Predictability and variability of East African rainfall seasons

Droughts and flooding over East Africa produce large scale humanitarian disasters such as famine. The recent 2010-11 drought led to an estimated 250,000 deaths in the region, whilst flooding also causes deaths, population displacement, and damage to infrastructure. A better understanding of East African rainfall variability, leading to improved seasonal forecasts, could drastically reduce the impact of these events.

The most widely used operational seasonal forecast in the region is the consensus based Greater Horn of Africa Climate Outlook Forum (GHACOF) forecast, produced using a combination of dynamical and statistical model forecasts alongside local knowledge. In this thesis, for the first time, East African rainfall forecasts from GHACOF are compared directly to dynamical seasonal forecasts from the UK Met Office Unified Model, and both are evaluated against observations. Both forecasts appear to show good skill at forecasting the short rains (October-December), whilst poor skill in forecasting the long rains (March-May) is found.

The drivers of variability in the long rains are studied, linking the long rains to zonal winds over the Congo basin on both inter-annual and decadal timescales, with westerly anomalies leading to more rainfall over East Africa. A source of variability in these zonal winds is found to be the North Atlantic Oscillation (NAO). A Rossby wave response in the mid-latitudes to pressure changes during NAO events propagates equatorward, eventually reaching the Congo basin. The Met Office seasonal forecast model is able to represent both the connection between zonal winds over the Congo and rainfall, as well as the NAO Rossby wave mechanism, in its ensemble members. However, the NAO amplitude in the ensemble mean is too small, and so the teleconnection linking the NAO and the long rains in the ensemble mean is hidden by noise, but these results offer hope for future skilful dynamical predictions of the long rains.