Rainfall is the most important climate element that affects the livelihood and wellbeing of the majority of Ethiopians. June-September, locally known as Kiremt, is the main rainy season in which about 85-95% of the country's food crop is produced. Because the entire agricultural activities and crop production of the nation hinge on the amount and distribution of June-September rainfall, accurate forecasting of this decisive Kiremt season is of extreme importance for mitigating the climate-related catastrophic events the country experiences frequently. This study utilizes for Ethiopia the novel monsoon rainfall prediction approach of Webster and Hoyos (2004, BAMS) that combines wavelet analysis and linear regression techniques to develop empirical models for predicting monthly totals and seasonal anomalies. At the heart of the wave banding technique are identification of the coherent modes of rainfall variability and their spectral separation through inverse wavelet filtering. Extensive diagnostic study of the filtered rainfall and identically wavelet-filtered global sea surface temperature (SST) provides the basis for developing reliable prediction models.

The results of this study show that large-scale SST variability over many parts of the tropical oceans is critically important in explaining Ethiopian rainfall variability. On seasonal to annual time-scales, the tropical Atlantic and the Arabian Sea exert strong influence on Ethiopian rainfall, while SST variations over the equatorial Pacific and the Indian Ocean become important for the biennial and ENSO modes. The results further show strong correlations between model-predicted and observed monthly rainfall totals/standardized all-Ethiopian June-September rainfall anomalies. For monthly rainfall predictions based on orthogonalized global SST with at least 3-month lead-time, the correlation between observed and predicted rainfall totals is +0.90 for 10 independent verification years. Predictions of all-Ethiopian standardized June-September rainfall anomalies also are highly successful. Based on orthogonolized global March SST predictions, the correlation between observed and predicted standardized June-September rainfall anomalies is +0.88 for 10-year retroactive verification.

The fact that the models explain more than 77% of the total rainfall variability when tested on independent data makes the wavelet-based predictions dependable and usable. Supported by knowledge of the requirements of the user community and decision makers, such accurate forecasts of monthly rainfall totals and seasonal anomalies months in advance would greatly help in combating the damaging effects of recurring droughts and reducing the adverse socioeconomic impacts of rainfall variability in Ethiopia.