kamsiah abdul wahab 09_24
TRANSCRIPT
UNIVERSITI TEKNOLOGI MARA
APPLICATION OF FUZZY LOGIC TO SIMULATION OF RAINFALL
IN KERAYONG RIVER CATCHMENT
KAMSIAH ABDUL WAHAB
Thesis submitted in fulfillment of the requirements for the degree of
Master of Science in Civil Engineering
Faculty of Civil Engineering
March 2009
CANDIDATE'S DECLARATION
I declare that the work in this thesis was carried out in accordance with the
regulations of University Teknologi MARA. It is original and is the result of my
own work, unless otherwise indicated or acknowledged as referenced work. This
thesis has not been submitted to any other academic institution or non-academic
institution for any order degree or qualification.
In the event that my thesis be found to violate the conditions mentioned above, I
voluntarily waive the right of conferment of my degree and agree to be subjected to
the disciplinary rules and regulations of University Teknologi MARA.
Name of Candidate
Candidate's ID No.
Programme
Faculty
Thesis Title
Kamsiah Binti Abdul. Wahab
2002200136
EC780
Faculty.of Civil Engineering
Application of Fuzzy Logic To Simulation
Signature of Candidate
Date 11-03-2009
COPYRIGHT © UiTM
APPLICATION OF FUZZY LOGIC TO SIMULATION
OF RAINFALL IN KERAYONG RIVER CATCHMENT
Abstract
Short term rainfall characteristics of Kerayong River catchment, a tributary of Klang
River were analysed in this research. As the urbanisation process and the population
increased in this catchment is inevitable, therefore knowledge on rainfall
characteristics is necessary and useful in designing future drainage system. There are
four rainfall and five stage monitoring stations established in this catchment. The
rainfall stations are equipped with data logger of 0.5 mm tipping bucket resolution
and record the data based on event mode. Time Dependent Data Analysis (TIDEDA)
program is used by Department of Irrigation and Drainage (DID) Malaysia to read
and edit the data for further analysis. The available data of Kerayong River
catchment were beginning from year 2001 and 2002 which continuous data were
available for all stations. The spatial and temporal distribution for Kerayong River
catchment is studied by plotting the storm event at one minute interval. The analysis
shows that the daily rainfall temporal pattern for the year 2001 and 2002 was found
to be relatively similar to all stations. However the magnitude of rainfall intensity
varies considerably at short time intervals. The result shows that there are missing
data occur during the thunderstorm at Kg. Cheras Baru station from 15th January to
16th May 2003. The missing data are related to the malfunctioning of the instruments
and vandalism. Therefore, the potential of fuzzy logic modeling i.e. Fuzzy Rule
Based Model (FRBM) and Normal Ratio Method (NRM) for filling the missing data
at Kg. Cheras Baru was investigated. A total of 420 daily rainfall values i.e. data for
year 2001 and 2002 are employed to construct the FRBM. The numbers of data sets
are selected in randomly and divided into training and verification sets. The rules are
adopted from the available daily rainfall record with the simple assumption. The
ii
COPYRIGHT © UiTM
daily rainfall at Kg. Cheras Baru is assumed to be missing and unseen during the
training session.
The NRM is used in comparing the results produced by FRBM. The accuracy of
this method is statistically measured using standard error and coefficient of
determination. The standard error of the constructed model with rain factors applied
is found to be below 4.0 mm while the coefficient of determination is 0.9. The rain
factors applied in FRBM i.e. 0.0 equals to no rain and 0.5 equals to rain event are
more superior in predicting the missing data compared to the NRM. The model was
verified using collected data from January to May 2003. However, the accuracy of
FRBM is governed by the set of rules provided. The fuzzy rules will increase when
large historical data was used. To achieve better results, other input parameter such
as time of rainfall occurrence and associated water level data should be considered
in FRBM. The FRBM proposed is this study has a potential in determining the
missing rainfall data as well as verifying the collected rainfall data at any site
concerned. The main advantage of FRBM is that the rules provided are written in a
simple natural human language and the concept is easy to understand.
