Mapping drought evolution in Ethiopia: trends, clustering, and Bayesian estimation of abrupt changes

Recent climate change has significantly impacted Ethiopian regions, leading to substantial alterations in precipitation patterns, with profound implications for local ecosystems and water resources. This study comprehensively analyzes the Standardized Precipitation-Evapotranspiration Index (SPEI) at 6, 12, and 24-month time scales across Ethiopia to explore drought trends. The Seasonal Kendall (SK) test, applied to SPEI data, reveals significant trends indicating increasing drought conditions in several regions. Specifically, 71.8% of data points for SPEI-24 show significant negative trends, which indicate more frequent and severe droughts, particularly in northwestern, western, and southern Ethiopia. In contrast, 9.8% of data points exhibit significant positive trends, suggesting a reduction in drought severity, mostly in northern and mid-western areas. The remaining 18.4% exhibit non-significant trends. Two clustering algorithms, K-means and Hierarchical, were used to identify homogeneous regions in terms of drought trends. Both algorithms divided Ethiopia into four clusters: Cluster C1 (central and northern Ethiopia) showed significant increasing drought trends, while C2 and C4, covering both southern and western Ethiopia, exhibited decreasing trends in drought intensity. Cluster C3 varied between algorithms: in K-means, it acted as a transitional zone, whereas in Hierarchical clustering, it was the westernmost cluster with a distinct decreasing trend. To assess abrupt changes in SPEI, the Bayesian Estimator of Abrupt Change, Seasonal Change, and Trend (BEAST) algorithm was used, identifying abrupt drought events such as those in 1984, 1999, and 2002/2003. These changes highlight historical drought variability across different regions of Ethiopia, offering insights into long-term climate impacts.