Improved skill of Coupled Model Intercomparison Project phase 6 over phase 5 models in reproducing weather regimes in East Asia

Abstract We evaluate the performance of climate models from the Coupled Model Intercomparison Project phases 5 (CMIP5) and 6 (CMIP6) in simulating the weather regimes over East Asia during the period from 1961 to 2005. The circulation patterns, frequencies and transition probabilities of the weather regimes are obtained by categorizing the daily 500‐hPa geopotential height (Z500) anomalies using the k ‐means clustering algorithm. We identify four winter (WWR1‐4) and three summer weather regimes (SWR1‐3) in both reanalysis datasets and CMIP5/6 models. Significant increasing (declining) frequency trends in WWR3 (WWR4) and SWR2 (SWR1) are found in the ERA5 and NCEP/NCAR reanalysis datasets. On the whole, both generations of models demonstrate reasonable abilities to reproduce the circulation patterns of weather regimes based on the Z500 anomaly fields but overall underestimate the intensities of weather regime patterns, underestimate the occurrence frequencies of WWR3, SWR2, and SWR3, and overestimate the frequencies of WWR1, WWR2, WWR4, and SWR1. In terms of both the ensemble‐mean and pairwise comparisons, the CMIP6 models generally perform better than their CMIP5 counterparts in reproducing the circulation pattern intensities, occurrence frequencies during the whole period and the linear frequency trends and transition probabilities of the weather regimes both in winter and summer. Further analysis shows a highly positive correlation between the total frequency and transition performances but a lower correlation between the biases in the total frequency and intensity of weather regime patterns. The improved skill of CMIP6 over the CMIP5 models in reproducing the weather regimes over East Asia may be partly related to the increased vertical atmospheric resolution in the model.