The climate of the East Africa exhibits marked interannual rnuctuations that provokerndroughts or rnooding which lead to enormous impact on socio-economic activities overrnthe region. Therefore, understanding of the mechanisms that produce this variabilityrnand developing both dynamical and statistical approaches for extended range forecastrnand projected information on future climate is of great importance. In the rst partrnof this study observational datasets and a series of Sea Surface Temperature (SST)rnforced Atmospheric General Circulation Model (AGCM) ensemble simulations for thernwhole 20th century are analyzed to investigate the physical mechanism and potentialrnpredictability of East African short rains variability.rnIt is found that there is substantial skill in reproducing the East African shortrnrains variability given the SSTs are known. Consistent with recent previous studies itrnis found that the Indian Ocean (IO) and in particular the western pole of the IndianrnOcean dipole (IOD) play a dominant role for the prediction skill, whereas SST outsidernthe IO play a minor role. The physical mechanism for the western IO inrnuence onrnEast African rainfall in the model is consistent with previous ndings and consists of arngill-type response to a warm (cold) anomaly that induces a westerly (easterly) low-levelrnrnow anomaly over equatorial Africa and leads to moisture rnux convergence (divergence)rnover East Africa. On the other hand a positive El Ni~no-Southern Oscillation (ENSO)rnanomaly leads to a spatially non coherent reducing e ect over parts of East Africa,rnbut the relationship is not strong enough to provide any predictive skill in our model.rnThe East African short rains prediction skill is also analyzed within a model derivedrnpotential predictability framework and it is shown that the actual prediction skill isrnbroadly consistent with the models potential prediction skill. Low frequency variationsrnof the prediction skill are mostly related to SSTs outside the IO region and likely duernto an increased interference of ENSO with the IO inrnuence on East African short rainsrnafter the mid-70s climate shift.rnBased on results from a series of AGCM experiments, the performance of dynamicalrnseasonal forecast systems are evaluated for the prediction of SSTAs over tropical IOrnand short rains anomalies over equatorial East Africa. The evaluation is based on observationalrndatasets and the Asia-Paci c Climate Center (APCC) Ocean-Atmosphere-rnLand coupled Multi-Model Ensemble (MME) retrospective forecasts (hindcasts) usingrncommon years for all models from 1982 to 2005.rnThe coupled climate models ensemble reproduces seasonal characteristics of lowrnlevel wind, the spatial distribution of SON mean rainfall and seasonal climate variationsrnover equatorial East Africa with further improvement in MME mean. Ensemble meanrnof individual coupled models and MME mean also show statistically signi cant skill inrnforecasting sea surface temperatures anomalies (SSTAs) over the western and easternrnparts of the tropical IO, giving signi cant correlation at 99% con dence level for IOD.rnMoreover, ve out of ten coupled models and MME mean show statistically signi cantrnskill in predicting equatorial East Africa short rains. The delity of hindcasts is furtherrnmeasured by Anomaly Correlation Coe cient (ACC) and four models as well as MMErnmean show signi cant skill over East Africa. It is shown that the reproduction of thernobserved variability in the East African region is mainly due to a realistic relationshiprnof East African rainfall with the IOD. Overall, the skill of the dynamical models isrnattributed to the fact that slowly evolving SSTs are the primary source of predictability,rnand to the fact that coupled climate models produce skillful predictions of SON SSTrnanomalies over tropical IO.rnThis study therefore provides insight into interannual rainfall variability and predictabilityrnover East Africa, in view of tropical Indian Ocean-Atmosphere climaternpatterns and underlying mechanisms. In addition, the information on coupled forecastrnsystems will open the possibility of using readily available seasonal forecasts as skillfulrnpredictions of equatorial East Africa short rains. On the whole, the results found inrnthis study will feed into real-time monitoring and forecasting at seasonal to interannualrntimescales to enhance early warning and disaster preparedness activities and minimizernthe impacts of climate-related catastrophes that are prevalent in the region