i i i
COPYRIGHT © UiTM
ACKNOWLEDGEMENT
First and foremost, praised to Allah The Almighty for giving me full of high spirits
to complete this thesis. My sincere thanks to my supervisor, Dato' Prof. Ir. Dr
Sahol Hamid Abu Bakar, Senior Director of the Ministry of Higher Education
(MOHE), Malaysia for giving me opportunity to be his graduate student. I am
deeply grateful and my utmost appreciation for his professionalism in guidance,
constructive comments and time at all stages of research work to complete this
thesis. To my employer, UiTM, thanks are attended for providing research facilities
and financial support. Thanks to Assoc. Prof. Dr Ismail Atan for his interesting
discussion and valuable suggestion. I convey my honest thanks to Dr Hj Mohd Nor
Hj Mohd Desa, Director of Humid Tropic Centre (HTC) Kuala Lumpur, who
supervised and guided the research throughout all stages from the beginning. I
would like to express my thanks to the entire staff of HTC, directly or indirectly for
their assistance, teamwork and helpfulness.
I would like to express my gratefulness to Mr. Hashim Harun and Mr. V.
Nadarajan from Hydrology Branch of the Department of Irrigation and Drainage,
Malaysia for their supervision and help during the site visit and providing various
data especially rainfall and water level data. To Mr. Salim Bacik, your
understanding and expertise in Time Dependent Data Analysis (TIDEDA) are very
much appreciated. Thanks to Technsource systems in providing Matrix Laboratory
(MATLAB) information with details especially to Mr. Chin Yik Sean. Thanks to
my good friend at work, Ms. Munira Akhmal Hashim, for her time and effort in
getting the information together where the experience we gets are very meaningful
to us and may our friendship remains forever. I wish to convey my special
gratefulness to my beloved mother, Mrs. Siti Khadijah Ahmad for her loving,
patience and support for me to complete this thesis. My special appreciations extend
to all my family members: Zarifah, Ridzuan, Haslina, Subhah, Azman and
Rahmat for their love, constant and encouraging support throughout my study in
iv
COPYRIGHT © UiTM
UiTM. Finally, I would like to thank all my friends directly or indirectly involved in
this thesis for their support and kindness especially to Ms. Fairudz Yahya. Thank
for your love, affection and advice. Their help and effort cannot be repaid and may
God bless them.
v
COPYRIGHT © UiTM
TABLES OF CONTENTS
page TITLE PAGE
CANDIDATE'S DECLARATION
ABSTRACT ii
ACKNOWLEDGEMENTS iv
TABLES OF CONTENTS vi
LIST OF TABLES x
LIST OF FIGURES xiii
LIST OF PLATES xv
NOTATION AND ABBREVIATION xvi
CHAPTER 1 : INTRODUCTION
1.1 General 1
1.2 Background and problem statement 3
1.3 Objective of the study 5
1.4 Scope of the study 5
1.5 Significance of the research 7
1.6 Assumption and limitation 8
CHAPTER 2 : LITERATURE REVIEW
2.1 Introduction 9
2.2 Rainfall study 10
2.3 Missing and reconstruction of data 11
2.4 Fuzzy logic theory with other AI approaches 15
CHAPTER 3 : THEORETICAL CONSIDERATION
3.1 Study area 20
3.2 Network establishment 22
vi
COPYRIGHT © UiTM
3.2.1 Rainfall stations 22
3.2.2 Water level stations 24
3.3 Instrumentation 27
3.3.1 Tipping bucket-rain gauge 27
3.3.2 Hydrologger 28
3.3.3 Static random access memory (SRAM) card 28
3.3.4 Hydro reader unit 29
3.3.5 Water level recorder 29
3.3.6 SEBA strip chart 30
3.4 Data collection and processing 31
3.5 Rainfall intensity 33
3.6 Missing data analysis 34
3.6.1 Normal ratio method 34
3.6.2 Fuzzy logic approach 35
3.7 Definition of terms in fuzzy logic 36
3.7.1 Universe of discourse 36
3.7.2 Fuzzy logic 36
