Introduction

As the most prominent interannual climate variability on the planet, the El Niño-Southern Oscillation (ENSO) affects the weather and climate over most of the globe through a chain of oceanic or atmospheric teleconnections1,2,3. For the East Asian region, the seasonal-to-interannual climate response to ENSO is robust and has been extensively studied. It has been demonstrated that ENSO exerts its influence in East Asia mainly by modifying the atmospheric circulation over the western North Pacific (WNP) with an anomalous low-level Philippine Sea anticyclone/cyclone4,5,6,7,8 and a Kuroshio anticyclone/cyclone9,10,11. During El Niño winters, the southerly wind anomalies embedded in the western part of these anticyclonic circulation anomalies bring abundant warm and moist air to East Asia, leading to abnormal wet and warm climate states and a weakened East Asian winter monsoon (EAWM)5,7,8,11,12,13. When La Niña events occur, a roughly opposite picture is generally observed, with non-negligible asymmetric characteristics14,15.

However, this ENSO influence in winter has been shown to be unstable on subseasonal timescales16,17. Compared to the late winter (January–February) EAWM, the early winter (November–December) EAWM has been shown to be more responsive to ENSO and thus more predictable in operational climate models9,18,19. This subseasonal change is suggested to be related to the sudden disappearance of the anomalous Kuroshio anticyclone/cyclone in January, which significantly weakens the ENSO influences thereafter9. As part of the North Pacific atmospheric circulation, the Kuroshio anticyclone/cyclone is influenced by both the tropical Central Pacific (CP) and the WNP precipitation anomalies. These two forcings have opposite effects during El Niño winters. While the former excites strong cyclonic circulation anomalies, the suppressed convection in the WNP is conducive to an anomalous anticyclone there and thus favors a Kuroshio anticyclone10,20,21. From early to late winter, the precipitation response to ENSO decreases in the WNP but increases in the tropical CP. These changes in the magnitude of convection anomalies result in the strengthening effect from the tropical CP precipitation and thus the subseasonal disappearance of the anomalous Kuroshio anticyclone10,20,21. Other studies argue that the disparate ENSO effects between the early and late winter are associated with the different impacts of ENSO on the tropical Walker circulation during the two periods18,19,22. During early winter, ENSO induces a double-cell anomalous Walker circulation over the tropics. This leads to pronounced precipitation responses in the tropical eastern Indian Ocean–western Pacific, which excites a systematic extratropical Rossby wave train from the tropical Indian Ocean towards East Asia, favoring an enhanced influence on the EAWM18,22. In contrast, during late winter, ENSO is accompanied by a single-cell anomalous Walker circulation over the tropics. The precipitation response in the tropical eastern Indian Ocean–western Pacific is thus weakened, which can hardly produce an obvious wave train to exert much influence on the EAWM19,22. In addition, from early winter to late winter, ENSO teleconnection to the Euro-Atlantic region undergoes a clear subseasonal change. The atmospheric response to a warm ENSO event shifts from a typical positive North Atlantic Oscillation (NAO) to a negative NAO23,24,25. Since the NAO is considered to be an important driver of the Eurasian winter climate variability26,27,28,29,30, the EAWM response to ENSO may also be influenced by this pathway16.

Although the above studies have documented the late winter weakening of the ENSO footprint in East Asia, the evolutionary feature of the EAWM responses to ENSO during the entire boreal winter season is still unclear. When, how, and why the ENSO influence begins to weaken requires further systematic investigation. In this paper, based on the latest high temporal-resolution observations and reanalysis datasets, we examine the daily evolution of the ENSO influence on the EAWM and find that the teleconnection is generally stable throughout the whole boreal winter season, but exhibits a peculiar break for about 10 days in the midwinter. The possible mechanisms responsible for this teleconnection collapse are then proposed.

