Bias evaluation in rainfall over Southeast Asia in CMIP6 models

This study evaluated the bias in rainfall over Southeast Asia in the historical simulations of the sixth phase of Coupled Model Intercomparison Project (CMIP6). Using 48 CMIP6 models, a suite of evaluation methods was applied to examine in detail - annual meridional march, annual climatology, long-term trend, and interannual variability. For this study, multi-source observational datasets were employed for comparisons against model simulations. The biases in the climatological state were decomposed into ensemble biases and the leading modes of inter-model uncertainty. The CMIP6 models could simulate well the climatological state of meridional march and annual mean with high pattern correlation coefficients of 0.88 and 0.59, respectively. However, the per-formance is poor in simulating the long-term trend that exhibits a correlation coefficient of 0.05 between simulation and observation in spatial patterns. The CMIP6 models could only reproduce about 23% of inter -annual variability. This evaluation not only revealed the systematic biases in 48 CMIP6 models but also illus-trated the biases in the different Empirical Orthogonal Function (EOF) modes of inter-model uncertainty and the relative performance ranking for a particular model. Based on the total performance index, the best 3 models are EC-Earth3, EC-Earth3-Veg, E3SM-1-0 and relatively poor performance were seen in INM-CM4-8, AWI-ESM-1-1-LR and MCM-UA-1-0. The study provides an assessment of the overall uncertainties in CMIP6 models in the historical simulations and could also be a useful feedback to the global climate modellers in optimizing the models for better performance.