3.1 Asymmetric ENSO amplitude during different AMO phases
Figure 1 shows the 21-year low-pass-filtered AMO index and 21-year sliding standard deviation (STD) of the 10-year high-pass-filtered Niño3, Niño3.4 and Niño4 indices in the previous winter. According to the AMO time series, we select 1900–1904, 1925–1962 and 1998–2015 as the positive AMO phases, and take 1905–1924 and 1963–1997 as the negative AMO phases. The STD of the Niño3.4 index, as a measure of ENSO amplitude, exhibits prominent interdecadal variation, with stronger ENSO amplitude during the negative AMO phase, which is consistent with the published studies (Dong et al. 2006; Kang et al. 2014; Gong et al. 2020; Xu et al. 2023). Similarly, the STD of the Niño3 index (i.e., the SST variability in the eastern equatorial Pacific) shows the interdecadal variation, although with smaller amplitude. Although the STD of the Niño4 index in the central equatorial Pacific maintains an overall negative correlation with the AMO, this relationship is less significant and becomes unstable compared to that for the Niño3.4 and Niño3 indices. It should be noted that even though the 21-year sliding STD can reflect the interdecadal variability of SST amplitude, it may overlook the potential impacts from the ENSO cycle and the different strength between El Niño and La Niña. Therefore, to better understand the modulation effect of the AMO on ENSO amplitude, we will comprehensively consider different phases and zonal positions of the ENSO. The asymmetric ENSO amplitude during different AMO phases and the potential causes will be explored through composite analysis.
Next, we investigate different characteristics of various ENSO events during different AMO phases. Figure 2 shows the composite differences of SSTA in the boreal winter for CP-La and EP-El events. It is clearly seen that the cold center of CP-La is close to the dateline (Figs. 2a and b), while the warm center of EP-El is located in the eastern equatorial Pacific (Figs. 2c and d). These two types of ENSO events are significantly stronger during the negative AMO phase (Figs. 2b and d). To better illustrate the development of the ENSO, we use Figure. 3 to exhibit the evolution of composite zonal SSTA and the minimum/maximum values within the equatorial Pacific during different ENSO events. We can see that both the distribution of SSTA in the equatorial Pacific and the associated extreme values indicate that the intensity of CP-La and EP-El is stronger during the negative AMO phase. This larger SSTA extends beyond the winter of the mature ENSO phase and almost covers the entire developing period. Moreover, there exists a notable longitudinal deviation (~ 50 degrees) between the SSTA centers of CP-La and EP-El. This confirms that the classifications of CP-ENSO and EP-ENSO in our study are reasonable and reliable. Interestingly, the ENSO amplitude differences during different AMO phases are not observed in EP-La and CP-El events (Supplementary Fig. 1). This indicates that the modulation of the AMO on ENSO amplitude is likely to be asymmetric between different phases and zonal positions of the ENSO, which has not been addressed previously.
3.2 Asymmetric responses of ENSO-related air-sea variables during different AMO phases
It has been widely known that both convection and precipitation play crucial roles in tropical regions. Figure 4 shows the time evolution of anomalous precipitation and outgoing longwave radiation (OLR) composites for CP-La and EP-El during different AMO phases. Corresponding to the SSTA mentioned above, stronger negative precipitation anomalies and positive OLR anomalies can be seen in the central-western equatorial Pacific from the developing summer to the decaying spring for CP-La events during the negative AMO phase (Figs. 4a and b), indicating significantly suppressed convection in the region. Similarly, for EP-El events, positive precipitation anomalies and negative OLR anomalies in the central-eastern equatorial Pacific are larger during the negative AMO phase (Figs. 4c and d), suggesting stronger convective activity. Additionally, all the precipitation anomaly centers are located in the west of the SSTA center (Figs. 3 and 4), which reflects a Gill response to the anomalous heat source related to SSTA. This also indicates the warmer and more moist mean-state background in the western and central tropical Pacific (near the warm pool) would favor a more sensitive response of anomalous convection (and also precipitation) to SSTA. During the developing period of the ENSO, persistent local precipitation anomalies in the tropical Pacific often change vertical motion through a positive feedback mechanism (i.e., upward motion - increased precipitation - latent heat release - further enhancement of upward motion), thereby altering the Walker circulation and low-level zonal wind. Through the Bjerknes positive feedback, the ENSO is continuously intensified. Therefore, the intensity of tropical Pacific precipitation during the ENSO developing phase may be coupled with ENSO intensity, and the zonal position of the anomalous precipitation corresponds to the type of an ENSO event.
