Abrupt ENSO–NAO Teleconnection Reversal
We begin by examining the regressed Euro-Atlantic (70°–0°W) zonal mean daily sea-level pressure (SLP) anomalies with respect to the Niño-3.4 index from December to January, as shown in Fig. 1a. There clearly exists a prominent NAO phase transition, from positive to negative, in early January of the ENSO warm phase winter. From December 1st to about January 5th (hereafter referred to as P1). The Euro-Atlantic region is characterized by a typical SLP anomaly pattern of a positive NAO phase in response to the El Niño forcing (supplementary Fig. 1a). However, from January 10th until about January 25th (hereafter denoted as P2), the atmospheric pattern abruptly reverses its distribution with evident positive SLP anomalies in the north and negative anomalies in the south, which project onto an evident negative NAO phase (supplementary Fig. 1b). The ENSO-regressed NAO index also shows a consistent feature, with an abrupt phase reversal, around January 8th.
We then display the Euro-Atlantic SLP anomalies during the two super El Niño winters, namely the winters of 1997–98 and 2015–16. It is clear that the anomalies also exhibit an obviously strong NAO phase reversal from positive to negative in early January (Fig. 1b), consistent with the former regression results. It should be noted that we do not show the NAO anomalies during another 1982–83 super El Niño winter because there was a strong El Chichón volcanic eruption that occurred in the spring of 198226. Such an event can exert remarkable impacts on the Euro-Atlantic atmospheric circulation for about 1–2 years after the eruption27,28.
To further consolidate this phenomenon, we adopt a suit of 26 AMIP-style simulations derived from CMIP6 experiments (see details in Methods) to conduct a parallel analysis. The results based on the multi-model ensemble mean (MME) are displayed in Fig. 1c. It is found that the simulated ENSO-regressed SLP pattern and NAO index also feature abrupt NAO phase transitions, from positive to negative, in early January of the El Niño winter. Even the transition timing agrees adequately with the observational counterpart (Fig. 1c). However, given the non-negligible inter-model spreads (supplementary Fig. 2a), we evaluated the performance of each model by comparing the ENSO-related Euro-Atlantic zonal mean (70°–0°W) daily SLP evolution pattern with the observations. We then select the ten best AMIP models (AMIP-B10) based on the pattern correlations because they consistently better reproduce this abrupt NAO phase reversal (supplementary Fig. 2).
It is well known that the NAO is a leading atmospheric mode over the Euro-Atlantic region and is responsible for considerable local weather and climate predictability29,30. The corresponding regional climate anomalies were thus examined. As expected, rapid changes in terms of the anomalous low-level winds and surface air temperatures from P1 to P2 are detected (Fig. 1d). Prominent temperature warming prevails in the regions of eastern Canada, Greenland, and North Africa, whereas strong northerly wind anomalies with rapid cooling occur in western Europe and eastern America. This is a typical climate response to a decrease in the NAO. The aforementioned results imply that this intraseasonal NAO phase reversal signal during ENSO in early January is robust enough to be applied to the operation of regional climate prediction.
Possible Physical Mechanisms
We now turn to explore the possible mechanisms responsible for this NAO phase reversal in response to ENSO forcing. It is unlikely that this abrupt intraseasonal ENSO teleconnection change can be explained by ENSO local SSTs because SST anomalies in the central-eastern tropical Pacific are highly persistent during the boreal winter season31. We first examine ENSO-related stratospheric daily anomalies because the stratosphere has been considered a major medium that leads to the late winter ENSO influence over the Euro-Atlantic sector16,23,24. However, no evident stratospheric signal propagated downward to affect the troposphere within the period from December to January (supplementary Fig. 3a, b). The tropospheric processes modulated by boundary conditions are then analyzed, which mainly include a delayed local TNA SST modulation19,20 and/or Rossby wave trains excited by the ENSO-induced tropical convection that propagates into the Euro-Atlantic region13,14,17. Unfortunately, the spatial patterns and amplitudes of ENSO-related TNA SSTs (supplementary Fig. 3c, d) and tropical convection anomalies (supplementary Fig. 4) during P1 and P2 are quite similar. These results imply that the previously proposed mechanisms, which are used to explain the various ENSO–NAO teleconnections during the early or late winter, are not the initiators of this abrupt ENSO–NAO teleconnection reversal in early January.
Next, we compare the atmospheric anomalies in the troposphere directly. Figure 2 shows the ENSO-regressed, 250-hPa geopotential height anomalies for P1 and P2. To illustrate the wave energy propagation, the associated wave activity flux (WAF) (see details in Methods) is also displayed. It can be seen that during El Niño P1, an evident Rossby wave propagates poleward from the tropical Pacific to the North Pacific, which is associated with the deepening of the Aleutian low. These wave energies continue heading northeastward across North America and then arrive in the North Atlantic region. This gives rise to the negative geopotential height anomalies around Iceland, which project onto a positive NAO pattern (Fig. 2a). The picture is almost the same during El Niño P2, but the Rossby wave characteristics in the North America–North Atlantic sector are dramatically different from those in El Niño P1. The positive geopotential height anomalies over North America during P2 head southeastward rather than northeastward before it enters the North Atlantic region. This results in the negative geopotential height anomalies shifting southward from Iceland to the Azores. The shift in anomalies projects onto a negative NAO phase and is therefore responsible for the NAO phase reversal (Fig. 2b). The change in Rossby wave propagation direction from P1 to P2 is also visible in the AMIP simulations (Fig. 2c, d, and supplementary Fig. 5). Therefore, we suggest that the proximate cause of NAO phase reversal in early January during ENSO events is the Rossby wave-propagating direction change over the northeastern North American region. How this change takes place is the next scientific question to be addressed.
