Eastern Peninsula Malaysia rainfall model identification using Balanced Stochastic Realization Algorithm

Rainfall forecasting is an important component of flood warning to the public especially in tropical climate region like Malaysia. The forecasts can be used to make decisions about whether warnings of floods should be issued to the general public in advance. In this research, this time series data of rainfall is firstly used to obtain the model using the system identification approach. Balanced Stochastic Realization (BSR) subspace algorithm is used to develop a rainfall model for Eastern Peninsula Malaysia site. The model is developed using a rainfall data obtained from Malaysia Meteorological Department (MMD). In this study, rainfall data in 2001 until 2005 from Kota Bharu station is used to simulate the model. The model is then used to estimate rainfall of the same station in 2006 until 2010. The results reveal good model performance and accuracy. The model is further used to estimate the neighboring state city of Kuantan and Kuala Terengganu. From analysis, the model also demonstrates good performance when tested to different data locations. The significant outcome from this study will able to assist further investigation on rainfall forecasting study in Malaysia, specifically, and tropical climatic region, generally.