Abstract
The variability of East African short rains (October-December) has profound socioeconomic and environmental impacts on the region, making accurate seasonal rainfall predictions essential. We evaluated the predictability of East African short rains using model ensembles from the multi-system seasonal retrospective forecasts from the Copernicus Climate Change Service (C3S). We assess the prediction skill for 1- to 5-month lead times using forecasts initialized in September for each year from 1993 to 2016. Although most models exhibit significant mean rainfall biases, they generally show skill in predicting OND (October-December) precipitation anomalies across much of East Africa. However, skill is low or absent in some northern and western parts of the focus area. Along the East African coasts near Somalia and over parts of the western Indian Ocean, models demonstrate skill throughout the late winter (up to December-February), likely due to the persistence of sea surface temperature anomalies in the western Indian Ocean. Years when models consistently outperform persistence forecasts typically align with the mature phases of El Niño Southern Oscillation (ENSO) and/or Indian Ocean Dipole (IOD). This latter mode, when tracked using the Dipole Mode Index, is generally able to predict the sign of the rainfall anomaly in all models. Despite East Africa's proximity to the west pole of the IOD, the correlation between short rains and IOD maximizes when both east and west are considered. This finding confirms previous studies based on observational datasets, which indicate that broader-scale IOD variability associated with changes in the Walker Circulation, rather than local SST fluctuations, is the primary driver behind East African rainfall.