Previous studies have shown that the Pacific meridional mode (PMM)43 is an effective conduit through which the NPO can eventually impact the tropics and ENSO variability44−46. It has been proposed that the NPO can reduce surface evaporation and increase SST in the subtropical northeastern Pacific by reducing the speed of the subtropical northeasterly trade winds. The warming of the subtropical northeastern Pacific would further reduce the trade winds and initiate a positive thermodynamic feedback among surface winds, evaporation, and SST, known as the wind–evaporation–SST (WES) feedback47. This WES feedback excites the PMM, which propagates positive SST anomalies from the subtropics into the central equatorial Pacific, where positive SST anomalies are conducive to the development of El Niño. In addition to the PMM, there are at least two other mechanisms by which NPO atmospheric anomalies can lead to the onset of El Niño through the weakening of off-equatorial trade winds48 or the excitation of the off-equatorial Rossby wave49. However, studies of these ENSO extratropical precursor dynamics have focused mainly on the effects of the NPO on the onset of El Niño rather than on the potential role of the NPO in prolonging the duration of El Niño. Thus, there is a need to develop an understanding of the dynamics underlying the link between the NPO and multi-year El Niño events.
Figure 2 shows the evolutions of SST and SLP anomalies composited for five multi-year El Niño events over a 3 yr period. The positive NPO forcing during JFM(0) (Fig. 2b) generates positive SST anomalies extending from the subtropical northeastern Pacific to the central equatorial Pacific, closely resembling a positive phase of the PMM43 (Fig. 2a), which maintains positive SST anomalies for several seasons and extends them equatorward into the central equatorial Pacific, finally leading to equatorial Pacific warming during the subsequent JFM(1) (Fig. 2c; see Supplementary Fig. 6 for more details). This process involves teleconnections from the extratropics to the tropics (termed “extratropical−tropical teleconnections”), consistent with previous findings28−31.
We then examine whether the equatorial Pacific warming can exert an influence on the extratropics. We note that the NPO-induced Pacific SST anomaly pattern during JFM(1) closely resembles that of the CP El Niño39−42, characterized by the center of the warming being located mainly in the central equatorial Pacific (Fig. 2c; see also Supplementary Fig. 7a). It has been recognized that the extratropical SLP response to ENSO is sensitive to the longitudinal position of maximum SST anomalies along the equatorial Pacific50. Unlike the projection of the Eastern Pacific (EP) El Niño onto the Aleutian Low, the CP El Niño projects onto a very different SLP pattern in the North Pacific, and is closely associated with negative SLP anomalies over the Hawaiian region51, 52 (Supplementary Fig. 8). The Hawaiian SLP variability, represented by the time series of SLP anomalies averaged over the Hawaiian region (SLPHI; 158°–135°W, 13°–24°N; red box in Fig. 2d), is closely linked to the NPO-like SLP variability and PMM (Supplementary Fig. 9). The composite SLPHI and NPO indices for multi-year El Niño events show two peaks: one during JFM(0) and the other during JFM(1) (Fig. 3a, b). The re-intensification of the SLPHI and NPO indices during JFM(1) is consistent with the warming in the central tropical Pacific associated with CP El Niño, implying that the NPO-induced central tropical Pacific SST variability can feed back into the North Pacific, generating negative SLP anomalies over the Hawaiian region (by extension of the positive phase of the NPO in the North Pacific). This process is referred to as tropical−extratropical teleconnections.
The anomalous low pressure over the Hawaiian region in response to central equatorial Pacific warming during JFM(1) produces an anomalous southwesterly flow on its eastern flank, which acts to re-weaken the off-equatorial trade winds, which in turn activate ENSO precursor dynamics such as the PMM43, resulting in a re-intensification of the PMM during March–April–May (MAM)(1) (Fig. 3c). This positive feedback between the PMM and CP El Niño on interannual timescales is consistent with the recent findings of Stuecker53. Through PMM dynamics, positive SST anomalies in the subtropical northeastern Pacific persist through June–July–August (JJA)(1) and propagate gradually from the subtropics into the central equatorial Pacific44−46 (Supplementary Fig. 6), thereby re-initiating the development of El Niño of year (1) (Fig. 2e), finally resulting in multi-year El Niño events. The results are robust among different SST and SLP data sets (Supplementary Fig. 10).
