Uneven Spatial Distribution of Arctic Warming in Boreal Winter and CMIP6 Historical Simulation Bias

Abstract Recent decades have seen extraordinary Arctic warming trends in boreal winter in both observations and simulations. However, the spatial distribution of Arctic warming as well as climate model simulation require assessment. In this study, we compare the spatial distribution of Arctic warming in the ERA5 reanalysis with that in 40 models from the Coupled Model Intercomparison Project Phase 6 (CMIP6) for boreal winters of 1965–2024 to detect and understand the spatial distribution of Arctic warming and its simulation bias in CMIP6 models. Based on the ERA5 reanalysis, the spatial distribution mainly reflects a faster warming in the Atlantic sector (1.24 K/decade) than in the Pacific sector (0.70 K/decade), a pattern of “Atlantic‐faster, Pacific‐slower.” However, CMIP6 models fail to simulate this pattern, underestimating the warming rate in the Atlantic sector and overestimating it in the Pacific sector. The stationary zonal wave‐1 and wave‐2 warming trends dominate the spatial distribution of Arctic warming, with variance contributions of 53.9% and 26.5%, respectively. The simulation bias in CMIP6 models is mainly attributed to the limitations in simulating the amplitude and the phase of wave‐1 as well as the amplitude of wave‐2 warming trends. Furthermore, the spatial distribution of Arctic warming depends on trends in large‐scale atmospheric circulations associated with wave‐1 and wave‐2. The simulation bias is related to simulation errors in these large‐scale atmospheric circulations and their impact on meridional temperature advection. It is suggested that caution is needed when using model outputs to describe the spatial distribution of winter Arctic warming.