3.7.3 Fuzzy sets 36
3.7.4 Membership function 37
3.7.5 Fuzzy numbers 37
3.7.6 Fuzzification 37
3.7.7 Defuzzification 38
3.7.8 Linguistic variables and linguistic terms 38
3.8 Fuzzy rule based model (FRBM) 3 8
3.8.1 Fuzzification 39
3.8.2 Fuzzy inference system (FIS) 39
3.8.3 Fuzzy rule 40
3.8.4 Defuzzification 41
CHAPTER 4 : RESEARCH METHODOLOGY
4.1 Hydrological process 42
vii
COPYRIGHT © UiTM
4.2 Site investigation 43
4.3 Installation of equipment 43
4.4 Data collection, data retrieval and data processing 44
4.5 Data analysis 45
4.5.1 NRM 45
4.5.2 FRBM 46
4.6 Parameter determination 48
4.6.1 Input and output 48
4.6.2 Fuzzification 50
4.6.3 Implication method - fuzzy If-Then rules 54
4.6.4 Defuzzification 55
4.7 Goodness of fit test 57
4.7.1 Standard error (SE) 57
4.7.2 Coefficient of determination^2) 58
CHAPTER 5 : ANALYSIS RESULTS AND DISCUSSION
5.1 Overall data record 59
5.2 Rainy days records 60
5.3 Rain days analysis 61
5.4 Rainfall temporal pattern 63
5.4.1 1-min 63
5.4.2 Daily 67
5.4.3 Monthly 70
5.5 NRM analysis 73
5.6 Fuzzy logic analysis 79
5.6.1 FRBM(I) result 79
5.6.2 FRBM (II) result 86
5.7 Comparison between NRM and FRBM result 94
5.8 Discussion of results 98
5.8.1 Discussion of NRM analysis 99
5.8.2 Discussion of FRBM(I) analysis 100
5.8.3 Discussion of FRBM (II) analysis 101
5.9 Verification data analysis 103
viii
COPYRIGHT © UiTM
CHAPTER 6 : CONCLUSION AND RECOMMENDATION
6.1 Rainfall study
6.2 Method of infilling data
6.3 The benefit of method use
6.4 Recommendation
6.4.1 Data accuracy
6.4.2 Station
6.4.3 Fuzzy logic
106
107
110
111
111
111
111
REFERENCES
APPENDICES
Appendix A
Appendix B
Appendix C
Appendix D
Appendix E
Appendix F
Appendix G
Land use of Kerayong River catchment
Site visit to Kerayong river catchment
Fuzzy rules
FIS properties
Rainfall and water level data at Taman Desa station
Normal ratio method calculation
Verification data
ix
COPYRIGHT © UiTM
LIST OF TABLES
Table Title Page
No. No.
2.1 Examples of rules to determine the sensitivity degree 16
3.1 Details of rain gauges establish in Kerayong River catchment 22
3.2 Details of water level establish in Kerayong river catchment 24
3.3 TIDEDA features 32
3.4 Rainfall and water level data collected 33
3.5 Categorization of rainfall intensity by DID 33
4.1 The classification of station number 49
4.2 Categorization of rainfall depth 49
4.3 Categorization of RF 49
4.4 The FRBM type with the linguistic variables of input and 50 output variables
4.5 Fuzzy classes for D5,D4,D3 and Dl 52
4.6 Fuzzy classes for RF 53
5.1 The portion of missing rainfall data 5 9
5.2 The portion of missing water level data 60
5.3 Number of rain days 60
5.4 Longest period in days without rain and with rain for the year 2001 61
5.5 Longest period in days without rain and with rain for the year 2002 62
5.6 Longest period in days without rain and with rain for the year 2003 62
5.7 Frequency analysis of rainfall recorded for the year 2001 62
5.8 Frequency analysis of rainfall recorded for the year 2002 63
5.9 Selected temporal pattern of 1-minute intervals 64
5.10 Highest intensity of observed rainfall (mm) 67
5.11 Highest rainfall record (mm) and date occurrence in subscript 69
5.12 Daily average (mm) value according to year 70
5.13 Daily average (mm) recorded at each rain gauge for all seasons 70
5.14 Monthly rainfall (mm) for the year 2001 71
x
COPYRIGHT © UiTM
5.15 Monthly rainfall (mm) for the year 2002 71
5.16 Monthly rainfall (mm) for the year 2003 71
5.17 Statistical characteristic of monthly rainfall data for the year 2002 72
5.18 NRM training results for the year 2001 according to a selected date 73
