Regionalization and Prediction of Seasonal Precipitation in Ethiopia

Abstract:

Rainfed agriculture continues to be an important part of Ethiopia’s livelihoods and economy. Highly variable inter-annual precipitation, however, presents a serious challenge to sustainable production and subsistence survival. An improved understanding of what drives hydroclimatic extremes and an effective prediction system may help to buffer resulting impacts through improved decision-making. Precipitation data from the National Meteorological Agency at 0.1 x 0.1 grids for 1983 – 2011 during the June-September rainy season over western Ethiopia is evaluated through a cluster analysis to investigate homogeneous regions with similar rainfall patterns for subsequent prediction of seasonal precipitation for each region. A k-means clustering method is applied with the optimal number of clusters (K) selected by the within cluster sum of square errors (WSS) metric. Homogenous regions are defined with relatively clear and smooth boundaries, low inter-cluster correlations, and high intra-cluster correlations. The precipitation prediction models are statistically based, with a seasonal total prediction for each cluster; grid-based predictions are subsequently conditioned on the cluster level prediction through regression. Prospective model predictors include large-scale ocean-land-atmospheric climate variables and local variables and conditions. These predictions will be used in economic and water management models.