Performance of the CORDEX-Africa Regional Climate Model in capturing precipitation and air temperature conditions in the Omo Gibe River Basin, Ethiopia

Abstract Using regional climate models (RCMs) and ensembles of multiple model simulation outputs without assessing their modeling performance did not always ensure the best agreement between observed and modeled climate variables. To this end, assessing the modeling performance of regional climate models (RCMs) is indispensable in selecting the most effective model to use for climate change impact studies. In this study, the performance of ten Coordinated Regional Climate Downscaling Experiments (CORDEX) in Africa was examined against observational datasets from 1986 to 2005 across the entire Omo Gibe River Basin (OGRB). The output of RCMs was evaluated based on their ability to reproduce the magnitude and pattern of monthly, seasonal, and annual precipitation and air temperature, precipitation characteristics, and statistical metrics. The results confirm the difference between RCMs in capturing climate conditions at both spatial and temporal scales. The spatial pattern of mean annual precipitation was better reproduced by the ensemble mean and RACMO22T (EC-EARTH). CCLM4-8-17 (MPI) and the ensemble mean reproduced the annual patterns of observed precipitation, even though the amounts were different. Except for peak precipitation, all RCMs simulated seasonal precipitation, and the pattern was reasonably captured by RACMO22T (EC-EARTH), CCLM4-8-17 (CNRM), RCA4 (CNRM), CCLM4-8-17 (MPI), and REMO2009 (MPI). The interannual and seasonal variability of precipitation was higher than the variability of air temperature. It was found that observed and RCM precipitation simulations using CCLM4-8-17 (MPI), REMO2009 (MPI), and RCA4 (CNRM) showed better agreement at several individual stations in the Omo Gibe River Basin (OGRB. Likewise, RCA4 (MPI) and CCLM4-8-17 (MPI) were superior in capturing minimum and maximum air temperatures. The cumulative distribution of extreme precipitation was better captured by RCA4 (MIROC5), and all RCMs, including their ensemble mean, overestimated the return period. Overall, the study emphasizes that the selection of robust RCMs that better reproduce observed climate conditions and the use of multi-model ensembles of models with the best performance after systematic bias correction are fundamentally necessary for any study of climate change impacts and adaptation in the OGRB.