 17 20.57 0.76 0.32 0.621989 2 22 3 21.71 2.24 0.44 0.31989 3 30 16 104.36 2.4 0.4 0.51989 8 18 9 1.52 1.59 0.44 0.551997 3 5 14 52.16 1.08 0.27 0.611997 5 1 8 113.26 0.23 0.94 0.021997 10 8 2 42.56 0.27 0.74 0.021997 10 20 0 42.87 1.27 0.9 0.041997 12 28 21 65.08 0.16 0.84 0.031998 1 28 14 93.32 0.42 0.44 0.241998 12 31 15 35.76 0.61 0.85 0.111999 2 7 11 14.13 1.11 0.33 0.651999 2 28 10 61.94 1.1 0.47 0.411999 3 18 7 18.23 0.43 0.69 0.271999 5 9 14 41.98 2.8 0.76 0.191999 5 18 19 14.52 0.7 0.46 0.51969 8 27 1 62.4 0.93 0.64 0.15 BREMER RIVER1970 12 7 3 62.29 1.83 0.57 0.251971 1 31 1 23.18 1.39 0.41 0.541971 2 3 11 0.24 2.63 0.6 0.41972 10 19 0 39.13 12.55 0.97 0.021972 10 28 1 25.75 0.2 0.13 0.611973 2 17 12 10.65 0.85 0.61 0.311974 1 9 23 25.99 0.74 0.47 0.351974 11 19 15 14.14 2.47 0.72 0.211975 2 24 22 116.98 1.81 0.71 0.091975 10 20 3 72.35 1.06 0.61 0.081975 12 19 20 45.31 0.32 0.27 0.521976 1 20 12 28.23 0.19 0.11 0.581977 5 10 22 35.82 1.95 0.39 0.181979 11 24 17 4.95 2.36 0.28 0.611980 2 5 21 49.01 0.09 0.25 0.051980 5 9 12 111 0.05 0.13 0.041981 2 7 6 70.38 1.43 0.4 0.33
Modelling losses in flood estimation E-8
Year month date hour IL(mm) CL(mm/h) PL VRC1983 1 3 6 42.8 1.33 0.75 0.031983 5 1 2 23.2 1 0.45 0.461983 5 27 11 23.25 0.63 0.14 0.681983 6 18 15 28.17 1.17 0.48 0.211983 6 21 11 15.39 0.65 0.27 0.611983 11 27 19 0.4 1.25 0.6 0.41984 7 26 16 30.89 0.69 0.32 0.41984 11 5 13 9.95 2.86 0.52 0.351986 11 30 23 66.75 0.02 0.18 01987 1 27 20 39.16 3.69 0.82 0.091987 10 16 15 74.28 1.19 0.42 0.161987 12 1 15 23.92 0.88 0.26 0.391988 2 14 16 82.74 0.41 0.73 01989 4 2 10 85.89 2.48 0.61 0.181989 5 16 16 39.04 0.34 0.31 0.291989 12 5 13 55.23 2.79 0.95 0.011990 1 15 20 44.14 0.47 0.35 0.091990 2 3 4 55.01 3.88 0.99 01990 2 6 20 24.46 9.44 0.88 0.041969 1 1 21 99.85 0.09 0.93 0 TENHILL CREEK1971 2 7 15 21.1 0.19 0.74 0.011971 11 28 21 34.95 1.49 1 01976 2 10 17 58.28 1.27 0.48 0.281976 10 31 21 44.38 0.89 0.74 0.061979 6 21 8 89.19 0.64 0.7 0.041981 2 7 7 104.39 2.12 0.71 0.11981 12 3 14 20.16 1.26 0.87 0.051982 1 21 3 51.84 0.37 0.45 0.041982 2 24 17 75.96 0.48 0.9 01983 5 3 1 81.99 1.7 0.52 0.051983 5 27 18 29.6 2.68 0.84 0.091983 6 21 22 6.86 3.9 0.86 0.131985 1 23 18 28.33 1.49 0.98 0.011988 4 5 11 14.55 0.83 0.7 0.241988 7 5 8 27.51 0.07 0.05 0.541990 5 22 21 58.44 0.64 0.74 0.051990 5 28 0 42.57 1.44 0.88 0.011991 12 12 11 125.46 0.02 0.04 0.051992 2 9 21 15.41 7.58 0.94 0.061995 11 20 8 42.21 2.51 0.97 0.011996 1 10 5 53.69 0.09 0.25 0.021999 2 9 3 97.67 1.34 0.89 0.031999 3 3 17 35.68 0.01 0.1 0.011978 3 21 15 0.2 2.91 0.51 0.49 LOGAN RIVER1978 4 2 9 5.79 0.1 0.03 0.781978 12 29 16 7.51 2.73 0.35 0.521979 3 4 13 47.99 0.73 0.97 0.011979 6 20 20 62.74 1.12 0.51 0.231979 11 20 18 16.46 6.24 0.94 0.041979 12 15 22 13.98 2.46 0.36 0.481981 2 5 23 13.98 2.73 0.89 0.11981 11 2 22 10.9 0.84 0.45 0.431982 2 24 23 83.07 0.53 0.32 0.081983 3 20 23 43.64 3.08 0.91 0.041983 5 1 9 32.41 0.79 0.29 0.561983 7 29 4 34.02 1.18 0.76 0.091983 10 15 20 0.65 9.63 0.82 0.181983 11 21 6 35.52 0.57 0.37 0.131985 4 3 16 38.99 0.41 0.69 0.021986 10 1 16 47.49 3.09 0.92 0.031987 12 1 20 64.7 3.62 0.74 0.051988 4 5 7 6.5 0.15 0.09 0.821988 7 4 21 32.28 1.13 0.28 0.571989 3 26 3 22.89 0.6 0.7 0.191989 4 1 14 25.12 0.7 0.27 0.58
Modelling losses in flood estimation E-9
