IMPROVING FORECASTS OF NILE FLOOD USING SST INPUTS IN TFN MODEL

Egypt depends on the Nile River for all of its water resources. Using the streamflows' history, the large fluctuations of the Nile flood cause the best predictions to be unsatisfactory. The purpose of this paper is to stochastically forecast the Nile summer runoff one-season-ahead using, as inputs, an El Nino-southern oscillation (ENSO) sea surface temperatures (SSTs) signal in the East Pacific and SSTs in the South Indian Ocean. Causality between inputs and outputs is established, and a multiple-input transfer function with noise (TFN) model is built for forecasting purposes. The model explains 63% of the variability of the Nile flood with relatively stable parameters. The mean of absolute percentage error of forecasts is 6% calculated on a data set that was not used in the parameter estimation. The model is parsimonious, and its behavior agrees with the most recent studies in climatology. The forecasting ability of the model is high for extreme floods and severe drought years, except when the South Atlantic Ocean displays a strong warm signal opposite to the El Nino-southern oscillation cold signal.