Gauging the performance of CMIP5 historical simulation in reproducing observed gauge rainfall over Kenya

Abstract

Abstract The detrimental impacts of climate change have drawn the interest of many climate scientists toward understanding the past climate for the sake of preparing for the future. The current study evaluates the performance of 12 rainfall models available in Coupled Model Intercomparison Project Phase 5 (CMIP5) in reproducing observed rainfall over Kenya from 1979 to 2005. Several statistical metrics were deployed in quantifying the disparities between CMIP5 models, in situ, and the Global Precipitation Climatology Centre (GPCC v7) rainfall datasets. The results show satisfactory skill of CMIP5 models' in simulating the bimodal rainfall regime despite exhibiting dry (wet) bias during March–May (MAM) and October–November (OND) season, respectively. The models' skills in reproducing the interannual variability is relatively weak. However, majority of the models captures the temporal pattern with reasonable skills in OND than in MAM and annual rainfall. The impacts of Indian Ocean Dipole (IOD) and El Nino southern oscillation (ENSO) are marked in observed OND rains with no significance link in MAM rains. Overall, CMIP5 models' skills in replicating the mean statistics and teleconnection links are relatively weak. Remarkably, the performance of models at different time scale, metrics, simulation of dynamical and teleconnection patterns are inconsistent among models. Nevertheless, based on skill score, the models are listed from top to bottom as; MPI-ESM-MR, CSIRO-MK3-6-0, GISS-E2-R, MRI-CGCM2.3.3, EC-EARTH, MIROC-ESM-CHEM, FGOALS-g2, BCC-CSM1.1-M, HADGEM-AO, CanESM2, GFDL-ESM2G, IPSL-CM5A-MR, and MME model. This study sheds light on the use of statistical metrics and teleconnection pattern to rank CMIP5 models and forms basis on model selection. Model parameterization over tropics is prudent, and bias correction is paramount in future projections and impacts studies.