Year month date hour IL(mm) CL(mm/h) PL VRC1989 5 15 23 1.48 0.42 0.19 0.791989 11 9 20 72.79 3.95 0.56 0.071989 11 11 23 26.33 1.77 0.77 0.031989 12 5 1 11.83 0.97 0.3 0.571990 1 15 20 38.56 0.19 0.22 0.11990 2 7 21 9.8 3.56 0.93 0.051991 1 26 19 29.35 1.38 0.74 0.151991 2 7 1 69.39 18.31 0.36 0.511991 2 7 16 8.29 6.66 0.43 0.551991 2 8 17 22.67 0.07 0.13 0.131991 12 12 3 77.55 4.14 0.61 0.251995 1 20 22 54.75 0.38 0.46 0.041995 2 17 19 25.46 5 0.93 0.041995 3 9 19 40.62 0.35 0.82 01996 7 28 4 44.57 1.06 0.63 0.011996 8 30 18 49.56 0.8 0.7 0.041996 12 7 17 27.39 2.35 0.88 0.071997 2 14 19 99.01 1.81 0.44 0.311997 12 23 22 52.51 2.76 0.27 0.121998 1 30 13 17.89 2.29 0.97 0.031978 1 28 16 34.29 3.36 1 0 RUNNING CREEK1978 3 21 16 0.76 3.75 0.45 0.551978 11 5 22 43.86 0.15 0.27 0.051978 12 26 22 38.19 1.53 0.84 0.051979 3 4 20 49.31 0.09 0.76 0.011979 6 21 2 80.57 0.32 0.64 0.041979 11 20 21 37.22 8.1 0.95 0.011979 12 24 18 39.51 1.91 0.87 0.031980 12 31 18 36.57 0.41 0.81 0.041981 2 6 18 33.43 2.2 0.82 0.131981 4 4 19 34.15 5.44 0.98 0.011982 1 20 16 22.81 0.88 0.43 0.331983 5 1 16 34.72 1.88 0.6 0.311983 6 21 13 16.55 1.22 0.36 0.541983 11 16 3 46.63 0.17 0.63 0.021984 3 3 19 47.2 1.83 0.99 01984 4 8 14 14.33 10.17 0.42 0.511989 1 6 23 1.78 0.01 0.16 0.191989 12 24 19 0 0.03 0.66 0.341995 3 1 21 0.64 0.08 0.62 0.141978 3 21 19 29.41 2.99 0.77 0.13 TEVIOT BROOK1978 4 2 9 5.79 1.74 0.45 0.441978 12 29 16 7.51 6.42 0.82 0.141979 6 20 15 45.49 1.8 0.75 0.161979 11 19 19 32.62 0.18 0.64 01979 11 20 20 26.62 6.99 0.9 0.031979 12 15 20 13.73 4.62 0.99 0.011980 2 5 20 51.27 0.09 0.16 0.051980 5 28 16 34.18 0.71 0.28 0.131981 2 7 13 91.9 1.18 0.5 0.121981 4 4 17 25.81 2.1 0.89 0.051981 11 2 14 3.55 0.63 0.4 0.561982 12 31 21 31.72 0.7 0.94 01983 5 1 6 25.28 0.69 0.25 0.621983 10 15 23 65.99 3.39 0.79 0.041987 1 2 17 1.5 3.73 0.85 0.141987 1 27 22 31.86 0.8 0.69 0.061987 3 3 6 53.58 0.71 0.82 0.021988 4 3 18 26.46 1.01 0.15 0.631988 4 5 5 6.46 1.32 0.98 0.021988 7 4 21 32.28 0.47 0.12 0.71989 3 26 7 33.9 0.01 0.02 0.421989 4 1 16 29.98 0.14 0.07 0.661989 10 27 5 40 0.81 0.99 01990 2 7 21 9.8 3.59 0.94 0.04
Modelling losses in flood estimation E-10
Year month date hour IL(mm) CL(mm/h) PL VRC1990 4 20 11 5.42 0.75 0.3 0.621990 12 21 20 31.75 0.22 0.97 01991 2 7 16 8.29 5.89 0.43 0.551991 12 12 1 61.21 2.92 0.42 0.411991 12 31 21 31.87 0.05 0.35 01995 2 17 20 41.64 0.29 0.2 0.11995 12 5 23 27.04 0.98 0.66 0.031995 12 23 15 35.93 0.54 0.69 0.041997 2 14 18 81.43 2.32 0.57 0.261998 1 30 5 14.7 1.24 0.65 0.291999 1 9 19 27.13 2.41 0.72 0.121999 3 3 12 4.77 3.15 0.47 0.451973 2 14 12 11.27 0.93 0.76 0.17 ALBERT RIVER1973 7 6 15 55.76 2.59 0.7 0.241974 1 24 23 137.95 6.95 0.7 0.251974 3 10 14 43.46 0.99 0.57 0.31974 4 21 19 10.17 0.39 0.87 0.041975 1 8 14 14.06 1.48 0.44 0.491975 2 24 21 110.27 0.61 0.07 0.511976 2 10 23 3.13 0.26 0.11 0.861976 2 27 23 30.46 2.15 0.96 0.011977 3 2 13 0.59 0.62 0.99 0.011978 12 23 13 5.69 2.39 0.98 0.021978 12 26 4 51.48 0.69 0.27 0.351981 2 4 18 11.08 2.57 0.59 0.381981 2 6 23 36.89 2.99 0.95 0.031981 5 21 8 165.84 0.4 0.34 0.111981 11 3 12 16.11 2.08 0.89 0.041981 12 3 18 25.84 1.3 0.61 0.321982 1 19 0 6.63 2.17 0.95 0.041982 3 15 13 1.77 0.69 0.44 0.261982 12 13 22 56.6 0.34 0.19 0.151983 5 27 13 131.03 0.18 0.11 0.271983 6 21 16 35 1.14 0.47 0.331989 5 27 2 93.63 1.67 0.93 0.031990 3 27 20 51.67 2.6 0.58 0.311990 4 4 19 12.84 0.56 0.28 0.461990 5 22 9 35.1 1.07 0.51 0.281990 5 27 12 94.19 1.21 0.83 0.041990 6 7 4 165.02 4.36 0.83 0.071991 12 12 11 70.31 2.47 0.88 0.051992 3 16 12 58.75 0.17 0.73 0.021994 3 1 5 75.07 0.12 0.6 0.061995 11 20 20 81.15 0.19 0.3 0.211996 1 3 16 88.91 0.1 0.67 0.011997 5 5 11 59.23 0.36 0.92 01997 5 17 0 9.32 0.49 0.49 0.241978 3 17 3 12.44 2.83 0.84 0.15 BACK CREEK1978 12 26 21 3.6 0.74 0.71 0.21979 12 16 2 0 0.52 0.86 0.141980 5 6 19 8.13 2.45 0.58 0.411981 11 2 12 12.49 2.53 0.93 0.061982 10 10 13 0 2.92 0.98 0.021987 3 5 6 49.55 2.82 0.78 0.141989 4 3 14 0.42 2.39 0.56 0.431991 11 30 18 0 1.64 0.99 0.011995 2 14 18 6.08 1.62 0.78 0.21970 10 24 16 28.02 1.48 0.85 0.05 MACINTYRE BROOK1970 12 9 9 0.31 3.11 0.84 0.161974 10 30 22 25.1 0.55 0.97 0.011974 11 1 19 2.7 1.55 0.78 0.211974 11 18 23 28.91 1.98 0.76 0.121975 2 24 14 88.38 1.55 0.78 0.021975 3 17 17 0.79 2.26 0.87 0.131975 12 23 13 93.34 5.52 0.82 0.03
