Fitting seasonal stochastic models to inflows of bekhme reservoir

Abstract EN

In this study time series analysis is applied to records of mean inflow to Bekhme reservoir in the north of Iraq for the period from water year 1933/1934 to water year 2001/2002.

Three Box-Jenkins seasonal multiplicative models are fitted to this series.

These are the (1, 1, 0)x(1, 1, 0)12, (2, 1, 0)x(1, 1, 0)12, and (0, 1, 1)x(0, 1, 1)12 models.

The unconditional sum of squares method is used to estimate the parameters of the models and to compute the sum of squared errors for each of the fitted model.

It is found that the model which corresponds to the minimum sum of squared errors is the (0, 1, 1)x(0, 1, 1)12 model with moving average parameters θ= 0.378 and Θ= 0.953.

The adequacy of this model is checked by plotting the normalized cumulative periodogram which does not indicate nonrandomness of the residuals.

The 95 and 99 percent confidence regions for the model parameters θ and Θ are shown graphically.

Forecasts of monthly inflow for the period from October, 2002, to September, 2006 are graphically compared with observed inflows for the same period and since agreement is very good adequacy of the selected model is confirmed.