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.

iii 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 ACKNOWLEDGEMENTS TABLES OF CONTENTS LIST OF TABLES LIST OF FIGURES LIST OF PLATES NOTATION AND ABBREVIATION ii iv vi x xiii xv xvi

CHAPTER 1 : 1.1 1.2 1.3 1.4 1.5 1.6 General

INTRODUCTION 1 3 5 5 7 8

Background and problem statement Objective of the study Scope of the study Significance of the research Assumption and limitation

CHAPTER 2 : LITERATURE REVIEW 2.1 2.2 2.3 2.4 Introduction Rainfall study Missing and reconstruction of data Fuzzy logic theory with other AI approaches 9 10 11 15

CHAPTER 3 : THEORETICAL CONSIDERATION 3.1 3.2 Study area Network establishment 20 22

vi

COPYRIGHT UiTM

3.2.1 3.3

Rainfall stations

22 24 27 27 28 28 29 29 30 31 33 34 34 35 36 36 36 36 37 37 37 38 38 38 39 39 40 41

3.2.2 Water level stations Instrumentation 3.3.1 Tipping bucket-rain gauge 3.3.2 Hydrologger 3.3.3 Static random access memory (SRAM) card 3.3.4 Hydro reader unit 3.3.5 Water level recorder 3.3.6 SEBA strip chart 3.4 3.5 3.6 Data collection and processing Rainfall intensity Missing data analysis 3.6.1 Normal ratio method 3.6.2 Fuzzy logic approach 3.7 Definition of terms in fuzzy logic 3.7.1 Universe of discourse 3.7.2 Fuzzy logic 3.7.3 Fuzzy sets 3.7.4 Membership function 3.7.5 Fuzzy numbers 3.7.6 Fuzzification 3.7.7 Defuzzification 3.7.8 Linguistic variables and linguistic terms 3.8 Fuzzy rule based model (FRBM) 3.8.1 Fuzzification 3.8.2 Fuzzy inference system (FIS) 3.8.3 Fuzzy rule 3.8.4 Defuzzification

CHAPTER 4 : RESEARCH METHODOLOGY 4.1 Hydrological process 42

vii COPYRIGHT UiTM

4.2 4.3 4.4 4.5

4.6

4.7

Site investigation Installation of equipment Data collection, data retrieval and data processing Data analysis 4.5.1 NRM 4.5.2 FRBM Parameter determination 4.6.1 Input and output 4.6.2 Fuzzification 4.6.3 Implication method - fuzzy If-Then rules 4.6.4 Defuzzification Goodness of fit test 4.7.1 Standard error (SE) 4.7.2 Coefficient of determination^ 2 )

43 43 44 45 45 46 48 48 50 54 55 57 57 58

CHAPTER 5 : ANALYSIS RESULTS AND DISCUSSION 5.1 5.2 5.3 5.4 Overall data record Rainy days records Rain days analysis Rainfall temporal pattern 5.4.1 1-min 5.4.2 Daily 5.4.3 Monthly NRM analysis Fuzzy logic analysis 5.6.1 FRBM(I) result 5.6.2 FRBM (II) result Comparison between NRM and FRBM result Discussion of results 5.8.1 Discussion of NRM analysis 5.8.2 Discussion of FRBM(I) analysis 5.8.3 Discussion of FRBM (II) analysis Verification data analysis 59 60 61 63 63 67 70 73 79 79 86 94 98 99 100 101 103

5.5 5.6

5.7 5.8

5.9

viii COPYRIGHT UiTM

CHAPTER 6 : CONCLUSION AND RECOMMENDATION 6.1 6.2 6.3 6.4 Rainfall study Method of infilling data The benefit of method use 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 No. 2.1 3.1 3.2 3.3 3.4 3.5 4.1 4.2 4.3 4.4 4.5 4.6 5.1 5.2 5.3 5.4 5.5 5.6 5.7 5.8 5.9 5.10 5.11 5.12 5.13 5.14

Title

Page No.

Examples of rules to determine the sensitivity degree Details of rain gauges establish in Kerayong River catchment Details of water level establish in Kerayong river catchment TIDEDA features Rainfall and water level data collected Categorization of rainfall intensity by DID The classification of station number Categorization of rainfall depth Categorization of RF The FRBM type with the linguistic variables of input and output variables Fuzzy classes for D5,D4,D3 and Dl Fuzzy classes for RF The portion of missing rainfall data The portion of missing water level data Number of rain days Longest period in days without rain and with rain for the year 2001 Longest period in days without rain and with rain for the year 2002 Longest period in days without rain and with rain for the year 2003 Frequency analysis of rainfall recorded for the year 2001 Frequency analysis of rainfall recorded for the year 2002 Selected temporal pattern of 1-minute intervals Highest intensity of observed rainfall (mm) Highest rainfall record (mm) and date occurrence in subscript Daily average (mm) value according to year Daily average (mm) recorded at each rain gauge for all seasons Monthly rainfall (mm) for the year 2001

16 22 24 32 33 33 49 49 49 50 52 53 59 60 60 61 62 62 62 63 64 67 69 70 70 71

COPYRIGHT UiTM

5.15 5.16 5.17 5.18 5.19 5.20 5.21 5.22 5.23 5.24 5.25 5.26 5.27 5.28 5.29 5.30 5.31 5.32 5.33 5.34 5.35 5.36 5.37 6.1 C-l C-2