5.19 NRM training results for the year 2002 according to a selected date 7 4
5.20 Statistic between observed and NRM in NE and SW monsoon 78
5.21 Accuracy of NRM analysis 78
5.22 FRBM (I) results with entire rules for year according to a selected 79 date
5.23 FRBM (I) results with entire rules for year according to a selected 80 date
5.24 Statistic between entire rules applied in FRBM (I) and observed data 82 in intermonsoon month
5.25 Statistic between 46-R of FRBM (I) and observed data in NE and SW 85 monsoon
5.26 Accuracy of FRBM (I) data analysis nr
5.27 First FRBM (II) training data result for the year 2001 (mm) 86
5.28 First FRBM (II) training data result for the year 2002 (mm) 87
5.29 Final FRBM (II) training result for the year 2001 and year 2002 88 (mm)
5.30 Statistic between entire rules applied for FRBM (II) and observed 90 data in intermonsoon month
5.31 Statistic between 46-R of FRBM (II) and observed data in NE and 93 SW monsoon
5.32 Accuracy of FRBM (II) data analysis to the observed value 93
5.33 Result comparison for the year 2001 and year 2002 (mm) 94
5.34 Statistics analysis of comparison result in NE monsoon month 97
5.35 Summary of Goodness of Fit Test for year 2001 and year 2002 97
5.36 Estimate missing data with the comparison of input used (mm) 98
5.37 Comparison between observed average rainfall and estimated value 105
6.1 The benefit of fuzzy logic modeling 110
C-l 'NR' rules system for the determination of rainfall depth at S1
C-2 ' VL' rules system for the determination of rainfall depth at S1
XI
COPYRIGHT © UiTM
C-3 'L' rules system for the determination of rainfall depth at SI
C-4 'L' rules system for the determination of rainfall depth at SI
C-5 'H' rules system for the determination of rainfall depth at SI
C-6 'VH' rules system for the determination of rainfall depth at SI
E-1 Rainfall data at Taman Desa station for year 2002
E-2 Water level data at Taman Desa station for year 2002
G-1 Infilling missing daily rainfall at Kg. Cheras Baru station for year 2003
xii
COPYRIGHT © UiTM
LIST OF FIGURES
Figure Title Page No. No. 2.1 The overlapping of membership function in FRBM 13
2.2 The type of overlapping membership functions 13
2.3 Membership function for extreme cumulative rainfall (C) and 14 excessive cumulative rainfall (C)
2.4 Membership function for density of population 16
3.1 Location of the study area and hydrological stations established 21 in Kerayong River catchment
3.2 Trapezoidal and triangular fuzzy numbers 37
3.3 Structure of FRBM 39
4.1 Research methodology chart 42
4.2 Schematic diagram of FRBM procedure 47
4.3 Membership function of triangular shape 51
4.4 Membership function of trapezoidal shape for 'very high' and ' 51 rain' classification
4.5 Membership function of trapezoidal shape for 'no rain' 52 classification
4.6 Membership function for D5,D4,D3 and Dl 53
4.7 Membership function for RF 53
4.8 The graphical example for inference diagram of Mamdani 56 method and defuzzification process
5.1 Number of rain day for year 2001 and year 2002 61
5.2 Selected temporal pattern for 1 -minute intensities of storm for the 65 year 2001
5.3 Selected temporal pattern for 1 -minute intensities of storm for the 66 year 2002
5.4 Daily rainfall temporal pattern for all stations in the year 2001 68
5.5 Daily rainfall temporal pattern for all stations in the year 2002 69
5.6 Total monthly rainfall for all stations in year 2002 72
5.7 Plot between observed and NRM in NE monsoon 75
5.8 Plot between observed and NRM in SW monsoon 77
xiii
COPYRIGHT © UiTM
LIST OF PLATES
Plate Title Page No. No.
