Evaluation of regional climate models performance in simulating rainfall climatology of Jemma sub-basin, Upper Blue Nile Basin, Ethiopia

Gebrekidan Worku, Addis Ababa University
Ermias Teferi, Addis Ababa University
Amare Bantider, Addis Ababa University
Yihun T. Dile, Texas A&M AgriLife
Meron Teferi Taye, Addis Ababa University


This study examines the performance of 10 Regional Climate Model (RCM) outputs which are dynamically downscaled from the fifth phase of Coupled Model Inter-comparison Project (CMIP5) GCMs using different RCMs parameterization approaches. The RCMs are evaluated based on their ability to reproduce the magnitude and pattern of monthly and annual rainfall, characteristics of rainfall events and variability related to Sea Surface Temperature (SST) for the period 1981–2005. The outputs of all RCMs showed wet bias, particularly in the higher elevation areas of the sub-basin. Wet bias of annual rainfall ranges from 9.60% in CCLM4 (HadGEM2-ES) model to 110.9% in RCA4 (EC-EARTH) model. JJAS (June-September) rainfall is also characterized by wet bias ranges from 0.76% in REMO (MPI-ESM-LR) model to 100.7% in RCA4 (HadGEM2-ES) model. GCMs that were dynamically downscaled through REMO (Max Planck Institute) and CCLM4 (Climate Limited-Area Modeling) performed better in capturing the rainfall climatology and distribution of rainfall events. However, GCMs dynamically downscaled using RCA4 (SMHI Rossby Center Regional Atmospheric Model) were characterized by overestimation and there are more extreme rainfall events in the cumulative distribution. Most of the RCMs’ rainfall over the sub-basin showed a teleconnection with Sea Surface Temperature (SST) of CMIP5 GCMs in the Pacific and Indian Oceans, but weak. The ensemble mean of all 10 RCMs simulations was superior in capturing the seasonal pattern of the rainfall and had better correlation with observed annual (Correl = 0.6) and JJAS season rainfall (Correl = 0.5) than any single model (S-RCM). We recommend using GCMs downscaled using REMO and CCLM4 RCMs and stations based statistical bias correction to manage elevation based biases of RCMs in the Upper Blue Nile Basin, specifically in the Jemma sub-basin.