Results

Midwinter breakdown of the ENSO influence on the EAWM

In Fig. 1a, we first display the ENSO-regressed spatial pattern of the boreal winter seasonal (NDJFM) mean 850-hPa wind anomalies to show the influence of ENSO on the East Asian atmospheric circulation. As demonstrated in many previous studies4,5,6,7,8,9,10,11, when El Niño events occur, there are significant anticyclonic circulation anomalies over the WNP, which leads to pronounced southerly wind anomalies over East Asia and thus a weakened EAWM. To explore the subseasonal variations of this influence, the monthly evolution of the EAWM response to ENSO is examined in Fig. 1b. It can be seen that, unlike the responses in November, December, February, and March, which show a significantly weakened EAWM, there is an evident breakdown of the ENSO influence on the EAWM in January. We further examine the daily evolution of the EAWM response to ENSO in Fig. 1c. The weakened EAWM can be detected stably throughout almost all wintertime days. An exception occurs during midwinter, January 14–23, when the EAWM is exclusively strengthened, suggesting a breakdown and reversal of the ENSO–EAWM teleconnection. Correspondingly, the East Asian precipitation and SAT responses exhibit similar characteristics with the EAWM, i.e., El Niño events favor an increase in precipitation and SAT in East Asia, but this relationship cannot be established or even reversed during the midwinter. In addition, composite analysis is conducted to examine the daily evolution of the East Asian winter climate anomalies during El Niño and La Niña winters in Supplementary Fig. 1. It is also characterized by a distinct midwinter collapse both in El Niño and La Niña winters, consistent with the results shown by regressions. We note that for East Asia, midwinter is typically a period with the strongest annual EAWM and coldest annual SAT (Fig. 1c). This distinct breakdown of ENSO influence may mean a less predictable January or midwinter climate in East Asia, which has already been manifested itself in climate prediction models31.

Fig. 1: Subseasonal variation of the ENSO influence on the EAWM.
figure 1

Regression coefficients of the winter a NDJFM mean 850-hPa wind anomalies (vectors in ms−1), b monthly mean EAWM index (bars in ms−1), and c daily mean EAWM index (bars in ms−1), precipitation index (blue curve in mmday−1), and surface air temperature (SAT) index (red curve in K) with respect to the December–February (DJF) mean Niño-3.4 index. The black box in (a) outlines the region used to define the EAWM index. The wind anomaly in (a) is shown only when its zonal or meridional component is significant at the 90% confidence level. Scale bars in (b) represent a range of 90% confidence level. The precipitation and SAT indices in (c) are defined as the area-averaged anomalies in the regions marked with the boxes in Fig. 2a and Supplementary Fig. 2a, respectively. Solid bars in (c) denote the corresponding regressions are significant at the 90% confidence level. The cyan dashed curve in (c) shows the daily evolution of the climatology for the SAT index. The vertical black dashed line in (c) roughly denotes the timing of the EAWM phase transition. The translucent red, blue and red polygons in (c), respectively, mark the periods defined as P1 (November 1 to January 10), P2 (January 14–23), and P3 (January 27 to March 31).

Full size image

To further illustrate this particular subseasonal change in the ENSO influence, we divide the entire wintertime into three periods, namely P1 (November 1 to January 10) and P3 (January 27 to March 31) with weakened EAWM responses, and P2 (January 14–23) with a strengthened EAWM anomaly (Fig. 1c). The spatial patterns of the ENSO-regressed atmospheric circulation (Fig. 2a–c and Supplementary Fig. 2a–c), precipitation (Fig. 2), and SAT (Supplementary Fig. 2) anomalies during these three periods are then presented. It can be seen that during P1, El Niño events are accompanied by large-scale anticyclonic circulation anomalies over the WNP, with one center over the Philippines (i.e., Philippines Sea anticyclone) and another over the Kuroshio extension region (i.e., Kuroshio anticyclone). The anomalous southerly wind in the western part of these anomalous anticyclones strongly dominates over most of East Asia, resulting in a weakened EAWM, increased precipitation, and warm SAT responses (Fig. 2a, d). During P3, while the Kuroshio anticyclone disappears, the Philippine Sea anticyclone is still evident and has a large spatial scale. Thus, we can also observe distinct southwesterly wind anomalies over East Asia and similar climate anomalies to those in P1 (Fig. 2c, f). However, the circulation pattern is different for P2. Northern East Asia is controlled by an anomalous cyclone rather than an anticyclone, which appears to be associated with the changes in the mid-to-high latitudes. The Philippine Sea anticyclone still exists but is mostly confined to the tropics. As a result, we can detect evident northwesterly wind anomalies over the East Asian region and a strengthened EAWM (Fig. 2b). Correspondingly, the precipitation anomaly in eastern China and South Korea is reduced (Fig. 2b, e), and there also exist colder SAT responses (Supplementary Fig. 2b, e). Based on these results, we suggest that during midwinter, the ENSO influence in East Asia undergoes a remarkable breakdown, which is manifested in both atmospheric circulations and surface climate responses.

Fig. 2: Midwinter breakdown of the ENSO climate influence in East Asia.
figure 2

Regression coefficients of the 850-hPa wind (vectors in ms−1) and precipitation anomalies (shadings in mmday−1) with respect to the DJF mean Niño-3.4 index for the a P1, b P2, and c P3. The wind anomaly is shown only when its zonal or meridional component is significant at the 90% confidence level. The dots denote the precipitation anomaly is significant at the 90% confidence level. Regression coefficients of the stational precipitation anomalies (shadings in mmday−1) with respect to the DJF mean Niño-3.4 index for the d P1, e P2, and f P3. The statistically significant and insignificant precipitation anomalies at the 90% confidence level are respectively marked with solid and hollow circles. The blue boxes denote the region used to define the precipitation index.