Modelling losses in flood estimation E-11
Year month date hour IL(mm) CL(mm/h) PL VRC1976 2 10 10 90.65 2.51 0.57 0.221978 9 6 21 54.99 0.69 0.86 0.031982 12 2 21 17.87 10.14 0.96 0.031983 5 1 10 31.52 1.4 0.61 0.291984 2 17 19 2.54 5.29 0.92 0.071984 4 8 16 88.11 1.5 0.46 0.071987 1 27 16 28.55 4.39 0.82 0.131988 4 3 16 19.11 2.36 0.67 0.211988 4 11 4 22.97 0.18 0.1 0.391989 3 26 11 47.25 1.09 0.78 0.121990 4 20 16 29.55 3.4 0.63 0.221991 2 7 20 35.65 0.59 0.72 0.111994 2 19 23 59.13 0.68 0.83 01995 2 17 19 7.55 3.91 0.97 0.021995 2 18 17 16.24 15.06 0.81 0.171995 10 26 4 5.28 4.57 0.98 0.021995 11 19 22 44.05 0.93 0.67 0.151996 1 9 0 47.18 0.15 0.21 0.221996 1 22 4 28.46 0.34 0.53 0.231996 5 2 12 83.54 2.8 0.7 0.131997 12 15 2 40.41 0.05 0.52 0.01 SPRING CREEK1998 7 24 21 0.5 1.76 0.99 0.011998 11 23 13 0.24 0.95 0.94 0.061999 1 28 14 5.94 0.33 0.94 0.051999 6 27 13 2.64 0.53 0.93 0.061999 11 6 22 16.06 0.93 0.99 0.011970 12 9 9 0.31 2.24 0.73 0.27 CANAL CREEK1972 10 28 13 0.07 3.56 0.94 0.061973 11 22 2 30.7 3.76 0.88 0.011974 11 1 18 3.05 1.64 0.73 0.251974 11 18 23 19.28 2.57 0.76 0.161975 2 24 14 104.29 2.56 0.81 0.021975 3 17 17 0.97 2.73 0.85 0.141975 12 23 13 116.41 6.84 0.81 0.031976 2 10 9 50.01 1.3 0.31 0.451978 4 2 11 46.26 0.39 0.92 01978 9 6 21 53.63 0.51 0.72 0.061982 12 2 21 14.03 6.12 0.92 0.051983 5 1 10 36.15 0.67 0.31 0.491983 5 27 15 12.54 1.32 0.39 0.451983 6 21 16 27.66 0.6 0.47 0.311984 4 8 16 85.01 0.18 0.07 0.11984 9 4 21 13.93 1.67 0.78 0.091987 1 27 16 9.24 0.37 0.21 0.551988 4 3 15 12.55 1.91 0.51 0.371988 4 11 4 26.66 1.1 0.29 0.41989 3 26 11 51.72 0.36 0.41 0.231989 12 22 3 27.25 0.13 0.19 0.281990 4 20 16 24.24 1.58 0.31 0.431990 5 22 14 22.72 0.85 0.51 0.251995 2 17 19 8.56 3.33 0.96 0.031996 1 9 0 17.71 0.29 0.32 0.391996 5 2 12 77.66 2.38 0.58 0.191978 11 5 23 30.98 0.36 0.94 0 CADAMINE RIVER1979 6 21 7 75.83 0.33 0.55 0.041980 2 5 19 39.38 0.56 0.99 01980 5 9 11 89.35 0.08 0.17 0.081981 12 3 10 8.99 3.2 0.87 0.091983 5 1 13 61.09 2.27 0.87 0.081983 5 27 16 20.58 0.51 0.46 0.371986 12 25 9 40.51 0.74 1 01987 5 15 11 20.84 1.17 0.79 0.141987 12 1 14 79.12 2.73 0.74 0.081988 4 1 3 85.87 1.27 0.8 0.121989 11 17 21 12.69 2.49 0.69 0.2
Modelling losses in flood estimation E-12
Year month date hour IL(mm) CL(mm/h) PL VRC1990 4 5 10 54.14 0.5 0.68 0.051990 4 20 21 44.63 0.92 0.65 0.061991 12 31 5 24.47 0.45 0.86 01995 11 20 17 85.14 0.46 0.76 0.021996 12 12 3 30.94 1.42 0.92 0.011997 11 3 17 38.83 0.14 0.28 0.031998 1 30 22 41.23 1.02 0.94 0.021998 7 28 8 23.04 2.02 0.95 0.011999 3 3 6 85.74 2.57 0.71 0.091974 1 19 16 1.45 2.8 0.83 0.17 PAROO CREEK1974 1 21 11 1.06 1.96 0.87 0.131974 2 3 14 25.64 0.64 0.63 0.131974 2 15 18 24.08 3.9 0.59 0.181974 2 16 22 2.23 1.51 0.81 0.181975 12 13 20 1.38 5.58 0.98 0.021971 3 3 4 46.23 1.06 0.6 0.28 GORGE CREEK1971 3 17 18 9.49 0.72 0.27 0.521971 4 12 7 30.31 0.69 0.55 0.231972 3 7 5 5.65 1.31 0.32 0.571973 3 27 19 6.2 1.46 0.97 0.031973 12 16 17 0.03 5.39 0.84 0.161974 1 31 15 0.62 2.01 0.55 0.451975 1 1 15 31.86 0.17 0.79 0.021975 2 14 12 4.51 0.73 0.93 0.061971 4 12 18 59.79 0.56 0.76 0.03 MALBON RIVER1972 3 6 5 21.58 7.05 0.81 0.141973 11 25 21 59.17 0.84 0.71 0.041973 12 16 20 34.21 14.61 0.96 0.021973 12 20 22 3.63 3.09 0.85 0.13 NORMAN RIVER1974 1 1 19 0.66 2.18 0.82 0.181974 1 3 20 16.61 4.36 0.83 0.121974 1 6 1 102.63 3.85 0.98 0.011974 1 30 18 43.18 3.46 0.93 0.041974 12 29 12 16.02 5.63 0.88 0.111975 1 16 19 14.46 3.2 0.6 0.31975 12 28 20 28.41 2.02 0.92 0.031977 2 3 14 37.69 0.54 0.56 0.061973 11 26 15 16.37 8.71 1 0 MOONLIGHT RIVER1973 12 25 21 33.83 0.45 0.29 0.151973 12 26 19 2.27 6.79 0.58 0.411974 1 2 9 36.62 2.4 0.64 0.081974 1 24 18 60.58 1.34 0.55 0.231974 3 9 1 20.46 0.62 0.16 0.551974 12 31 2 28.93 10.4 0.99 0.011973 11 11 23 31.22 1.77 0.8 0.03 AGATE CREEK1973 12 20 22 