Abstract: This study uses ERA5 daily reanalysis data from 1979 to 2019 to objectively classify the tropospheric polar vortex over the Northern Hemisphere during winter through SOM (self-organizing mapping) analysis. Further, the temporal variation characteristics of the tropospheric-polar-vortex weather are analyzed, and the causes of its long-term temporal variations are revealed. Results reveal the following: (1) Depending on the center position, the polar vortex can be classified into four types, namely circumpolar, dipole, Eurasian, and North American, with the circumpolar and dipole types being the predominant circulation types. The circumpolar type shows a significant long-term decrease, while the dipole type exhibits a notable increase, with both demonstrating obvious interannual and interdecadal variability. (2) The long-term decrease in the circumpolar type and the long-term increase in the dipole type are mainly attributed to the rapid warming of the Arctic region, which continuously reduces the meridional temperature gradient and weakens baroclinicity between the middle and high latitudes of the Northern Hemisphere. This results in a weakened circumpolar westerly circulation, making the tropospheric polar vortex in the Northern Hemisphere more prone to splitting. Using data from the pre-industrial control simulation (piControl) and a simulation with a 1% a⁻¹ CO₂ increase (1pctCO₂), which leverages the ocean–atmosphere coupled model (CESM2) for CMIP6 (the coupled model intercomparison project phase 6), SOM explores the relationship between the long-term temporal variability of circumpolar and dipole types and global warming. The analysis reveals that in both piControl and 1pctCO₂, the circumpolar and dipole types remain the dominant circulation patterns. While there is no significant variability trend for these types in piControl, 1pctCO₂ shows a significant decrease in the circumpolar type and an increase in the dipole type. This indicates a strong link between these long-term temporal variability trends and global warming. Concurrently, a comparative analysis using eleven other CMIP6 models confirms the reliability of the obtained results.