To verify the hypothesis mentioned above, we use Figs. 5 and 6 to exhibit the composite sea level pressure (SLP) and 850-hPa wind anomalies for CP-La and EP-El events during the developing stage of different AMO phases, respectively. During the negative AMO phase, there exist significant positive SLP anomalies over the central tropical Pacific and negative anomalies over the Maritime Continent and the eastern tropical Indian Ocean in the developing autumn and peaking winter of CP-La events. Thus, this enlarged zonal SLP gradient results in stronger anomalous easterly winds over the central tropical Pacific (Fig. 5). Moreover, these SLP anomalies are stronger during the negative AMO phase with a wider spatial range and a longer duration, as compared with those during the positive AMO phase. Therefore, the easterly wind anomalies are more remarkable during the negative AMO phase, favoring the upwelling of cold water in the central equatorial Pacific and thus strengthening CP-La events. Meanwhile, there is a stronger downward motion in the Niño4 region (Supplementary Figs. 2a and b), indicating an adjustment of the local Walker circulation. Similarly, for EP-El events, stronger high (low) SLP anomalies appear in the central-eastern tropical Pacific (the Maritime Continent and tropical Indian Ocean) during the negative AMO phase. The resultant westerly wind anomalies in the central-eastern tropical Pacific continuously transport warm water from the warm pool to the east, thus strengthening EP-El events. In addition, the tropospheric upward motion become stronger over the Niño3 region, suggesting a weakening of the Walker circulation (Supplementary Figs. 2c and d).
As mentioned previously, the anomalous lower-level zonal winds can change the subsurface ocean layer through the Bjerknes positive feedback. Figures 7 and 8 illustrate the evolution of the composites of subsurface ocean temperature and current anomalies with thermocline depth in the equatorial Pacific for CP-La and EP-El events, respectively, during different AMO phases. Significant westward current anomalies are found in the central and western tropical Pacific during the developing spring to autumn of CP-La under the negative AMO phase. In the early spring, the anomalous cold water at a depth of 100–200 m in the warm pool continuously migrates eastward and upward along the thermocline, which moves to about 100 m deep near 120°W in winter when the CP-La event matures, with the deeper parts of the warm pool replaced by anomalous warm water Figs. 7g-l). In contrast, CP-La-related subsurface ocean temperature and current anomalies are noticeably weaker during the positive AMO phase. During early spring, the anomalous cold water is located in the central tropical Pacific at a depth of around 100 m, which moves eastward and upward slowly above the thermocline (Figs. 7a-f).
During EP-El events of the negative AMO phase, significant and persistent eastward current anomalies appear in the central and western tropical Pacific, which continuously transport warmer water (located in the central tropical Pacific at a depth of 100–200 m in the early spring) eastward and upward along the thermocline. When the EP-El peaks, the anomalous warm water moves to a depth of about 50 m near 100°W, which is shallower and more eastward than the colder water during CP-La events (Fig. 8g-l). In contrast, the subsurface current and temperature anomalies during EP-El events are much weaker during the positive AMO phase, with no significant warmer water in the developing spring. Similar to CP-La events, the anomalous warm water during the developing period stays primarily above the thermocline (Fig. 8a-f).
3.3 The role of AMO in influencing ENSO amplitude asymmetry
To gain a deeper understanding of the dynamic and thermodynamic processes that contribute to ENSO amplitude asymmetry, we conduct a mixed layer heat budget analysis, with a focus on the ENSO developing period. Based on the time evolution of ENSO-related SSTA (Figs. 3c and f), we define the developing period of CP-La as from July to the following January, and the EP-El developing period as from May to December, consistent with the previous research (Gong et al. 2020). Note that these selections do not affect our main conclusions.