3.63 3.23 0.89 0.11973 12 25 13 3.29 3.71 0.95 0.051973 12 29 20 1.46 7.12 0.95 0.051974 1 2 0 6.37 2.65 0.81 0.161974 1 3 16 1.54 3.95 0.84 0.161974 1 15 18 0.93 2.83 0.77 0.221974 1 29 20 19.53 2 0.85 0.131974 2 2 17 32.28 0.23 0.94 01974 2 6 16 0.14 1.41 0.35 0.651974 2 15 20 6.11 1.04 0.54 0.391974 3 4 22 34.67 0.93 0.31 0.281974 12 29 6 2.8 4.68 0.89 0.111975 1 16 16 2.94 4.03 0.81 0.181975 2 23 0 5.54 2.89 0.9 0.081975 10 23 15 18.36 3.96 0.93 0.051975 12 19 17 14.12 1.09 0.99 0.011976 1 6 22 5.31 5.63 0.93 0.061977 1 20 2 10.7 2.23 0.9 0.081973 12 20 22 6.02 5.02 0.96 0.04 PERCY RIVER1974 1 3 17 1.32 3.04 0.9 0.09
Modelling losses in flood estimation E-13
Year month date hour IL(mm) CL(mm/h) PL VRC1974 1 5 23 3.27 4.96 0.93 0.061974 1 17 22 4.26 3.13 0.86 0.131974 1 29 15 24.85 1.01 0.39 0.51974 3 9 23 9.4 3.19 0.94 0.041975 1 14 23 10.72 1.27 0.8 0.161977 2 3 10 0.11 0.04 0.36 0.581973 11 24 22 37.14 0.25 1 0 ELIZABETH CREEK1974 1 3 16 13.9 4.03 0.97 0.031974 1 17 17 26.68 1.77 0.79 0.181974 1 20 17 42.09 1.92 0.8 0.171974 1 29 17 24.49 1.45 0.81 0.111974 2 22 15 12.23 2.15 0.65 0.231974 3 4 16 2.74 3.88 0.89 0.111975 12 26 16 16.22 2.17 0.98 0.011974 1 29 14 29.55 1.09 0.79 0.03 ROUTH RIVER1974 2 6 21 26.04 4.06 0.51 0.321974 3 20 22 6.57 3.96 0.85 0.131974 12 30 0 36.21 0.67 0.13 0.261975 1 14 22 7.9 3.83 0.77 0.191977 3 1 20 51.77 1.44 0.62 0.021978 12 13 23 35.19 1.16 0.89 0.051969 12 7 19 38.31 4.08 0.78 0.03 PALMER RIVER1972 1 7 22 55.86 1.79 0.91 0.051972 2 4 14 1.82 5.6 0.99 0.011972 2 15 23 53.07 0.08 0.09 0.081975 3 20 22 31.29 8.62 0.92 0.031974 1 23 9 5.19 5.8 0.83 0.16 NORTH PALMER RIVER 1974 1 30 19 25.28 2.57 0.82 0.071974 2 13 16 2.26 7.48 0.78 0.211974 2 15 19 14.51 10.95 0.91 0.071975 3 2 17 0.8 7.51 0.99 0.011975 3 31 21 46.19 0.3 0.85 0.011975 12 24 22 44.94 3.99 0.86 0.011970 12 31 0 87.05 0.26 0.92 0.01 HOLROYD RIVER1972 4 13 6 17.43 1.28 0.42 0.511973 2 8 16 62.04 7.92 0.97 0.011973 2 22 22 24.16 16.04 0.94 0.051973 2 24 23 42.01 1.09 0.71 0.021973 3 3 17 6.04 1.32 0.55 0.411973 3 5 14 1.91 7.01 0.65 0.341973 3 23 3 40.98 0.74 0.46 0.241973 11 21 20 63.13 2.31 0.41 0.21973 12 12 12 69.02 0.92 0.97 01973 12 25 1 3.03 1.06 0.76 0.231974 3 14 19 80.16 0.84 0.15 0.261975 10 22 4 90.31 1.48 0.8 0.011976 3 27 23 37.6 0.75 0.79 0.121977 2 11 16 22.3 0.92 0.23 0.541977 3 6 16 31.43 1.48 0.73 0.211971 2 26 12 13.13 6.91 0.95 0.04 COEN RIVER1971 3 12 17 6.97 2.17 0.49 0.451972 4 12 19 29.41 1.31 0.32 0.61973 3 3 22 11.78 1.19 0.35 0.561973 3 8 14 30.94 1.63 0.3 0.521973 3 22 14 0.41 3.59 0.68 0.321974 1 8 14 24.52 2.2 0.62 0.161974 2 26 13 3.53 5.11 0.86 0.131974 3 19 12 27.82 2.2 0.88 0.061974 3 24 16 36.38 1.59 0.54 0.111974 3 29 18 6.64 1.78 0.92 0.071975 2 6 18 45.12 1.51 0.35 0.431975 2 19 22 39.11 2.92 1 01975 10 22 5 35.62 3.07 0.7 0.031975 12 13 15 23.47 9.45 0.94 0.031975 12 19 17 49.63 0.2 0.3 0.02
Modelling losses in flood estimation E-14
Year month date hour IL(mm) CL(mm/h) PL VRC1976 2 18 10 14.93 1.96 0.44 0.461977 3 14 19 43 4.46 0.89 0.021978 2 18 19 0.26 9.11 0.8 0.21979 1 10 6 31.57 2.16 0.4 0.541979 3 23 3 10.3 3.29 0.34 0.541980 2 15 0 51.24 1.37 0.84 0.11981 1 23 13 0.27 2.62 0.76 0.241981 1 26 12 0.74 4.08 0.99 0.011981 2 25 13 13.84 2.66 0.68 0.271982 2 25 16 44.06 2.8 0.3 0.271982 3 31 18 36.13 0.55 0.14 0.381984 2 15 12 45.71 1.12 0.86 0.081984 2 23 13 1.77 4.16 0.46 0.531984 3 27 20 0.3 4.53 0.6 0.391985 2 17 23 3.19 2.7 0.41 0.561985 3 8 1 42.43 1.75 0.64 0.081985 3 31 22 25.93 1.55 0.63 0.291986 1 24 21 21.77 3.4 0.53 0.281986 2 8 18 62.85 0.62 0.21 0.151987 1 31 17 37.55 0.6 0.99 01987 2 2 21 21.38 1.95 0.81 0.141988 3 9 1 24.86 1.42 0.56 0.331988 12 13 13 25.25 3.76 0.92 0.061989 2 8 20 0.73 1.46 0.72 0.271991 2 14 19 7.52 1.18 0.96 0.041991 2 23 5 52.4 0.81 0.47 0.31992 1 11 8 26.45 2.96 0.92 0.051992 2 26 