Teleconnection of rainfall time series in the central Nile Basin with sea surface temperature

We investigated teleconnections of rainfall time series in the central Nile Basin (Sudan and South Sudan) with localities in the global sea surface temperature (SST) field, using monthly rainfall data from 11 gauging stations from 1960 to 1999. Annual rainfall ranged from 100 mm in the north to more than 700 mm in the south, and all stations had a strong contrast between rainy and dry seasons with rainless dry periods of several months. Rainfall time series at the stations were categorized as strongly seasonal, with precipitation concentration index exceeding 16 and seasonality index exceeding 0.9. The rainfall stations were classified into four zones on the basis of annual rainfall, seasonality, and cross-correlations among the stations. We calculated cross-correlations of interannual rainfall time series in summer (July and August) with the global SST field. For short lag times (0 or 1 month), summer rainfall in Zones I and II (northern arid regions) had significant correlations with SST over the eastern Mediterranean Sea and southern Indian Ocean, summer rainfall in Zone III (semiarid region) had significant negative correlations with SST over the Indian Ocean, and summer rainfall in Zone IV (southern wet region) had significant correlations with SST over tropical areas and the southwestern Pacific Ocean. For long lag times (3–6 months), Nile Basin summer rainfall time series had significant correlations with SST in various regions of the Atlantic and Indian Oceans but not the Pacific Ocean. Rainfall in Zones I and II had positive correlations (significance level?<?0.01) with SST south of Greenland and around the Azores Islands and negative correlations with SST south of Madagascar; rainfall in Zone III had negative correlations with SST in parts of the Indian Ocean; and rainfall in Zone IV had significant positive correlations with SST southwest of South Africa and negative correlations with SST in the southwestern Indian Ocean. In sum, rainfall in three of the zones (I, II, and IV) had significant positive and negative correlations with SST in parts of the Indian and Atlantic Oceans. For each of these zones, one positive correlation and one negative correlation were selected and correlations with the time series of the difference between the two SST records were calculated. Correlations of Nile Basin rainfall with the SST differences were stronger than the original positive and negative correlations. The resulting time series of SST difference were applied to an artificial neural network to predict summer rainfall, yielding satisfactory correlation coefficients between the observed and predicted summer rainfall (r?>?0.70).