3.1 Rain gauge at Kg. Cheras Baru 22
3.2 Rain gauge at Taman Sg. Besi 23
3.3 Rain gauge at Taman Desa 23
3.4 Telemetry rainfall station at Pandan Indah 24
3.5 Water level station at Taman Sg. Besi 25
3.6 Water level station at Taman Desa 25
3.7 Water level station at Kg. Cheras Baru 26
3.8 Stick gauge 26
3.9 Tipping bucket with 0.5 mm resolution 27
3.10 Hydrologger for logging rainfall and water level data 28
3.11 Hydro reader unit 29
3.12 Shaft encoder 30
3.13 SEBA strip chart 31
B-l The main drain at Taman Sg. Besi station
B-2 Secondary drain at Taman Desa station
B-3 Kerayong River at the confluence of Cheras River
B-4 Downstream of Kerayong River through to Klang River
B-5 The garbage trap near Sg. Besi Highway
B-6 Construction work to remove the plan tree
B-7 During the site visit at Kg. Cheras Baru station to retrieve data
B-8 The development at Taman Desa station
B-9 The damages of stick gauge at Taman Miharja station
B-10 The damages of solar because of human influence
xv
COPYRIGHT © UiTM
NOTATION AND ABBREVIATION
ANN Artificial Neural Network
AND Fuzzy operator
AI Artificial Intelligence
Centroid Centre of area method
D Rainfall depth
DOF Degree of fulfillment
DID Department of Irrigation and Drainage
FIS Fuzzy Inference System
FLS Fuzzy Logic System
FRBM Fuzzy Rule Based Model
GUI Graphical User Interface
H High
HTC Humid Tropic Centre
/ intensity (mm/min)
km kilometer
L Light
LV Linguistic variable
m Number of membership function
mm millimeter
M Medium
MATLAB Matrix laboratory
MFs Membership function
n Number of input
N Number of training data
NE Northeast monsoon
NIWA National Institute of Water Atmospheric Research
NR No Rain
NRM Normal Ratio Method
xvi
COPYRIGHT © UiTM
Obs
r2
R
RF
S
SE
SW
SRAM
SMART
T
TR
TIDEDA
Trimf
Trapmf
U
H(x)
VL
V
Oi
o Pi
%
X
X
Observed
Coefficient of determination
Rain
Rain factor
Station
Standard error
Southwest monsoon
Static random access memory card
Storm Management and Road Tunnel
Triangular fuzzy number
Trapezoidal fuzzy number
Time Dependent Data Analysis
Triangular membership function
Trapezoidal membership function
The universe of discourse
Membership function for the fuzzy set
Very Light
Very High
Observed data
Observed mean data
Estimated data
Percentage
Mean data
Summation
xvii
COPYRIGHT © UiTM
CHAPTER 1
INTRODUCTION
1.1 General
Rapid urbanisation contributes to frequent detrimental effects to the hydrological
cycle. The distinct impact is that change in the land use for development purpose
result in decreasing infiltration, increasing run off volume and later on accompanied
by changes in rainfall pattern. These factors contribute to critical flash flood in urban
areas especially in Kuala Lumpur. Generally flooding happens within a short
duration of time commonly resulted from high intensity storms. In the year of 2002,
there are eight flash floods were recorded in Kuala Lumpur by the Department of
Irrigation and Drainage (DID) Malaysia, an increase in number from previous year.
As such, it is important for engineers to be knowledgeable about the natural passage
of excess runoff and the rainfall characteristics such as the intensity, duration and
frequency in order to conceptualize and predict their effects to the drainage network
design whereby any changes in space and time may influence the planning progress.
Kerayong River catchments are chosen as the study area and as an experimental
basin by DID, Malaysia because the surface characteristics are mainly dictated by
urbanisation process as more than half of the areas are developed as business and
commercial centers or organized residential areas. Most rainfall events in the area
are of short durations which occur in only a small part of the total storm duration.
For that reason the rainfall variability in space and time at this catchment area are
being studied especially at short time scale where the storm events are analyze at
one minute interval. Therefore, this knowledge is one of the ways in tackling the
problems related to the storm drainage in the region of rapid urbanisation and the
1
COPYRIGHT © UiTM
understanding on spatial and temporal characteristics is necessary for better
management practice.
In determining the rainfall pattern during a thunderstorm a very dense, so efficient
and reliable network are required. For a better result, long historical series of
hydrological data are needed as a key requirement in designing drainage network.
The main challenge for engineers and planners in designing the drainage network is
to maintain good quality data standards which are acceptable to all users. The main
problem in adding the data and knowledge to the applied system are many
incompatible, inconsistent and missing data. Methodologies or techniques that help
to improve the accuracy of estimating missing observations are highly significant
and useful for the process of hydrological modeling, which requires complete data
sets. Therefore, filling in missing section of data is one of the important issues in
urban hydrology, which is to be addressed in this research. A continuous hydrologic
simulation is recommended as an alternative to the traditional design event approach
for the analysis and the hydraulics design. There are two approached methods
applied to be filled in the missing records i.e. Normal Ratio Method (NRM) and
Fuzzy Rule Based Model (FRBM). The FRBM is a new approach and theory based
on mathematical equation using linguistic variable is being developed and
investigated in this study for filling the missing records and to reconstruct the data.
