Evaluation of CMIP6 models for rainfall simulation in Central Eastern Africa using extreme precipitation indices

Accurate simulation of extreme precipitation is crucial for understanding climate variability and informing adaptation techniques in Central Eastern Africa, a region with complex climatic and topographic situations. This study evaluates the performance of 19 CMIP6 models in simulating precipitation and its extremes using five indices: PRCPTOT, CDD, R95p, R20mm, and SDII-based on their spatial and temporal variability during 1981–2014. Using Taylor diagrams for spatial patterns and interannual variability skill for temporal characteristics, we identified the most skilful models (“Good models”) and the least skilful models (“Bad models”). The outcomes show that good models are ACCESS-ESM1-5, EC-Earth3-Veg, HadGEM3-GC31-MM, MRI-ESM2-0, IPSL-CM6A-LR and UKESM1-0-LL model simulations show closer agreement with the observed spatial and temporal patterns of precipitation, these models exhibit smaller biases and narrow uncertainty ranges with smaller absolute errors. Conversely, bad models like CNRM-ESM2-1, MIROC6, and KACE-1-0-G, these models exhibit higher biases with higher mean absolute errors, for used indices. In addition, good models provide reliable spatial distributions and temporal variability, closely matching observations, while bad models deviate significantly, especially in regions with complex terrain. Ensemble analysis reveals that good models offer smaller errors and fewer statistically significant biases, highlighting their potential as benchmarks for improving climate simulations. This study highlights the importance of model selection and improvement for accurately simulating extreme precipitation in CEA, offering brainstorms to improve climate modelling frameworks and inform climate resilience schemes in this exposed region.