Best Arima Models for Forecasting Inflow of Hit Station

Abstract

Time series analysis for hydrological phenomena hasan important role in water resources engineering. In this study,seven models of ARIMA family are tested for forecasting themonthly discharge at Hit station on Euphrates river in Iraq. Thestatistical analyses were done for models with help of IBM SPSSstatistics 21 software, The number of observations used is equalto 480 reading, start from October 1932 and end at September1972, this period represents the near-natural stream flow of theriver before the construction of dams in Syria and Turkey.Statistical tests such as T-test and F-test were used to detect anychange in Mean and Variance at 95% significant probabilitylevel. Results showed that the best model is (2,0,1)×(0,1,1)12which gives a minimum error and good agreement betweenobserved and forecast discharge.