Historical evaluations and simulations of precipitation over East Africa from Rossby centre regional climate model

This study assesses the performance of ten Regional Climate Model (RCMs) from the latest version of Rossby Centre of Atmospheric models (RCA4) in the simulation of precipitation over Greater Horn of Africa (GHA) from 1951–2005. The evaluation was performed against observed data from the Climatic Research Unit (CRU) and Global Precipitation Climatology Centre (GPCC). Results for mean seasonal analyses demonstrate an underestimation of March–May (MAM) and June–September (JJAS) precipitation whilst October to December (OND) precipitation is overestimated. Further assessment on the annual scale depicts underestimation of rainfall. However, the west to east gradient representing heavier to lighter precipitation and bimodal patterns of the north to south rainfall band is well captured by most models. The models fairly reproduce precipitation variability over the southeast region as compared to the northwest parts of the study domain. The mean ensemble invariably outperforms the individual RCA4 models due to its minimal probability deviance in precipitation in each zone and throughout the GHA region. The overall evaluation shows weak correspondence of the model data with observed CRU based on statistical metrics. The top five performing models are: MIROC5, CSIRO, CM5A-MR, MPI-ESM-LR, and EC-EARTH. Large variations of model performance are noted from one model to model, and from one region to the other. The ensemble mean of the outperforming RCMs reproduces the rainfall climatology over study domain with reasonable skill and the findings of this study will be a base for the study of extreme floods/droughts events in the region. •This study appraise the performance ten Rossby Centre of Atmospheric models (RCA4) in the simulation of precipitation over GHA domain during 1951-2005•Results for mean seasonal analyses demonstrate an overestimation of OND and JJAS rainfall whilst an underestimation amidst the MAM season.•Annual scale depicts underestimation of rainfall.•The overall evaluation shows weak correspondence of the model data with observed CRU based on statistical metrics.•The best five models are as follows: MIROC5, CSIRO, CM5A-MR, MPI-ESM-LR, and EC-EARTH.