Climate regionalization using objective multivariate clustering methods and characterization of climatic regions in Ethiopia

Objective climate regionalization is essential in environmental and climate studies, particularly over regions with complex terrain and meteorological conditions. The aim of this study was to define and characterize homogenous climatic regions over Ethiopia using a combination of principal component analysis (PCA) and K-means clustering as well as PCA and Ward's clustering. We used Climate Hazards Group Infrared Precipitation with Stations (-6 km resolution) and TerraClimate (-4 km resolution) data obtained between 1985 and 2018. Additionally, data from weather stations provided by the National Meteorology Agency of Ethiopia were applied to assess seasonal and annual precipitation and temperature trends across climatic regions in the 1985-2018 period. Homogenous climatic regions were defined by applying PCA-K-means and PCA-Ward's clustering methods on a matrix derived from precipitation and a combination of precipitation and maximum and minimum temperatures. The trends in seasonal rainfall and maximum and minimum temperatures over the respective regions were computed by fitting a linear regression model to each grid cell. Significant differences in the trends were assessed using the Mann-Kendall test. The results show that it is sufficient and reasonable to define four homogeneous climatic regions. These homogeneous climatic regions have distinct annual cycles, seasonal rainfall and temperature trends, and annual rainfall anomalies. The heterogeneity of the climatic regions between the two time windows (1985-2001 and 2002-2018) is negligible, demonstrating the robustness of the regionalization methods. The seasonal rainfall during the short rains has increased by 50 mm/decade in the southwestern region. The mean annual and seasonal temperature have increased between 0.3 and 0.66 degrees C/decade in all climatic regions. Climate regions defined in the present study are reliable and can be used in various studies at both national and regional levels in evaluation of seasonal forecasts and downscaling global forecasts and could facilitate the development of agricultural plans and strategies for food security enhancement.