19 49.77 0.29 0.29 0.11992 12 21 19 15.53 2.42 0.85 0.141992 12 30 17 3.21 3.22 0.19 0.781993 1 23 23 15.62 0.96 0.48 0.461995 3 17 1 30.76 3.98 0.81 0.091996 3 3 14 14.78 4.07 0.48 0.411996 3 7 8 3.66 1.38 0.91 0.091996 3 10 9 0.82 1.43 0.51 0.491996 10 10 2 81.89 3.63 0.74 0.131996 12 24 23 9.07 3.13 0.89 0.11997 1 19 17 19.59 1.41 0.88 0.091997 3 4 13 14.26 2.53 0.86 0.121998 11 20 19 55.64 0.22 0.54 0.011999 1 1 19 63.04 0.08 0.31 0.011999 1 8 4 47.2 0.57 0.8 0.041999 1 29 21 31.59 0.3 0.15 0.21975 3 8 11 7.8 0.03 0.07 0.05 DULHUNTY RIVER1975 3 23 12 0.11 1.16 0.64 0.361976 1 16 12 4.67 5.91 0.96 0.041976 4 10 1 4.8 2.32 0.4 0.571979 1 10 20 0.32 1.83 0.37 0.631979 4 9 2 16.29 1.72 0.63 0.311980 1 7 14 0 0.53 0.51 0.491982 4 2 20 2.97 0.85 0.78 0.171984 2 10 13 0.72 1.48 0.88 0.091984 2 25 10 2.56 2.02 0.86 0.131986 1 15 14 0 3.54 0.77 0.231990 3 20 16 0.42 0.45 0.64 0.31975 3 6 21 0.2 0.18 0.95 0.04 BERTIE CREEK1975 3 8 9 0 2.27 0.83 0.171975 3 23 12 0.11 1.11 0.61 0.381976 1 16 13 4.67 6.85 0.95 0.041979 4 9 0 1.2 1.43 0.48 0.521983 5 20 20 5.03 1.33 0.98 0.021984 2 25 12 4.01 2.69 0.98 0.021986 1 15 10 0 1.84 0.93 0.07
Modelling losses in flood estimation E-15
APPENDIX F
Gow rie River
0
2
4
6
8
10
12
1 10 100 1000Duration (h)
CL
(mm
/h)
Blunder Creek
0
2
4
6
8
10
12
1 10 100 1000Duration (h)
CL
(mm
/h)
Stone River
02468
10121416
1 10 100
Bohle River
0123456789
1 10 100Duration (h)
CL
(mm
/h
Duration (h)
CL
(mm
/h)
)
Broughton River
0
2
4
6
8
10
1 10 100Duration (h)
CL
(mm
/h)
Blacks Creek
0
5
10
15
20
1 10 100 1000Duration (h)
CL
(mm
/h)
Sande Creek
0
2
4
6
8
10
1 10Durat ion (h)
100
Monal Creek
0
2
4
6
8
10
1 10 100Duration (h)
CL
(mm
/h)
Modelling losses in flood estimation F-1
Mary River
00.5
11.5
22.5
33.5
44.5
1 10 100 1000Duration (h)
CL
(mm
/h)
North Maroochy River
0
1
2
3
4
1 10 100 1000Duration (h)
CL
(mm
/h)
Bremer River
0
2
4
6
8
10
1 10 100Duration (h)
CL
(mm
/h)
Tenhill Creek
0
0.5
1
1.5
2
2.5
3
1 10Duration (h)
CL
(mm
/h100
)
Logan River
0
5
10
15
20
1 10 1Duration (h)
CL
(mm
/h
00
Running Creek
0
2
4
6
8
10
12
1 10 100Duration (h)
CL
(mm
/h))
Teviot Brook
0
1
2
3
4
5
6
7
8
1 10Duration (h)
CL
(mm
/h
100
)
Albert River
00.5
11.5
22.5
33.5
44.5
5
1 10 100
Duration (h)
CL
(mm
/h)
Macintyre Brook
0
2
4
6
8
10
12
14
16
1 10
Duration (h)
CL
(mm
/h)
100
Canal Creek
0
1
2
3
4
5
6
7
8
1 10 100Duration (h)
CL
(mm
/h)
Modelling losses in flood estimation F-2
Cadamine River
00.5
11.5
22.5
33.5
1 10 100Duration (h)
CL
(mm
/h)
Agate Creek
0
2
4
6
8
1 10Duration (h)
CL
(mm
/h)
100
Holroyd River
02468
1012141618
1 10 100
Coen River
0
2
4
6
8
10
1 10 100 1000Duration (h)C
L (m
m/h
)Duration (h)
CL
(mm
/h)
Dulhunty River
0
1
2
3
4
5
6
7
1 10Duration (h)
CL
(mm
/h
100
)
Bertie Creek
0
2
4
6
8
1 10 100Duration (h)
CL
(mm
/h)
Modelling losses in flood estimation F-3
APPENDIX G
Initial loss (mm) (ILs f or catchment ID 117003)
200.5181.5
162.5143.5
124.5105.5
86.567.5
48.529.5
10.5
Fre
quen
cy
10
8
6
4
2
0
initial loss (mm) (ILc for 117003)
160.0140.0120.0100.080.060.040.020.00.0
Freq
uenc
y
12
10
8
6
4
2
0
Modelling losses in flood estimation G-1
Initial loss (mm) (ILs for catchment ID120014)
68.059.150.241.232.323.314.45.5
Freq
uenc
y
8
6
4
2
0
Initial loss (mm) (ILc for catchment ID 120014)
68.059.150.241.232.323.314.45.5
Freq
uenc
y
8
6
4
2
0
Modelling losses in flood estimation G-2
Initial loss (mm) (ILs for catchment ID 145003)
83.558.132.77.3
Freq
uenc
y
30
20
10
0
Initial loss (mm) (ILc for catchment ID 145003)
87.367.046.726.46.1
Freq
uenc
y
30
20
10
0
Modelling losses in flood estimation G-3
Initial loss (mm) (ILs for catchment ID 138110)
90.676.863.049.235.521.77.9
Freq
uenc
y
14
12
10
8
6
4
2
0
Initial loss (mm) (ILc for catchment ID 138110)
91.579.467.355.343.231.219.17.0
Freq
uenc
y
16
14
12
10
8
6
4
2
0
Modelling losses in flood estimation G-4