Monthly rainfall (mm) for the year 2002 Monthly rainfall (mm) for the year 2003 Statistical characteristic of monthly rainfall data for the year 2002 NRM training results for the year 2001 according to a selected date NRM training results for the year 2002 according to a selected date Statistic between observed and NRM in NE and SW monsoon Accuracy of NRM analysis FRBM (I) results with entire rules for year according to a selected date FRBM (I) results with entire rules for year according to a selected date Statistic between entire rules applied in FRBM (I) and observed data in intermonsoon month Statistic between 46-R of FRBM (I) and observed data in NE and SW monsoon Accuracy of FRBM (I) data analysis First FRBM (II) training data result for the year 2001 (mm) First FRBM (II) training data result for the year 2002 (mm) Final FRBM (II) training result for the year 2001 and year 2002 (mm) Statistic between entire rules applied for FRBM (II) and observed data in intermonsoon month Statistic between 46-R of FRBM (II) and observed data in NE and SW monsoon Accuracy of FRBM (II) data analysis to the observed value Result comparison for the year 2001 and year 2002 (mm) Statistics analysis of comparison result in NE monsoon month Summary of Goodness of Fit Test for year 2001 and year 2002 Estimate missing data with the comparison of input used (mm) Comparison between observed average rainfall and estimated value The benefit of fuzzy logic modeling 'NR' rules system for the determination of rainfall depth at S1 ' VL' rules system for the determination of rainfall depth at S1

71 71 72 73
74

78 78 79 80 82 85 nr 86 87 88 90 93 93 94 97 97 98 105 110

XI

COPYRIGHT UiTM

C-3 C-4 C-5 C-6 E-1 E-2 G-1

'L' rules system for the determination of rainfall depth at SI 'L' rules system for the determination of rainfall depth at SI 'H' rules system for the determination of rainfall depth at SI 'VH' rules system for the determination of rainfall depth at SI Rainfall data at Taman Desa station for year 2002 Water level data at Taman Desa station for year 2002 Infilling missing daily rainfall at Kg. Cheras Baru station for year 2003

xii COPYRIGHT UiTM

LIST OF FIGURES

Figure Title No. 2.1 The overlapping of membership function in FRBM 2.2 2.3 2.4 3.1 3.2 3.3 4.1 4.2 4.3 4.4 4.5 4.6 4.7 4.8 5.1 5.2 5.3 5.4 5.5 5.6 5.7 5.8 The type of overlapping membership functions Membership function for extreme cumulative rainfall (C) and excessive cumulative rainfall ( C ) Membership function for density of population Location of the study area and hydrological stations established in Kerayong River catchment Trapezoidal and triangular fuzzy numbers Structure of FRBM Research methodology chart Schematic diagram of FRBM procedure Membership function of triangular shape Membership function of trapezoidal shape for 'very high' and ' rain' classification Membership function of trapezoidal shape for 'no rain' classification Membership function for D5,D4,D3 and Dl Membership function for RF The graphical example for inference diagram of Mamdani method and defuzzification process Number of rain day for year 2001 and year 2002 Selected temporal pattern for 1 -minute intensities of storm for the year 2001 Selected temporal pattern for 1 -minute intensities of storm for the year 2002 Daily rainfall temporal pattern for all stations in the year 2001 Daily rainfall temporal pattern for all stations in the year 2002 Total monthly rainfall for all stations in year 2002 Plot between observed and NRM in NE monsoon Plot between observed and NRM in SW monsoon

Page No. 13 13 14 16 21 37 39 42 47 51 51 52 53 53 56 61 65 66 68 69 72 75 77

xiii
COPYRIGHT UiTM

LIST OF PLATES

Plate No. 3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9 3.10 3.11 3.12 3.13 B-l B-2 B-3 B-4 B-5 B-6 B-7 B-8 B-9 B-10 Rain gauge at Kg. Cheras Baru Rain gauge at Taman Sg. Besi Rain gauge at Taman Desa

Title

Page No. 22 23 23 24 25 25 26 26 27 28 29 30 31

Telemetry rainfall station at Pandan Indah Water level station at Taman Sg. Besi Water level station at Taman Desa Water level station at Kg. Cheras Baru Stick gauge Tipping bucket with 0.5 mm resolution Hydrologger for logging rainfall and water level data Hydro reader unit Shaft encoder SEBA strip chart The main drain at Taman Sg. Besi station Secondary drain at Taman Desa station Kerayong River at the confluence of Cheras River Downstream of Kerayong River through to Klang River The garbage trap near Sg. Besi Highway Construction work to remove the plan tree During the site visit at Kg. Cheras Baru station to retrieve data The development at Taman Desa station The damages of stick gauge at Taman Miharja station The damages of solar because of human influence

xv COPYRIGHT UiTM

NOTATION AND ABBREVIATION

ANN AND AI Centroid D DOF DID FIS FLS FRBM GUI H HTC / km L LV m mm M MATLAB MFs n N NE NIWA NR NRM

Artificial Neural Network Fuzzy operator Artificial Intelligence Centre of area method Rainfall depth Degree of fulfillment Department of Irrigation and Drainage Fuzzy Inference System Fuzzy Logic System Fuzzy Rule Based Model Graphical User Interface High Humid Tropic Centre intensity (mm/min) kilometer Light Linguistic variable Number of membership function millimeter Medium Matrix laboratory Membership function Number of input Number of training data Northeast monsoon National Institute of Water Atmospheric Research No Rain Normal Ratio Method

xvi COPYRIGHT UiTM

Obs r2 R RF S SE SW SRAM SMART T

TR

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

TIDEDA Trimf Trapmf U

H(x)

VL V
Oi

o
Pi
% X

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