The NRM is used in comparing the results produced by FRBM.
For this study, both models use daily rainfall as an input parameter and the complete
hydrological data recorded at Kerayong river catchments is available from June
2001 until June 2003. The available data are employed to construct fuzzy rules in
FRBM until the best rules estimates are obtained. There are some missing data occur
during a thunderstorm at Kg. Cheras Baru station from 15th January to 16th May
2003. Therefore, the daily rainfall at Kg. Cheras Baru is assumed to be missing and
unseen during the training session, as the number of sets data are selected in
randomly to perform the analysis. The analysis is continued for verification sets to
2 COPYRIGHT © UiTM
find the actual missing value occur at Kg. Cheras Baru station and the missing value
is estimated using the other complete time series of nearby stations. Therefore, the
potential of fuzzy logic modeling i.e. Fuzzy Rule Based Model (FRBM) and Normal
Ratio Method (NRM) for filling the missing data at Kg. Cheras Baru was
investigated.
1.2 Background and problem statement
The rapid population growth with dynamic change in the land use pattern has an
impact on the hydrological processes. This linked to the occurrence of flash floods.
Much of the original forest cover has been replaced by urban land use and
development (see Ruzardi, 2002) which now encroached into foothills and this
have remitted in surface erosion and an increase in flash flood incidence. One of the
effects of urbanization is the increase in impervious cover where it will decrease the
infiltration capacity and increase the magnitude of surface runoff. The increased
surface runoff is manifested by higher runoff volume, higher peak discharge and
shorter time of concentration. Kuala Lumpur is one of the areas experiencing rapid
development whereby today more than 80% of its area has been developed (Desa,
1997)8. With the increasing of population, Kuala Lumpur emerging problems are
high water demand and the occurrence of flash flood. Flooding happens quickly and
varies greatly in intensity and duration depending on the prevailing storm patterns.
The flood events are normally localized and usually occur with little or no warning
after heavy rain and they can reach peak level in a short time within 1-3 hours
duration. High rainfall volume may result from different combinations of intensities
and durations. In the early part of the storm period, the storm depths and the
occurrence of high intensity is phenomenon of short convective storm. Therefore, an
investigation in the dynamic short term rainfalls (storm velocity and direction) is
necessary in understanding the hydrological systems.
3
COPYRIGHT © UiTM
Rainfall is an important input in any hydrology analysis where most hydrological
models need a complete and sufficient data for simulation purpose and statistical
analysis. The reliability and quality of the results depend on the quality of the input
data itself. The gaps in the data are quite common for some hydrological stations
especially during major storms due to the instrument failures, absence of observers
or communication breakdown. When the data are missing there should be a method
to estimate the missing value. However, the methods to fill in these gaps with proxy
information are few and inadequate. The conventional method, NRM and
continuous simulation called Fuzzy Logic, one of the Matrix Laboratory
(MATLAB) toolboxes has been developed to estimate missing value based on
previous intensive research in filling missing record. MATLAB was chosen for its
excellent data visualisation features and its support for simulation of dynamic
system. Fuzzy logic theory is based on mathematical equation and linguistic
variable has provided a methodology for computing to replace the conventional
method and currently used in many applications especially in determining the
missing value. This technique is a qualitative modeling scheme by which one
describes the hydrological behavior of catchments from the rainfall characteristics
data as an input by using a natural language. We can know the cause and effect of
certain phenomenon by constructing mathematical relationship based on observed
trends.
An adaptive FRBM proposed in this study deals with using observational data
obtained from adjacent stations for the reconstruction of missing rainfall records.
Initially, the most important aspect in FRBM is to determine their rules. To
construct a fuzzy model, the main problem is to establish a relationship between
missing data station and stations complete data to form the fuzzy rules. The number
of fuzzy rules is rapidly increasing when more parameters are considered in the
model. To overcome this problem the membership functions are constructed to
support the whole parameter involved in the model. The simple fuzzy rules are
classified using a simple human language.
4 COPYRIGHT © UiTM
1.3 Objective of the study
An initial attempt has been made to include a new knowledge on the characteristics o
and movement of storm cells by Desa(1997) but there remain a knowledge gap in
estimating missing rainfall data based on supplied and observed information. Hence
the primary purpose of this study is to collect and analyse short term rainfall data in
small urban catchment, i.e. Kerayong River (see Figure 3.1). The main objectives
are:-
a) To determine missing short term rainfall intensities and to reconstruct the
incomplete data using NRM and FRBM;
b) To study the spatial and temporal distribution of short term rainfall data i.e.
to collect, process and analyze rainfall data by plotting the storm event at one
minute interval; and
c) To apply fuzzy logic in the analysis and simulation of existing and collected
data.