Initial loss (mm) (ILs for catchment ID 145011)
86.072.959.846.733.720.67.5
Freq
uenc
y
40
30
20
10
0
Initial loss (mm) (ILc for catchment ID 145011)
86.875.363.952.541.029.618.26.7
Freq
uenc
y
50
40
30
20
10
0
Modelling losses in flood estimation G-5
Initial loss (mm) (ILs for catchment ID 416410)
87.978.869.660.551.342.233.023.914.75.6
Freq
uenc
y
50
40
30
20
10
0
Initial loss (mm) ( ILc for catchment ID 416410)
87.978.869.660.551.342.233.023.914.75.6
Freq
uenc
y
50
40
30
20
10
0
Modelling losses in flood estimation G-6
Initial loss (mm) ( ILs for catchment ID 422338)
100.076.753.430.16.8
Freq
uenc
y
40
30
20
10
0
Initial loss (mm) (ILc f or catchment ID 422338)
92.783.073.463.754.144.434.825.115.55.8
Freq
uenc
y
40
30
20
10
0
Modelling losses in flood estimation G-7
Initial loss (mm) (ILs f or catchment ID 422394)
87.978.869.660.551.342.233.023.914.75.6
Fre
quen
cy
40
30
20
10
0
Initial loss (mm) ( ILc f or catchment ID 422394)
83.274.565.957.248.639.931.322.614.05.3
Fre
quen
cy
40
30
20
10
0
Modelling losses in flood estimation G-8
Initial loss (mm) (ILs f or catchment ID 922101)
64.413.7
Fre
quen
cy
60
50
40
30
20
10
0
Initial loss (mm) (ILc f or catchment ID 922101)
76.755.133.411.8
Freq
uenc
y
50
40
30
20
10
0
Modelling losses in flood estimation G-9
APPENDIX H
α = 0.0080.008
Y M D Timestep Streamflow Rainfall Difference1 x1=(diff1*f1) BF=(Storm.+X1) Baseflow Netflow LogQ LogBF1990 5 22 1 0 1.5 0.12 0 0 1.5 1.5 0 0.176091 0.1760911990 5 22 2 1 1.5 0.11 0 0 1.5 1.5 0 0.176091 0.1760911990 5 22 3 2 1.5 0.12 0 0 1.5 1.5 0 0.176091 0.1760911990 5 22 4 3 1.5 0.1 0 0 1.5 1.5 0 0.176091 0.1760911990 5 22 5 4 1.5 0.08 0 0 1.5 1.5 0 0.176091 0.1760911990 5 22 6 5 1.5 0.16 0 0 1.5 1.5 0 0.176091 0.1760911990 5 22 7 6 1.4 0.56 0 0 1.4 1.4 0 0.146128 0.1461281990 5 22 8 7 1.4 2.54 0 0 1.4 1.4 0 0.146128 0.1461281990 5 22 9 8 1.4 1.59 0 0 1.4 1.4 0 0.146128 0.1461281990 5 22 10 9 1.4 0.39 0 0 1.4 1.4 0 0.146128 0.1461281990 5 22 11 10 1.4 0.39 0 0 1.4 1.4 0 0.146128 0.1461281990 5 22 12 11 1.3 0 0 0 1.3 1.3 0 0.113943 0.1139431990 5 22 13 12 1.2 3.48 0 0 1.3 1.3 -0.1 0.079181 0.1139431990 5 22 14 13 1.2 3.57 0 0 1.2 1.2 0 0.079181 0.0791811990 5 22 15 14 1.3 2.76 0 0 1.3 1.3 0 0.113943 0.1139431990 5 22 16 15 1.3 3.23 0 0 1.3 1.3 0 0.113943 0.1139431990 5 22 17 16 1.4 3.36 0 0 1.4 1.4 0 0.146128 0.1461281990 5 22 18 17 1.4 4.28 0 0 1.4 1.4 0 0.146128 0.1461281990 5 22 19 18 1.5 3.28 0 0 1.5 1.5 0 0.176091 0.176091
0
10
20
30
40
50
60
0 50 100 150Time step
Q
-0.5
0
0.5
1
1.5
2
0 50 100 150
Time step
Log
Q
Modelling losses in flood estimation H-1
Y M D Timestep Streamflow Rainfall Difference1 x1=(diff1*f1) BF=(Storm.+X1) Baseflow Netflow LogQ LogBF1990 5 22 20 19 1.6 5.6 0 0 1.6 1.6 0 0.20412 0.204121990 5 22 21 20 2.1 8.93 0 0 2.1 2.1 0 0.322219 0.3222191990 5 22 22 21 3.1 13.79 1 0.008 2.108 2.108 0.992 0.491362 0.3238711990 5 22 23 22 4.9 10.92 2.792 0.022336 2.130336 2.130336 2.769664 0.690196 0.3284481990 5 22 24 23 6.5 0.39 4.369664 0.03495731 2.165293312 2.165293312 4.334706688 0.812913 0.3355171990 5 23 25 0 6.8 0.19 4.634706688 0.03707765 2.202370966 2.202370966 4.597629034 0.832509 0.342891990 5 23 26 1 12.2 0.2 9.997629034 0.07998103 2.282351998 2.282351998 9.917648002 1.08636 0.3583831990 5 23 27 2 31.2 0.17 28.917648 0.23134118 2.513693182 2.513693182 28.68630682 1.494155 0.4003121990 5 23 28 3 36.5 0.31 33.98630682 0.27189045 2.785583636 2.785583636 33.71441636 1.562293 0.4449161990 5 23 29 4 35.3 0.2 32.51441636 0.26011533 3.045698967 3.045698967 32.25430103 1.547775 0.4836871990 5 23 30 5 35.2 0 32.15430103 0.25723441 3.302933376 3.302933376 31.89706662 1.546543 0.51891990 5 23 31 6 48.7 0.2 45.39706662 0.36317653 3.666109909 3.666109909 45.03389009 1.687529 0.5642051990 5 23 32 7 51.7 0.19 48.03389009 0.38427112 4.050381029 4.050381029 47.64961897 