Evaluation of CMIP6 model simulations of extreme precipitation in China and comparison with CMIP5

Based on the 32 global climate models that participated in the phase 6 of the Coupled Model Intercomparison Project (CMIP6), 27 global climate models that participated in CMIP5 and the observational dataset CN05.1, this study evaluates the performances of these CMIP6 and CMIP5 models on the simulation of extreme precipitation index over China for 1961—2005. Eight extreme precipitation indices defined by the Expert Team on Climate Change Detection and Indices (ETCCDI) are used to represent extreme precipitation events. Results indicate that the multi-model ensemble (MME) median of CMIP6 overall has lower relative errors of both climatological mean (with an average of 29.94%, reduced by 2.95%) and relative variability (with an average of 10.10%, reduced by 5.45%) than that of CMIP5. Generally, CMIP6 performs better than CMIP5 in simulating climatological condition of China, especially over the arid region (the error was reduced by 12.15% compared to CMIP5). Further analyses suggest that the MME median of CMIP6 has large spatial correlation coefficients and small root-mean-square errors. Based on the Taylor skill (TS) score, both CMIP6 and CMIP5 models are ranked to evaluate relative model performance. CMIP6 models have higher ranks than CMIP5 models, with an average TS score of 0.78 (0.75) for CMIP6 (CMIP5), and four out of the five highest-scored models are CMIP6 models. Regarding the homologous models, the TS scores of CMIP6 models (an average of 0.91) are larger than their earlier versions in CMIP5 (an average of 0.68), indicating a prominent improvement in CMIP6. Further analyses reveal that the performances of CMIP6 models differ in the simulation of extreme precipitation over different regions of China. Generally, compared to CMIP5, CMIP6 models perform better in simulating extreme precipitation events over China.