Development of artificial neural network (ANN) flood model for Jeti Kastam station based on significant rainfall locations using Z-score / Khairah Jaafar

A review of floods modelling in use today shows that new techniques are required that solves the problems of flood prediction and damage estimation. It is generally acknowledged in the environment sciences that the choice of computational model impacts the research results. Such models mainly used to simulate rainfall and water level prediction involve that use an intelligent model. This study aims to explore the use of an Artificial Neural Network (ANN) models to predict the water level at Jeti Kastam. Rainfall and water level were considered as the primary factor influencing the likelihood of flood and a number of ANN architectures were evaluated as flood prediction models. The Kelantan river area is one such region which has been vulnerable to flood disasters. This study predicts river water level from rainfall and water level data. The data collection start with rainfall and water level data provide by Department of Irrigation and Drainage Malaysia (DID). Then it followed by identifying its significant station using Z-score technique. After that, ANN model was developed based on significant data stations of rainfall and water level at Kelantan river. The use of six stations of rainfall and water level along Kelantan River from upstream to downstream which are Kuala Koh, Tualang, Kuala Krai, Kusial, Dataran Air Mulih and Jeti Kastam were coded as S1 to S6. It specifically analysed the historical data of rainfall and water level stations in Kelantan river basin from January 2013 until December 2015. The total of 105,120 rainfall data and water level data were collected per year. It means, the total of 315,360 data of rainfall and water level were used for analysis. Z-score technique was proposed to identify the significant data stations of rainfall and water level in order to increase the performance of ANN training networks model. This study shows the comparison results between with Z-score and without Z-score technique that will affect the performance of ANN prediction model. Z-score technique highlighted that four data stations of rainfall and two data stations of water level stations were recognized. For water level stations, they are Kuala Koh, Tualang, Kuala Krai and Jeti Kastam