Stream Flow Forcasting on Pahang River by Time Series Models, ARMA, ARIMA and SARIMA

Abstract

Stream flow forecasting is essential in resources planning and flood response. However, it is a challenge to generate an accurate forecast. In this study, time series models, ARIMA, ARMA and SARIMA are adopted to develop forecast series of two stations, Temerloh and Lubok Paku of Pahang River. SARIMA was chosen as the best forecasting model. SARIMA (2,0,3)(0,1,0)12 for Temerloh generated forecast series with MAPE of 18.35 % which indicative of good accuracy modelling. Besides, SARIMA (0,0,0)(1,1,2) for Lubok Paku demonstrated its MAPE values of 6.54 % which characterized the model as high accuracy. The ability of SARIMA deal with the seasonality of streamflow had increased the precision of forecasting. Besides, ARMA models revealed a lower MAPE values than ARIMA model due to the over-differencing of ARIMA model that lower its accuracy.