University of Wisconsin–Madison

Predicting the 2015 Drought in Ethiopia

Partners: Johns Hopkins University

Funding Agency: NSF-PIRE

Ethiopia is the second most populous country in Africa, next to Nigeria with a population doubling in every 20 years. Per the CIA fact book (2017), agriculture employs about 80% of the Ethiopian population. In Ethiopia, more than 90% of crop production is rain-fed. Failure of precipitation during crop growing seasons: Belg (March – May) and Kiremt (June – September), often result in food shortages. The 2015 drought for example, leaves more than 10 million people at risk of food shortage because of below normal precipitations during the Belg and Kiremt seasons. Scientific research conducted since mid-1980’s identified association of El Niño Southern Oscillation (ENSO) with the Kiremt season precipitation. Figure 1 shows correlation of at JJA ENSO with JJAS precipitation at different lead times and the strongest correlation is at concurrent time (Figure 1d). The correlation map is developed by using the CHIRPS precipitation data set and Niño 3.4 index.

Figure 1: Correlation maps of JJAS precipitation with different lag times of ENSO.

Figure 1: Correlation maps of JJAS precipitation with different lag times of ENSO.

The region shown by a dashed line in Figure 1d (referred to here as study area) is ENSO sensitive. Below normal precipitation magnitudes both during the Belg and Kiremt seasons in this region are associated with food shortage in the country. A simple linear precipitation prediction model based on the JJA Niño 3.4 as a predictor is developed to predict the Kiremt season precipitation. Ensemble JJA Niño3.4 predictions constitute dynamical and statistical models, issued in May by the International Research Institute (IRI), are used to predict the JJAS precipitation. The precipitation prediction model seems to perform well when the predictions are compared with the CHIRPS data set over the same study area. The prediction ensemble of the 2015 Kiremt precipitation by this model shows dry precipitation conditions with more than 75% probability. The model’s performance when tested from 2003 to 2005 exhibits a correlation coefficient of more than 0.85. Considering the prediction skill of the model, its usability for forecast based decision making is promising. A decision framework which incorporated the Belg season observed precipitation coupled with the prediction of the Kiremt season in May provides information for decision makers to allocate and mobilize appropriate resources 5 months ahead to avert the likely humanitarian catastrophe.