Through a careful examination of each of the individual multi-year El Niño events (Supplementary Fig. 11), we note that the individual evolution of each event is similar to the composite results presented above. That is, the positive NPO event (defined by a 1.0 SD threshold of the JFM(0) NPO index) is followed by a typical CP El Niño pattern39−42 or a mixed pattern42 of CP and EP El Niño in JFM(1), characterized by maximum warming near the central equatorial Pacific. Furthermore, there are negative SLP anomalies between Hawaii and western North America that are consistent with a typical pattern of the SLP response to CP El Niño SST forcing. The negative SLP anomalies project again onto a weakening of the off-equatorial trade winds and then a strengthening of the PMM, resulting in multi-year evolution of El Niño.
The schematic of Fig. 4 presents a two-way feedback mechanism between the tropics and extratropics associated with the CP El Niño phenomenon39−42 to explain the dynamics underlying the linkage between the NPO and multi-year El Niño events. The boreal winter NPO induces a CP-type El Niño event over the equatorial Pacific during the subsequent winter through its effect on the PMM43. This CP-type El Niño in turn feeds back into the North Pacific to force changes in atmospheric circulation over the Hawaiian region, which re-activate the PMM to favor the development of another El Niño event. This process continues until it is disrupted by negative feedbacks, as suggested by the ENSO cycle54 or noise in the air–sea coupled system9. The two-way feedback mechanism between the tropics and extratropics proposed here to explain the sources of multi-year El Niño events echoes that previously identified by Di Lorenzo et al.33 to explain the sources of tropical Pacific decadal (timescale > 8 years) variability. Di Lorenzo et al.33 suggested that the NPO atmospheric variability, which acts as the extratropical stochastic forcing, may enhance the low-frequency variance of ENSO through a chain of extratropical−tropical feedback processes. The present study demonstrates that a similar dynamical chain may also validly explain the dynamics underlying the emergence of multi-year El Niño events, which are dominated primarily by interannual variations in the 3 to 7 yr ranges. Taken together, it would appear that NPO atmospheric forcing, through extratropical−tropical feedbacks, may contribute to enhance the low-frequency variability in ENSO over a wide spectrum of time scales (interannual-to-decadal timescales; >3 years).
The schematic of Fig. 4 highlights the potential importance of CP El Niño in linking the NPO to multi-year El Niño events. Previous studies have suggested that the NPO tends to induce the CP-type El Niño, but not in all cases31. Equatorial ocean dynamics, such as zonal advection in the tropical Pacific, can extend NPO-induced warming in the central equatorial Pacific eastwards, leading to EP El Niño events55. Several single-year El Niño events (1965/66, 1972/73, 1991/92, and 1997/98), which are also preceded by a positive NPO event during JFM(0) (Supplementary Fig. 12d), are characterized by a typical EP El Niño pattern during JFM(1) (Supplementary Fig. 12b). Consistent with the position of maximum warming, there is only a weak positive SLP response over the Hawaiian region (Supplementary Fig. 12e, h). As a result, the PMM is very weak (PMM index < ±0.5 SD) during MAM(1) (Supplementary Fig. 12i), and the warming in the equatorial central−eastern Pacific decays rapidly to near La Niña conditions by JFM(2) (Supplementary Fig. 12c), thereby leading to a single-year El Niño event.
Although CP El Niño is important in bridging the NPO to multi-year El Niño events, this does not mean that all CP El Niño events can evolve as a multi-year event. For the NPO-preceded CP El Niño events, positive SST anomalies extend more westward (west of 170°E) into the western Pacific warm pool where the deep convection can more easily be excited37 (Supplementary Fig. 13a), thereby favoring a stronger SLP response over the Hawaiian region (Fig. 3b). In contrast, for the non-NPO-preceded CP El Niño events (1977/78, 1994/95, 2002/03, and 2009/10), positive SST anomalies are confined mainly east of 170°E and do not extend sufficiently into the western Pacific warm pool (Supplementary Fig. 13b). Therefore, the non-NPO-preceded CP El Niño events are all accompanied by a weaker SLP response over the Hawaiian region during JFM(1) (Supplementary Fig. 14h) and a weaker PMM-related SST anomaly band during MAM(1) (Supplementary Fig. 14i). As a result, El Niño conditions are not maintained for an additional year (Supplementary Fig. 14c), leading to single-year evolution. The results presented above suggest that both the NPO and its induced CP El Niño are necessary for generating multi-year El Niño events, consistent with the dynamical hypothesis that we have proposed (Fig. 4).