The Stationarity of Lead-Lag Teleconnections with East Africa Rainfall and its Incidence on Seasonal Predictability

Abstract

East Africa experiences large interannual rainfall variations, which can lead to severe droughts, as in 1984 in Ethiopia, or extensive flooding, as in 1997 in Somalia and Kenya. In the last 15 years, significant advances have been made in relating these variations to large-scale ocean-atmosphere anomalies, of which those associated to ENSO (El Nino Southern Oscillation) come first. The additional importance of sea-surface variations in the Tropical Atlantic and Indian Oceans is becoming established. Although 20(th) century East African precipitation series did not experience decadal-scale trends as significant as those found in the sahelian belt, the correlation with ENSO, the Indian monsoon or other large-scale climatic indices is time-dependent. For instance, as in several other tropical regions, we find a marked weakening of the teleconnections with the Southern Oscillation over the years 1920-1950. The assessment and understanding of such changes is an important issue for seasonal rainfall predictability. Using 20- and 30-yr moving correlations, we examine the stationarity of the relationship between selected global and regional climatic indices and east African rainfall time-series (1901-1997) for Ethiopia and Kenya / Uganda. The results are discussed through a consideration of the space and time structures of the signals. Some of the changes are associated with variations in the amplitude of large-scale circulation anomalies, and others with shifts in the mean background climate.