1.713491 0.6074961990 5 23 33 8 48.8 0.19 44.74961897 0.35799695 4.408377981 4.408377981 44.39162202 1.68842 0.6442791990 5 23 34 9 44.4 0 39.99162202 0.31993298 4.728310957 4.728310957 39.67168904 1.647383 0.6747061990 5 23 35 10 40 0 35.27168904 0.28217351 5.01048447 5.01048447 34.98951553 1.60206 0.699881990 5 23 36 11 36.1 0 31.08951553 0.24871612 5.259200594 5.259200594 30.84079941 1.557507 0.720921990 5 23 37 12 32.8 0 27.54079941 0.2203264 5.479526989 5.479526989 27.32047301 1.515874 0.7387431990 5 23 38 13 30.1 0 24.62047301 0.19696378 5.676490773 5.676490773 24.42350923 1.478566 0.754081990 5 23 39 14 28.1 0 22.42350923 0.17938807 5.855878847 5.855878847 22.24412115 1.448706 0.7675921990 5 23 40 15 26.2 0 20.34412115 0.16275297 6.018631816 6.018631816 20.18136818 1.418301 0.7794981990 5 23 41 16 24.2 0 18.18136818 0.14545095 6.164082762 6.164082762 18.03591724 1.383815 0.7898681990 5 23 42 17 22.5 0 16.33591724 0.13068734 6.2947701 6.2947701 16.2052299 1.352183 0.798981990 5 23 43 18 21.2 0 14.9052299 0.11924184 6.414011939 6.414011939 14.78598806 1.326336 0.807131990 5 23 44 19 20.1 0 13.68598806 0.1094879 6.523499843 6.523499843 13.57650016 1.303196 0.8144811990 5 23 45 20 18.9 0.6 12.37650016 0.099012 6.622511844 6.622511844 12.27748816 1.276462 0.8210231990 5 23 46 21 17.8 0.5 11.17748816 0.08941991 6.71193175 6.71193175 11.08806825 1.25042 0.8268481990 5 23 47 22 16.9 0.51 10.18806825 0.08150455 6.793436296 6.793436296 10.1065637 1.227887 0.832091990 5 23 48 23 16 0.1 9.206563704 0.07365251 6.867088805 6.867088805 9.132911195 1.20412 0.8367731990 5 24 49 0 15.2 0.1 8.332911195 0.06666329 6.933752095 6.933752095 8.266247905 1.181844 0.8409681990 5 24 50 1 14.5 0.2 7.566247905 0.06052998 6.994282078 6.994282078 7.505717922 1.161368 0.8447431990 5 24 51 2 14 0.03 7.005717922 0.05604574 7.050327822 7.050327822 6.949672178 1.146128 0.8482091990 5 24 52 3 13.4 0.07 6.349672178 0.05079738 7.101125199 7.101125199 6.298874801 1.127105 0.8513271990 5 24 53 4 12.8 0.01 5.698874801 0.045591 7.146716197 7.146716197 5.653283803 1.10721 0.8541071990 5 24 54 5 12.3 0.1 5.153283803 0.04122627 7.187942468 7.187942468 5.112057532 1.089905 0.856605
Modelling losses in flood estimation H-2
Y M D Timestep Streamflow Rainfall Difference1 x1=(diff1*f1) BF=(Storm.+X1) Baseflow Netflow LogQ LogBF1990 5 24 55 6 11.7 0 4.512057532 0.03609646 7.224038928 7.224038928 4.475961072 1.068186 0.858781990 5 24 56 7 11.1 0 3.875961072 0.03100769 7.255046617 7.255046617 3.844953383 1.045323 0.860641990 5 24 57 8 10.6 0 3.344953383 0.02675963 7.281806244 7.281806244 3.318193756 1.025306 0.8622391990 5 24 58 9 10 0 2.718193756 0.02174555 7.303551794 7.303551794 2.696448206 1 0.8635341990 5 24 59 10 9.7 0 2.396448206 0.01917159 7.322723379 7.322723379 2.377276621 0.986772 0.8646731990 5 24 60 11 9.3 0 1.977276621 0.01581821 7.338541592 7.338541592 1.961458408 0.968483 0.865611990 5 24 61 12 8.9 0 1.561458408 0.01249167 7.35103326 7.35103326 1.54896674 0.94939 0.8663481990 5 24 62 13 8.7 0 1.34896674 0.01079173 7.361824994 7.361824994 1.338175006 0.939519 0.8669851990 5 24 63 14 8.5 0 1.138175006 0.0091054 7.370930394 7.370930394 1.129069606 0.929419 0.8675221990 5 24 64 15 8.4 0 1.029069606 0.00823256 7.37916295 7.37916295 1.02083705 0.924279 0.8680071990 5 24 65 16 8.2 0 0.82083705 0.0065667 7.385729647 7.385729647 0.814270353 0.913814 0.8683931990 5 24 66 17 8 0 0.614270353 0.00491416 7.39064381 7.39064381 0.60935619 0.90309 0.8686821990 5 24 67 18 7.9 0 0.50935619 0.00407485 7.394718659 7.394718659 0.505281341 0.897627 0.8689221990 5 24 68 19 7.7 0 0.305281341 0.00244225 7.39716091 7.39716091 0.30283909 0.886491 0.8690651990 5 24 69 20 7.6 0 0.20283909 0.00162271 7.398783623 7.398783623 0.201216377 0.880814 0.869161990 5 24 70 21 7.5 0 0.101216377 0.00080973 7.399593354 7.399593354 