1.4 Scope of the study
The scope of the study is to collect and process short term rainfall data both in space
and time for Kerayong River catchment. This entails the establishment of four rain
gauge stations and five monitoring stations to capture extreme events. Each of the
rain gauges is installed with 0.5 mm tipping resolution and 203 mm of funnel
diameter rain gauges. All the stations are set to event mode. Each time the rain
gauge bucket tips, a signal is sent to the data logger and recorded in the non-volatile
ring memory. For water level, the shaft encoder is designed to interface the data
logger and send the signal at each time the encoder mechanism rotate. The raw data
for rainfall and water level were collected and retrieved using Static Random Access
Memory (SRAM) card at each time site visit has been done. A regular site visit was
conducted at least once a month to inspect the catchments condition in respect to any
5 COPYRIGHT © UiTM
land use changes introduced. This activity was done in conjunction with retrieval of
data from loggers and subsequent resetting of this logger. Hydro reader software is
used to read all the hydrological data stored in SRAM card and it capable to convert
the data files into Micro-Time Dependent Data Analysis (TIDEDA) file and main
frame TIDEDA format. The collected data then was processed using TIDEDA
program which is used by the DID Malaysia. This program was used to store the
database and analyse the hydrological data to any given interval of time. For this
study one minutes interval is used to construct as temporal pattern of the storm
event. The short term rainfall characteristics of Kerayong River catchment
especially the spatial and temporal distribution are studied and analysed in this
research. Other relevant information such as catchment physically details and land
used are obtained from the Department of Survey and Mapping Malaysia. Accuracy
of the features is counter checked by inspection in the catchments.
The campaign for data collection for Kerayong River was started in 1998 until now.
The historical rainfall data are obtained from the hydrological data bank of DID,
Malaysia. For this study, the rainfall data used for the analyses is from June 2001
until June 2003 because within this period the rainfall data is available for all
stations. There are real missing rainfall data occurred at one of the station i.e. Kg.
Cheras Baru station from 15th January to 16th May 2003. To estimate the missing
value, there are two approaches adopted i.e. NRM and FRBM. In order to have an
accurate model, both methods need to be classified by using training data. The daily
rainfall data is used as an input parameter and the numbers of training data chosen is
selected in randomly. A total of 420 daily rainfall values i.e. 150 rainfall data for
year 2001 and 270 rainfall data for year 2002 is employed to construct the FRBM
and NRM. The rules are adopted from the available daily rainfall record with the
simple assumption in FRBM analysis. In training session, the missing data at Kg.
Cheras Baru station are assumed missing and unseen during the training sets and
estimated using the complete time series at the nearby stations. The models need to
be evaluated by testing the model against the testing data to know the accurate
results. If the model cannot give an accurate result, the models parameters need to be
6
COPYRIGHT © UiTM
adjusted and classify the unseen data according to the model until the best estimated
results performed. The results of both models should always be checked, and to
measure the model efficiency in estimating the missing data, the Goodness of Fit test
i.e. standard error (SE) and coefficient of determination (r2) are calculated.
1.5 Significance of the research
The continuous and accurate records of rainfall data is an important parameter for
hydrological model. The dynamics of short term rainfall can play an important role
in designing the drainage network particularly in small urban catchment. This issue
has been a subject of intense discussion by water resources agencies and researchers.
The present knowledge in evaluating, understanding and defining spatial and
temporal variability is insufficient to progress further in filling missing rainfall data.
This study described and developed various methods for infilling the missing
rainfall data i.e. NRM and Fuzzy logic. The purpose of this study was to explore the
potential of fuzzy logic as a new approach in dealing with missing rainfall data. The
significance of this study is therefore to arrive at new technique and approach i.e.
FRBM to fill in missing data in proper way. It is a good tool and very useful in
manipulating information, in a situation that is difficult to be described by
mathematical models. Therefore this research is to explore the potential viability of
the alternative approach based on fuzzy logic system which is a significant step for
future work.
7 COPYRIGHT © UiTM