Future asymmetry of the Arctic stratospheric polar vortex
In the late 21st century, tropospheric warming and stratospheric cooling are expected in both hemispheres (Fig. 1a). However, two major differences are found in both hemispheres: the middle stratospheric cooling above 50 hPa is much weaker in the NH than in the SH, and the lowermost stratospheric cooling at 150–200 hPa is centred on the pole in the SH, whereas in the NH near 60–70°N. The former difference is most pronounced in boreal winter and suggests the competing effects of radiative cooling17,18 and dynamical Arctic warming due to a strengthened Brewer-Dobson circulation14,18,19. While the middle stratospheric cooling over the Arctic is the strongest in summer, the lowermost stratospheric cooling only occurs in autumn and winter (Fig. 1b), which produce the contrasting temperature responses between the lowermost and middle stratosphere.
Interestingly, in the lowermost (150 hPa) and middle (30 hPa) stratosphere, the Arctic cooling is the strongest over eastern Eurasia, whereas there is little model consensus on the cooling over North America (Fig. 1c, d). The future changed temperature difference between Eurasia and North America reaches approximately 5°C at 30 hPa, which is responsible for weaker Arctic stratospheric cooling in winter. In the lowermost stratosphere, the Pacific–North American-like (PNA) teleconnection pattern appears to be dominant (Fig. 1e). In the middle stratosphere, however, the polar vortex strengthens over Eurasia and weakens over North America, with the eastward shift compared with the climatological wave (Fig. 1f), indicating the shift of the polar vortex towards Eurasia2,15 or the enhancement of asymmetry between North America and Eurasia. These changes are also confirmed by a Student’s t-test at the 99% confidence level (Supplementary Fig. 1). The wavenumber 1-like response to climate change has been suggested by some models21,22, but its cause remains unclear.
Simulated polar vortex response to sea surface warming
To explore what induces the asymmetric polar vortex response, we make use of existing simulations23 with an atmospheric general circulation model (AGCM). Although there are many causes that drive polar vortex variability5,6,8–14, our simulations enable to quantify the atmospheric response derived from (1) global SST warming and sea ice loss (GLSST), (2) Arctic sea-ice loss (AICE), (3) midlatitude SST warming (MLSST), and (4) tropical SST warming (LLSST), respectively (see Methods and Supplementary Fig. 2).
First, we discuss the vertical section of the zonal-mean temperature responses (Supplementary Fig. 3). The simulated GLSST response reproduces well the projected temperature response (Fig. 1a), except for the middle stratosphere. The difference between the simulated GLSST and projected model responses is produced by the direct radiative component and ocean-atmosphere coupling23. The polar vortex is weakened by AICE and strengthened by MLSST, which have opposing influences, consistent with previous studies5,6,8–10. The AICE- and MLSST-changed polar vortexes are centred near the pole in the middle stratosphere (Supplementary Fig. 4). The LLSST response captures well the simulated lowermost stratospheric cooling with a peak over 60°–70°N in GLSST, accounting for the projected response. In the GLSST response the lowermost stratospheric cooling from autumn further into spring overestimates the projected response (Fig. 1b), due to AICE and MLSST, whereas the LLSST response reproduces well the seasonal evolution. In late winter, however, strong stratospheric warming descends from the middle to the lower stratosphere and is also slightly visible in GLSST, weakening the polar vortex.
Also in the horizontal distributions, the GLSST and LLSST responses capture well the projected lowermost stratospheric cooling with a peak over eastern Eurasia (Fig. 2a, b compared with Fig. 1c). Surprisingly, in the middle stratosphere, the GLSST and LLSST responses simulate Eurasian cooling and the warming over the North Pacific and North America (Fig. 2c, d). These results demonstrate that tropical SST warming induces the projected asymmetric temperature response (Fig. 1d), independent of the direct radiative cooling effect, Arctic sea ice loss, and midlatitude SST warming (Supplementary Fig. 4). The Eurasian cooling in the lowermost stratosphere is accompanied with a deepened Aleutian low (Fig. 2f) through the PNA pattern (Fig. 2j), corresponding with the projected circulation response (Fig. 1e). In the middle stratosphere, geopotential height responses show significant decreases over Eurasia along the Arctic coast and increases over North America (Fig. 2g, h). The wavenumber 1-like responses reproduce well the projected response with the eastward shift compared to the climatological wave (Fig. 2k, l), leading to the asymmetric polar vortex. These features in GLSST and LLSST correspond well with ENSO impacts on the warming and weakening of the polar vortex; the deepened Aleutian low associated with the positive PNA pattern enhances vertical wave propagation into the stratosphere, which weakens the polar vortex11–14.
Role of the eastern equatorial Pacific warming
Returning to coupled model simulations, in the late 20th century, there are little significant SST correlations with the stratospheric temperatures over North America and Eurasia among the individual models in tropical oceans (Supplementary Fig. 5). In the late 21st century, however, SST correlation with the North American temperature among the individual models positively increases in whole tropical oceans, and the Eurasian temperature negatively correlates with the western and eastern equatorial Pacific SST (Supplementary Fig. 5). These results support our AGCM simulations that tropical ocean warming induces the asymmetric polar vortex. In particular, the eastern equatorial Pacific, which plays the leading role in ENSO-induced PNA pattern, is expected to most significantly warm in the tropical Pacific24.
