ΑEvaluation of extreme precipitation over Asia in CMIP6 models

Based on four reanalyses or gridded data sets (ERA5, 20CR, APHRODITE and REGEN), we provide an overview of 23 Historical and 7 HighResMIP experiments’ performance from the Coupled Model Intercomparison Project Phase 6 (CMIP6) (for short, 6-Hist, HighRes) in simulating seven extreme precipitation indices over Asia defned by the Expert Team on Climate Change Detection and Indices (ETCCDI). We compare them with 28 Historical experiments in CMIP5 (5-Hist). CMIP5 and CMIP6 models are generally able to reproduce extreme precipitation’s spatial distribution and their trend patterns in comparison to the benchmark data set (APHRODITE). The overall performance of individual model is summarized by a “portrait” diagram based on four statistics for each index. We divide all 58 models into three groups (A, the top 20%; B, the median 60% and C group, the last 20%) according to MR rankings (the comprehensive ranking measure). Based on the “portrait” diagram and MR rankings, models that perform relatively well for all seven extreme precipitation indices include HadCM3, HadGEM2-AO, HadGEM2-CC and HadGEM2-ES from 5-Hist, EC-Earth3, EC-Earth3-Veg from 6-Hist and ECMWF-IFSHR, ECMWF-IFS-LR, ECMWF-IFS-MR from HighRes. The simulated performance of CMIP6 is polarized, for the top four and the last fve ranking models are both from CMIP6. Compared with the counterpart models in CMIP6 and CMIP5, the improvement of PCC (pattern correlation coefcient) is more obvious. Furthermore, the dry biases of CMIP6 (both 6-Hist and HighRes) in Southern China and India and the wet biases of CMIP6 in Tibet are reduced compared to CMIP5. This may beneft from the improvement that CMIP6 models can capture the characteristics of meridional moisture fux convergence, and improve the overestimation or underestimation of meridional and zonal specifc humidity eddies compared to CMIP5.