FLOOD FORECASTING AT KINABATANGAN RIVER, SABAH BY UTILIZING ARTIFICIAL NEURAL NETWORK (ANN) WAN NURULHAFIZAH BT ABD RAZAK Report submitted in fulfillment of the requirements for the award of the degree of Bachelor of Engineering (Hons.) in Civil Engineering Faculty of civil Engineering and Earth Resources UNIVERSITI MALAYSIA PAHANG JUNE 2015 iv ABSTRACT Flood event is among the most influential disaster in Malaysia .Therefore, the developing of flood forecasting model is to minimize the effects of flood and to achieve a model with high accuracy by utilizing Artificial Neural Network (ANN). Artificial Neural Network is a highly non-linear and can capture the complex interactions among input variables in a system without any prior knowledge about the nature of these interactions. Nowadays, ANN is widely used in prediction and forecasting in water resources. The area of study the flood forecasting is carried out at the Balat Station, Kinabatangan River, Sabah where the hourly water level data is collected from Department of Irrigation and Drainage (DID) from year 2000 until 2014. The results indicated that the model develop a highest accuracy is 6 hour time interval for 4000 iteration where the NSC result is 0.996 with lower RMSE 155.341 compared to others iteration and time interval. This modal achieved 100% at allowable error less than 500 mm which is show the prediction of water level. As a conclusion, this model shows high accuracy and water level can be used alone. This can be applied in the real world to give out warning on imminent flood. v ABSTRAK Peristiwa banjir adalah antara bencana yang paling berpengaruh di Malaysia. Oleh itu, membangunkan model ramalan banjir adalah salah satu cara bagi mengurangkan kesan banjir dan mencapai model dengan ketepatan yang tinggi dengan menggunakan rangkaian neural tiruan atau artificial neural network (ANN). ANN adalah tidak linear dan boleh menangkap interaksi kompleks antara pembolehubah input dalam sistem tanpa apa-apa pengetahuan terlebih dahulu tentang sifat interaksi ini. Pada masa kini, ANN digunakan secara meluas dalam ramalan dan ramalan dalam sumber air. Kawasan kajian peramalan banjir dijalankan di Stesen Balat, Kinabatangan River, Sabah di mana data paras air setiap jam dikumpulkan daripada Jabatan Pengairan dan Saliran (JPS) dari tahun 2000 hingga 2014. Hasil kajian menunjukkan bahawa model membangunkan ketepatan tertinggi adalah selang masa 6 jam untuk lelaran 4000 di mana hasil NSC ialah 0.996 dengan rendah RMSE 155.341 berbanding dengan lelaran serta selang masa yang lain. Modal ini mencapai 100% pada kesilapan yang dibenarkan kurang daripada 500 mm yang menunjukkan ramalan paras air. Kesimpulannya, model ini menunjukkan ketepatan yang tinggi dan boleh menggunakan paras air sahaja. Ini juga boleh digunakan dalam dunia sebenar untuk memberi amaran pada banjir yang akan berlaku vi TABLE OF CONTENTS Page SUPERVISOR’S DECLARATION i STUDENT DECLARATION ii ACKNOWLEDGEMENT iii ABSTRACT iv ABSTRAK v TABLE OF CONTENTS vi LIST OF FIGURES ix LIST OF TABLES x CHAPTER 1 INTRODUCTION 1 1.1 Background 1 1.2 Problems Statement 4 1.3 Objectives 4 1.4 Scope of Study 5 1.5 Research Significant 5 vii CHAPTER 2 LITERATURE REVIEWS 6 2.1 Introduction 6 2.2 Water Cycle 7 2.2.1 Precipitation 7 2.2.2 Runoff 7 2.2.3 Process of Water Cycle 7 2.3 Types and Causes of Flood 8 2.4 Impacts of Flood 9 2.4.1 Impacts on Human 9 2.4.2 Impacts on Property and Infrastructure 9 2.4.3 Impacts on Environment 9 2.5 Flood Forecasting Malaysia 10 2.5.1 Flood Forecasting System 10 2.6 Artificial Neural Network (ANN) 12 CHAPTER 3 METHODOLOGY 14 3.1 Study Area and Data Collection 14 3.2 Artificial Neural Network 16 3.3 Modeling Method 17 3.4 Flow Chart 19 viii CHAPTER 4 RESULTS AND DISCUSSION 20 4.1 1000 Iteration 20 4.1.1 3 Hours Time Interval 20 4.1.2 6 Hours Time Interval 23 4.2 4000 Iteration 26 4.2.1 3 Hours Time Interval 26 4.2.2 6 Hours Time Interval 29 4.3 Discussion 32 CHAPTER 5 CONCLUSIONS 33 REFERENCES 35 APPENDICES 36 A Approval Letter from Faculty 36 B Data Supply Form from DID 37 ix LIST OF FIGURES Figure. No Title Page 1.0 Flood Prone Areas in East Malaysia 2 1.1 Flooding In Kinabatangan, Sabah 3 2.0 Water Cycle 7 2.1 Classification of Water Level at Flood 11 Warning Centre, Malaysia 2.2 McCulloch Pitts Neuron (1943) 12 3.1 Kinabatangan River, Sabah (DID) 15 3.2 Water Level Elevation statuses at 15 Kinabatangan River Section 3.3 Neuron Scheme 16 3.4 Feed Forward Diagram 17 4.1 (a) Forecasting Performances Modal for 4-4-1 22 4.1 (b) Forecasting Performances Modal 23 4.2 (a) Forecasting Performances Modal for 6-6-1 24 4.2 (b) Forecasting Performance Modal 25 4.3(a) Forecasting Performances Model for 7-7-1 27 4.3(b) Forecasting Performance Model 28 4.4 (a) Forecasting Performances Model for 5-5-1 30 4.4 (b) Forecasting Performances Model 31 x LIST OF TABLES Table No Title Page 4.1 Forecasting Performances Result for Six ANN 20 Architectures for 3 Hours Time Interval and 1000 Iteration 4.2 Forecasting Performances Result for Six ANN 23 Architectures for 6 Hours Time Interval and 1000 Iteration 4.3 Forecasting Performances Result for Six ANN 26 Architectures for 3 Hours Time Interval and 4000 Iteration 4.4 Forecasting Performances Result for Six ANN 29 Architectures for 6 Hours Time Interval and 4000 Iteration 1 CHAPTER 1 INTRODUCTION 1.1 BACKGROUND Flood is the most common occurrence in the Malaysia and it also known as the worst flood especially in the state of Pahang, Terengganu and Sabah. This event was affecting in term of population, property, financial cost, frequency and social economic damaged. Flood also may cause the death, damaged and destructive. The population in Malaysia about 30 million of people and the area of Malaysia is about 330,000km2. In Peninsular Malaysia, the average mean rainfall is 2,400 mm while Sabah is about 2,600 mm and 3.800mm for Sarawak (DID, 2010). Malaysia has 189 basins including Sabah and Sarawak with the average mean rainfall over 2,000 mm to 4,000 mm per year. It is also mostly exposed to the flood event where the basin receives the amount of rainfall that may cause the basin overflow to the surrounding of area. Flood is a natural disaster event and temporary condition whether partial or completes inundation of normally dry land areas from overflow of inland or tidal waters from the unusual and rapid accumulation or runoff of surface waters from any source where flood is occurred from the combination of heavy rainfall that causing the rive or streams to overflow from their banks. 2 Sabah is the largest state of Malaysia after Sarawak with the area about 73,631 km2 with the population about 3, 117,405 of people. Sabah is a state enriched with the natural nature but it is also one of the commonly occurs of flood. Kinabatangan, Sabah is have been known as a very serious flooding occurred such as village of Balat, Barek Manis, Tangkulap and Bukit Garam due to human error in term of the development of infrastructure. Every year, people at Kinabatangan faced the same problems where the landslide also occurs and it is the one of effect from flood. Figure 1.0: Flood Prone Areas in East Malaysia