Evaluation of CMIP6 historical simulations over IGAD region of Eastern Africa

Abstract

AbstractThe Accuracy of model simulations is critical for climate change and its socio-economic impact. This study evaluated23 Global climate models participating in the Coupled Model Intercomparison Project phase 6 (CMIP6). The main objective was to identify the top 10 best performance models in capturing patterns of rainfall for the 1981–2014 period over the Intergovernmental Authority on Development (IGAD) region of Eastern Africa. The total rainfall, annual cycle, continuous, categorical and Volumatic statistical metrics, scatter plots, Cumulative Distribution Function (CDF), and colored code portrait were used to assess the patterns of total rainfall. Results indicate that most CMIP6 models generally capture the characteristics of the observed climatology pattern of total rainfall, bimodal and unimodal rainfall regimes. The majority of models over Arid and Semi-Arid Lands (ASALs) in Kenya, Somalia, Ethiopia, and Sudan scored the lowest skills, highest bias, and over-estimated rainfall, and lower skills over June–September (JJAS) compared to March–May (MAM) and October-December (OND). Quantitatively, a high percent of bias exceeding 80% scored over ASALs, a high correlation coefficient ranging between 0.6 and 0.7 across Ethiopia’s highlands, and a 5–40 as the lowest Root Mean Squared Error scored over the majority of the region. In addition, 21 out of 23 CMIP6 over-estimated rainfall over most parts of the region. The ACCESS-ESM1-5 and MIROC6 are the most over-estimated models as opposed to CNRM-CM6-1HR as the most model under-estimated rainfall, highest bias, and RMSE values. The regional and sub-national analysis showed it is inconclusive to select best-performed models based on individual metrics and sub-national analysis. Out of 23 models, the INM-CM5-0, HadGEM3-GC31-MM, CMCC-CM2-HR4, IPSL-CM6A-LR, KACE-1-0-G, EC-Earth3, NorESM2-MM, GFDL-ESM4, TaiESM1, and KIOST-ESM are the best 10 performance models over IGAD region. These findings highlight the importance of selecting the best performance models for mapping present and future hotspots and extreme rainfall events over the IGAD region of Eastern Africa.