Figure 9 shows composite differences of each heat budget term between the negative and positive AMO phases for CP-La and EP-El events, respectively. A negative (positive) T’t indicates the CP-La (EP-El) growing tendency is greater during the negative AMO phase. We can see that meridional advective feedback (term 5), thermocline feedback (term TH) and net heat flux (term Qnet) play significant roles in strengthening CP-La during its developing period in the negative AMO phase (Fig. 9a), while zonal advective feedback (term ZA) acts to suppress the intensification of CP-La. For EP-El events, the strengthening is primarily driven by zonal advective feedback, meridional advective feedback and thermocline feedback, with an opposite effect from the net heat flux Fig. 9b). Chen et al. (2016) demonstrated that the meridional advective feedback mainly serves as an amplifier that relies on the net effect of the other terms. Moreover, as EP-El events are recognized as the traditional ENSO events, diagnostic results for these events are consistent with previous findings (e.g., Gong et al. 2020). Therefore, we will mainly focus on the differences in zonal advective feedback, thermocline feedback and net heat flux for CP-La events during different AMO phases next.
Previous studies illustrated that the tropical Pacific thermocline would flatten under the negative AMO phase (Dong et al. 2006; Levine et al. 2017; Xu et al. 2023), that is, the thermocline shoals (deepens) in the central (eastern) equatorial Pacific, corresponding to upward (downward) background vertical current deviation. Compared to the positive AMO phase, the low-frequency background eastward current in the central equatorial Pacific accelerates significantly during the negative AMO phase (Fig. 10a). Such a current speed divergence in this region can trigger upward motion that prompts the upwelling of colder water from deeper layers, thereby facilitating the emergence and growth of CP-La events. Conversely, in the eastern equatorial Pacific, the eastward current becomes weaker and the associated convergence induces downward motion to warm the subsurface sea water, favoring the development and intensification of EP-El events. Given this, we also show the differences of Ekman pumping velocity (EPV), calculated by using wind stress, in the tropical Pacific between the two AMO phases (Fig. 10b). In contrast to those in the positive AMO phase, there are notable positive EPV anomalies in almost the entire western and central equatorial Pacific and negative EPV anomalies around 145°-85°W. This would favor the development of CP-La and EP-El events.
Figure 11a presents the low-frequency SST and 850-hPa wind regressed onto the AMO index. Note that regression results are reversed in sign to effectively represent the background conditions during the negative AMO phase. During the negative AMO phase, relatively colder SSTs occur in the western tropical Pacific (i.e., the warm pool), while the SSTs in the east are narrowly and slightly warmer. Concurrently, anomalous westerly (easterly) winds prevail over the central-western (eastern) tropical Pacific, indicating a weaking (strengthening) of the trade wind. These results are consistent with the previous findings (Dong et al. 2006; Gong et al. 2020). Moreover, these low-frequency anomalies of SST and low-level wind suggest that during the negative AMO phase, the zonal mean-state SST gradient is weaker than that during the positive AMO phase, thus increasing the air-sea coupling instability and thus favors the development of the ENSO (Dong et al. 2006; Gong et al. 2020). In addition, this weaker zonal SST gradient would enhance the zonal advection feedback during the developing period of EP-El, but suppress it during the development phase of CP-La, consistent with the diagnostic results of the ocean heat budget.
Finally, we investigate the difference of Qnet term in Eq. (1) for CP-La events (subplots in Fig. 9). Generally, for La Niña events, the central tropical Pacific SST gradually decreases during the developing period. We can see that the Qnet term is positive near the cold SSTA center, indicating that the ocean receives heat from the atmosphere to suppress the development of CP-La events. Interestingly, the Qnet difference between the two AMO phases is negative (Fig. 9a), favoring the development of CP-La during the negative AMO phase. Furthermore, this Qnet difference mainly comes from the LH difference (subplot in Fig. 9a) with a strong upward LH anomaly in the Niño4 region during the negative AMO phase, but shows little LH anomaly during the positive AMO phase (Supplementary Fig. 3).
As LH is closely related to air-sea humidity difference, we show the regression map of 2-m specific humidity onto the AMO index (Fig. 11b). Clearly, the entire tropical Pacific region becomes significantly drier during the negative AMO phase, except for weak moist anomalies in the central tropical Pacific. When combined with the low-level winds shown in Fig. 11a, it can be observed that both the westerly wind anomaly in the central and western tropical Pacific and the easterly wind anomaly in the east jointly produce a pronounced dry advection toward the central tropical Pacific. As the ocean stays in a saturated state, this strong dry advection in the atmosphere would enlarge the air-sea humidity difference in central tropical Pacific. Consequently, during the developing period of CP-La, the ocean releases more LH to the atmosphere in the central tropical Pacific to further decrease the SST there, which promotes the strengthening of CP-La during the negative AMO phase.