Relative performance of CMIP5 and CMIP6 models in simulating rainfall in Peninsular Malaysia

Abstract

This study evaluated the skills of global climate models (GCMs) of the fifth and sixth Coupled Model Intercomparison Project (CMIP5 and CMIP6) in simulating observed rainfall climatology, seasonal variability, and probability distribution function (PDF) in Peninsular Malaysia. Monthly rainfall records of eighty stations for 1975 − 2005 were employed for this purpose. The Kling-Gupta efficiency was applied to estimate GCMs’ skill to reconstruct rainfall climatology and seasonal variability, while Perkins skill score to replicate PDF. The GCMs of individual CMIP were initially ranked based on the individual metric, and then a compromise rating matric was then employed for the grading. Finally, the highest-ranking CMIP6 GCMs were identified and employed for rainfall projections over Peninsular Malaysia for different shared socioeconomic pathways (SSPs). Results revealed higher bias in CMIP6 GCMs than CMIP5 GCMs but the better association in replicating rainfall climatology and seasonal variability. The EC-ERATH was the best performing model in CMIP5, followed by MPI-ESM-LR, FGOALS-g2, and CanESM2. In contrast, MPI-ESM-MR showed the highest skill among CMIP6 models, followed by MPI-ESM-LR, MIROC-ESM, and GFDL-ESM2M. The employment of the most skilled four GMIP6 GCMs in projecting rainfall in the peninsula revealed a non-linear rainfall change for SSPs—an increase in rainfall for SSP1-26 and SSP5-85 and a decrease for SSP2-45 and SSP3-70. Overall, rainfall was projected to increase in the northwest and central south by 10 − 20% and decrease in the northeast and far south by 1 to 30%.