Forecasting the annual flow rate of the Tigris river using stochastic modelling

Abstract

The monthly and annual time series of the flow of the Tigris River at Al-Amara barrage was analyzed in order to forecast the annual and monthly future values of discharge of that river. The monthly series from 1980-2010 was used as data for this analysis. Non-homogeneity was detected in the series with significant positive jump’s observed in the periods 1987-1988 and 1992-1997. The non-homogeneity was removed using the method suggested by Reference [1]. A Box and Cox [2] transformation was then used to normalize the homogeneous series. The dependent stochastic component was obtained from the series after removing the periodic component, which was observed using harmonic analysis after fitting the normalized series. A first-order auto-regressive model (Markovian chain) was then used to model the last obtained component. The data in the period 1980-2000 was used to conduct that analysis. For generation of a future series, the model was verified using the remaining 9-years, 2001-2010. Some statistical properties were obtained for both the forecasted and the observed series and then used to compare the two series. The comparison indicates that the model is capable of forecasting acceptable future data.