0.100406646 0.875061 0.8692081990 5 24 71 22 7.4 0 0.000406646 3.2532E-06 7.399596607 7.399596607 0.000403393 0.869232 0.8692081990 5 24 72 23 7.2 0 -0.199596607 -0.0015968 7.397999834 7.2 0 0.857332 0.8573321990 5 25 73 0 7.1 0 -0.297999834 -0.002384 7.395615835 7.1 0 0.851258 0.8512581990 5 25 74 1 7 0 -0.395615835 -0.0031649 7.392450909 7 0 0.845098 0.8450981990 5 25 75 2 6.9 0 -0.492450909 -0.0039396 7.388511301 6.9 0 0.838849 0.8388491990 5 25 76 3 6.8 0 -0.588511301 -0.0047081 7.383803211 6.8 0 0.832509 0.8325091990 5 25 77 4 6.7 0 -0.683803211 -0.0054704 7.378332785 6.7 0 0.826075 0.8260751990 5 25 78 5 6.6 0 -0.778332785 -0.0062267 7.372106123 6.6 0 0.819544 0.8195441990 5 25 79 6 6.5 0 -0.872106123 -0.0069768 7.365129274 6.5 0 0.812913 0.8129131990 5 25 80 7 6.4 0 -0.965129274 -0.007721 7.35740824 6.4 0 0.80618 0.806181990 5 25 81 8 6.3 0 -1.05740824 -0.0084593 7.348948974 6.3 0 0.799341 0.7993411990 5 25 82 9 6.2 0 -1.148948974 -0.0091916 7.339757382 6.2 0 0.792392 0.7923921990 5 25 83 10 6.1 0 -1.239757382 -0.0099181 7.329839323 6.1 0 0.78533 0.785331990 5 25 84 11 6.1 0 -1.229839323 -0.0098387 7.320000608 6.1 0 0.78533 0.785331990 5 25 85 12 6 0 -1.320000608 -0.01056 7.309440604 6 0 0.778151 0.7781511990 5 25 86 13 5.9 0 -1.409440604 -0.0112755 7.298165079 5.9 0 0.770852 0.7708521990 5 25 87 14 5.8 0 -1.498165079 -0.0119853 7.286179758 5.8 0 0.763428 0.7634281990 5 25 88 15 5.7 0 -1.586179758 -0.0126894 7.27349032 5.7 0 0.755875 0.7558751990 5 25 89 16 5.6 0 -1.67349032 -0.0133879 7.260102397 5.6 0 0.748188 0.748188
Modelling losses in flood estimation H-3
Y M D Timestep Streamflow Rainfall Difference1 x1=(diff1*f1) BF=(Storm.+X1) Baseflow Netflow LogQ LogBF
1990 5 25 90 17 5.6 0 -1.660102397 -0.0132808 7.246821578 5.6 0 0.748188 0.7481881990 5 25 91 18 5.5 0 -1.746821578 -0.0139746 7.232847006 5.5 0 0.740363 0.7403631990 5 25 92 19 5.4 0 -1.832847006 -0.0146628 7.21818423 5.4 0 0.732394 0.7323941990 5 25 93 20 5.3 0 -1.91818423 -0.0153455 7.202838756 5.3 0 0.724276 0.7242761990 5 25 94 21 5.2 0 -2.002838756 -0.0160227 7.186816046 5.2 0 0.716003 0.7160031990 5 25 95 22 5.1 0 -2.086816046 -0.0166945 7.170121517 5.1 0 0.70757 0.707571990 5 25 96 23 5 0 -2.170121517 -0.017361 7.152760545 5 0 0.69897 0.698971990 5 26 97 0 5 0 -2.152760545 -0.0172221 7.135538461 5 0 0.69897 0.698971990 5 26 98 1 4.9 0 -2.235538461 -0.0178843 7.117654153 4.9 0 0.690196 0.6901961990 5 26 99 2 4.8 0 -2.317654153 -0.0185412 7.09911292 4.8 0 0.681241 0.6812411990 5 26 100 3 4.7 0 -2.39911292 -0.0191929 7.079920017 4.7 0 0.672098 0.6720981990 5 26 101 4 4.7 0 -2.379920017 -0.0190394 7.060880656 4.7 0 0.672098 0.6720981990 5 26 102 5 4.7 0 -2.360880656 -0.018887 7.041993611 4.7 0 0.672098 0.6720981990 5 26 103 6 4.6 0 -2.441993611 -0.0195359 7.022457662 4.6 0 0.662758 0.6627581990 5 26 104 7 4.6 0 -2.422457662 -0.0193797 7.003078001 4.6 0 0.662758 0.6627581990 5 26 105 8 4.5 0 -2.503078001 -0.0200246 6.983053377 4.5 0 0.653213 0.6532131990 5 26 106 9 4.4 0 -2.583053377 -0.0206644 6.96238895 4.4 0 0.643453 0.6434531990 5 26 107 10 4.4 0 -2.56238895 -0.0204991 6.941889838 4.4 0 0.643453 0.6434531990 5 26 108 11 4.3 0 -2.641889838 -0.0211351 6.92075472 4.3 0 0.633468 0.6334681990 5 26 109 12 4.2 0 -2.72075472 -0.021766 6.898988682 4.2 0 0.623249 0.6232491990 5 26 110 13 4.2 0 -2.698988682 -0.0215919 6.877396772 4.2 0 0.623249 0.6232491990 5 26 111 14 4.2 0 -2.677396772 -0.0214192 6.855977598 4.2 0 0.623249 0.6232491990 5 26 112 15 4.2 0 -2.655977598 -0.0212478 6.834729778 4.2 0 0.623249 0.6232491990 5 26 113 16 4.1 0 -2.734729778 -0.0218778 6.812851939 4.1 0 0.612784 0.6127841990 5 26 114 17 4.1 0 -2.712851939 -0.0217028 6.791149124 4.1 0 0.612784 0.6127841990 5 26 115 18 4 0 -2.791149124 -0.0223292 6.768819931 4 0 0.60206 0.60206
Modelling losses in flood estimation H-4