Because the propagating direction change is confined mostly to the meridional component of the WAF (i.e., Fy), we then decompose this Fy into four terms (see details in Methods) for these two periods, over the target region. The latitudinal distributions of the Fy and its four constituents for P1 and P2 are displayed in Fig. 3a, b. As we can see, Fy over northeastern North America is positive in P1 but negative in P2, which corresponds well with the respective northward and southward WAF propagation in the two periods. Whereas “term 3” and “term 4” are small, “term 1” and “term 2” both show larger amplitudes and are jointly responsible for the Fy anomalies. According to Eq. (2), which is described in Methods, “term 1” and “term 2” are closely associated with the aeolotropy of the geostrophic stream function anomalies. When the anomaly shows a pattern with a northwest–southeast tilt, these two terms will be positive and therefore prompt the Rossby wave energy to head northward. On the contrary, it will propagate southward if the anomaly has a northeast–southwest tilt. This inference is confirmed by the spatial patterns of the ENSO-related atmospheric anomalies over the northeastern North American region (Fig. 2a, b). We indeed observe that the positive geopotential anomalies exhibit a northwest–southeast tilt in P1 but a northeast–southwest tilt in P2. However, why do the atmospheric circulation anomalies over northeastern North America have spatial structures with different tilting directions in the two periods? This is another scientific question to be resolved.
We assume that this change in the Rossby wave-propagating direction is associated with the atmospheric mean state alteration between the two periods. The climatological distribution of the 250-hPa zonal wind and its spatial change from P1 to P2 are displayed in Fig. 3c. During P1, it shows an evident Atlantic jet over the western Atlantic region, with a central wind speed higher than 40 m/s. To the north of this Atlantic jet, the westerly wind speed decreases with latitude. The westerly is therefore stronger in the south and relatively weaker in the north, over the northeastern North American region. This zonal wind distribution is conducive to forming atmospheric anomalies with a northwest–southeast tilt. From P1 to P2, however, the westerly wind speed accelerates at about 60°N but decelerates at about 40°N. This climatological change weakens the meridional shear of the Atlantic jet on its north edge, which favors tilting of the atmospheric anomaly in a northeast–southwest direction.
To consolidate our hypothesis, we define a zonal wind meridional shear index (U250_shear index) as the climatological zonal wind difference between the regions of 40°–50°N, 270°–300°E, and 50°–65°N, 240°–260°E, which are identified in Fig. 3c. The time evolutions of this U250_shear index and the area-averaged Fy, “term 1” and “term 2” over northeastern North America are displayed in Fig. 3d. We can see that all four of these variables consistently show sharp changes around the NAO transition time. With the climatological decrease in U250_shear, Fy and its components “term 1” and “term 2” also decrease and eventually reverse their signs from positive to negative around January 8th. This coincides well with the NAO phase reversal during ENSO winters. The AMIP simulations also simulate changes in the two WAF components, and the associated climatological zonal wind shifts to a large extent (supplementary Fig. 6). Now, it is quite clear why we can observe negative anomalies in the regions of the Azores and Iceland during El Niño P1 and P2, respectively. However, these anomalies are related to just one of the two lobes of the NAO dipolar pattern. The mechanisms responsible for the establishment of another NAO polarity remain to be uncovered.
It is well known that the NAO is an intrinsic atmospheric mode that is closely involved in local synoptic eddy–low-frequency flow feedback32,33. This strong eddy–low-frequency flow feedback generally follows the “left-hand rule.” In other words, the eddy vorticity fluxes are directed preferentially about 90 degrees toward their left-hand side so that they converge into the cyclonic flow and diverge from the anticyclonic flow34,35. Therefore, if we have negative atmospheric anomalies over the North Atlantic, the eddy vorticity fluxes organized by the cyclonic flow will work to amplify and facilitate the NAO dipolar structure. To confirm this theoretical inference, we display the ENSO-regressed anomalous 250-hPa stream function and eddy vorticity fluxes for P1 and P2 in Fig. 4. The eddy vorticity fluxes indeed follow the “left-hand rule” as they are directed toward the left-hand side of the low-frequency flow. During El Niño P1, the eddy vorticity fluxes converge into the anomalous cyclonic flow at high latitudes. This inevitably leads to a divergence in the south and thus generates a positive stream function anomaly there, which forms a complete positive NAO response. During El Niño P2, the eddy vorticity fluxes converge into the anomalous cyclonic flow in the middle latitudes, which is induced by the Rossby wave trains. This also results in a divergence in the north and thus generates a positive atmospheric anomaly there, which projects onto a complete negative NAO dipolar pattern. The amplifying role played by North Atlantic eddy–low-frequency flow feedback can also be clearly detected in AMIP simulations (supplementary Fig. 7). Therefore, we suggest that this North Atlantic intrinsic positive synoptic eddy feedback serves as an additional mechanism that facilitates a complete NAO dipolar response during the two periods of ENSO winter. In addition, the presence of such strong positive eddy feedback helps the NAO phase reversal more sharply than the basic state changes.