Figure 3a shows longitude–vertical section of regressed anomalies of temperature averaged 60°–70°N onto the Niño-3 SST, as a representative of the eastern equatorial Pacific14. Eurasian cooling and North American warming are evident, indicating that warmer models in the eastern equatorial Pacific enhance the contrasting temperature response, which is consistent with the projected stratospheric temperature response (Fig. 3b). Remarkably, despite the strong radiative cooling, more than 80% of the individual models simulate stratospheric warming over North America approximately 1°C at 70 hPa (Fig. 3b), which is consistent across the models (Supplementary Fig. 1). The contrasting temperature response to the Niño-3 SST warming is associated with the asymmetric polar vortex (Fig. 3c). The negative geopotential height anomalies display a westward tilt with altitude roughly from the date line between Eurasia and North America, indicative of vertical wave propagation25. This result implies that the planetary wave response to the Niño-3 SST warming interferes constructively with the climatological wave and enhances vertical wave propagation into the middle stratosphere, similar mechanism to recent shift of the polar vortex towards Eurasia15.
To validate the asymmetric polar vortex response to the eastern equatorial Pacific warming, we diagnose anomalous planetary waves using a linear baroclinic model (LBM) (see Methods). When an idealized heating is centred over the equator at 100°W under the present climatology, the LBM response also shows a westward tilt with altitude from the date line, enhances vertical wave propagation into the stratosphere, and shifts the polar vortex towards the Eurasian continent, leading to the asymmetric polar vortex with larger amplitude over North America (Fig. 3d). The eastward shifted and deepened Aleutian low through the PNA pattern plays a role in strengthening the planetary wave (Supplementary Fig. 6). Changes in the location or strength of the tropical thermal forcing can considerably affect wave propagation into the stratosphere11,14. When the idealized heating is centred at 120°W, the LBM response still shows the asymmetric polar vortex, but at 140°W in the central Pacific, only weakens the polar vortex, and at 160°E in the western Pacific, has little influence on the polar vortex (Supplementary Fig. 6). Consequently, while the deepening of the Aleutian low weakens the polar vortex12–14, the eastward shift of the Aleutian low leads to the asymmetry by shifting the polar vortex to Eurasia. The eastward-shifted polar vortex response compared with the climatological wave (Figs. 1f, 2k, l, and 3c) is consistent with the eastward shift of the Aleutian low.
In observations, it is difficult to discern the polar vortex responses to the central and eastern Pacific ENSO14,26. In the future, however, the eastward-shifted PNA pattern with the deepened Aleutian low is expected as a robust change24,27,28. The eastern equatorial Pacific warms faster than the surrounding regions24, exceeding the SST threshold for tropical convection27,29, which in turn induces the eastward-shifted PNA pattern and leads to the asymmetric polar vortex. Indeed, both coupled models and our AGCM simulations indicate equatorial wave response over the eastern tropical Pacific (Supplementary Fig. 7), as a result of increased precipitation in a comparable magnitude with the central equatorial Pacific (Supplementary Fig. 8).
Possible impact on surface climate in the near future
The observed asymmetric polar vortex begins to develop in the late 20th century2,15. The multi-model mean appears to capture recent weakening and shift of the polar vortex, and then the weakening becomes moderate by the 2030s (Fig. 4b), possibly due to the opposing effects of Arctic sea ice loss5,6,9,10 and midlatitude SST warming8 (Supplementary Fig. 4), although unforced internal variability might be sufficiently strong2,4,7. From the 2030s, however, the polar cap height rapidly increases over North America and decreases over Eurasia, asymmetrically enhancing the polar vortex. While Eurasian cooling increasingly becomes strong, North American cooling is suppressed and maintained close to the present temperature by 2100 (Fig. 4a). This rapid change in the 2030s suggests that tropical Pacific warming takes the place of Arctic sea ice loss and midlatitude SST warming as the dominating forcing of the polar vortex.
The stratospheric variations can have a strong impact on surface weather and climate through the stratosphere–troposphere coupling5,6,13,15,30. In the late 21st century, subseasonal surface temperature variability (i.e., cold extremes) is projected to significantly decrease over the mid- to high-latitude NH due to Arctic amplification31. However, the cold extremes are likely to increase during a few decades from the 2030s (see Fig. 4 of ref. 31). Despite recent debate on the impact of Arctic sea ice loss7, in a few decades, the asymmetric polar vortex could act to increase the cold extreme events, such as recent cold-air-outbreak over Eurasia3. Indeed, composite analysis based on AGCM simulations (see Methods) shows that surface cooling (warming) is enhanced over Eurasia associated with a polar vortex shift towards Eurasia (no shift) (Supplementary Fig. 9), basically consistent with the impact of recent shift towards Eurasia on surface temperature15. Our results here provide a new insight into Arctic climate changes and important implications for the near future projection on surface climate. Contrary to highly uncertain future projections on the polar vortex7,18,20, the asymmetry is a robust change to be expected in the future.