Abstract:In this paper, the main regional spatial modes of winter 2 m air temperature in China were obtained by rotating empirical orthogonal function analysis based on historical simulation of global climate models for phase 6 of the Coupled Model Intercomparison Project (CMIP6). Results show that:the main regional spatial modes of winter 2 m air temperature in China based on CMIP6 multi-model ensemble mean can be classified as four types, namely Central-Northwest mode, South mode, Northeast mode and Yunnan-Tibet mode. Among them, Northeast mode and Yunnan-Tibet mode are relatively stable spatial modes, which exist stably in the selected CMIP6 models and observation data, and are the modes with high consistency in previous studies. However, there are divergences in the spatial classification of winter 2 m air temperature in Xinjiang, Northwest, central and South China, and the consistency of the classification of these regions is poor in previous studies. The South mode appears in all the selected CMIP6 models, while the Central-Northwest mode has great divergences within the models. In the observation data of recent 40 years, the spatial distribution of 2 m air temperature can classify as one mode in Xinjiang, Northwest and South China. However, in the CMIP6 multi-model ensemble mean, the spatial distribution of air temperature in Xinjiang and Northwest China is more closely related to the Central Plains of China. The CMIP6 model has good performance in simulating the location of the trough in the middle troposphere, which is directly related to the spatial distribution of low air temperature in winter. When the CMIP6 models and the observed winter temperature show the same spatial mode, the main circulation system in the middle troposphere associated with the mode is also consistent in the observation and CMIP6 models.