Full size image

Possible mechanisms responsible for the midwinter breakdown

We now turn to explore the possible mechanisms responsible for this midwinter breakdown of the EAWM response to ENSO forcing. It is unlikely that this abrupt subseasonal EAWM change can be directly interpreted by ENSO-related SST anomalies as they are highly persistent during the boreal winter season32. Although the ENSO-related SSTs show some subseasonal variations from November to March, for example, the wintertime SST anomaly gradually decreases in the tropical CP but increases in the tropical Indian Ocean (Supplementary Fig. 3a, c), they can hardly account for the abrupt collapse of the ENSO influence. The tropical precipitation anomalies associated with ENSO are then analyzed. They are characterized by positive precipitation anomalies in the tropical CP and western Indian Ocean (WIO), and negative precipitation anomalies in the western Pacific (WP) (Supplementary Fig. 3b). Previous studies have suggested that the subseasonal variation in the response of the East Asian atmospheric circulation to ENSO is associated with the convective forcings in the tropical CP and WP9,10, and the precipitation anomaly in the Indian Ocean can also play a role18,19,25,33. To test these hypotheses, we define three precipitation indices to measure the precipitation variability in the tropical CP (Pr_CP), WP (Pr_WP), and WIO (Pr_WIO), respectively (see “Methods”). Their time evolution during the boreal winter season is displayed in Supplementary Fig. 3d. Compared to those in P1 and P3, the precipitation indices show little distinct or abrupt change in intensity during P2, suggesting that the subseasonal changes in the magnitude of the ENSO-induced tropical precipitation anomalies cannot directly explain the ENSO–EAWM teleconnection breakdown in P2. We next display the partial-regressed atmospheric circulation patterns with respect to the three tropical precipitation indices to examine the contributions of the changes in their effects (Fig. 3). It is found that while the CP convection exerts a cyclonic circulation anomaly over the North Pacific (Fig. 3a–c), the negative WP precipitation anomaly is associated with an anomalous anticyclone there (Fig. 3d–f), which is consistent with previous studies9,10. In comparison, the effects of the WIO precipitation are relatively weak (Fig. 3g–i). Regarding to the influence change among the three periods, while the patterns related to the WP precipitation anomaly remain almost unchanged, the CP precipitation forcing exerts some different effects over East Asia. In particular, the northerly wind anomalies over East Asia during P2, which are crucial for the EAWM response collapse, are mostly contributed by the ENSO-related local CP convective forcing. Therefore, it seems that the ENSO local forcing is more important to produce this midwinter breakdown. But how this ENSO local forcing leads to the anomalous northerly wind over East Asia is the next scientific question to be resolved.

Fig. 3: Subseasonal variations of the atmospheric response to ENSO-induced tropical convection.
figure 3

Partially-regressed coefficients of the anomalous 500-hPa geopotential height (shadings and contours in m) and 850-hPa horizontal winds (vectors in ms−1) with respect to the Pr_CP index during the a P1, b P2, and c P3. d–f Same (a–c), but for the partial regressions with respect to the Pr_WP index. g–i Same (a–c), but for the partial regressions with respect to the Pr_WIO index. Dots are displayed when the geopotential height anomalies are significant at the 90% confidence level. The wind anomaly is shown only when its zonal or meridional component is significant at the 90% confidence level.

Full size image

We then directly examine the regressed pattern of the northern hemispheric atmospheric circulation with respect to the Niño-3.4 index during P1, P2, and P3 to explore the possible clues (Fig. 4, note that the regression with respect to the Pr_CP index (not shown) shows almost the same pattern). It can be seen that in P1 of the El Niño winter (Fig. 4a), there are pronounced large-scale low-level anticyclonic circulation anomalies over the Philippine Sea and the Kuroshio region. As a result, strong southerly wind anomalies prevail over East Asia, weakening the EAWM and producing enhanced precipitation. From a large-scale perspective, we can observe prominent Rossby wave trains extending from the Indo-western Pacific and the Central Pacific to the North Atlantic, which projects onto a significant positive NAO response. However, the pattern in P2 shows a dramatically different picture (Fig. 4b). First, the Kuroshio anticyclonic circulation is replaced by an anomalous cyclone, and the Philippine Sea anticyclone is also observed but shrinks to the south of East Asia. Second, the Rossby wave train from the Indo-western Pacific disappears. Only the wave train from the central Pacific reaches the North Atlantic, which is associated with the negative phase of the NAO pattern.

Fig. 4: Subseasonal variations of the ENSO-related atmospheric circulation anomalies in mid-to-high latitudes.
figure 4

Regression coefficients of the 850-hPa wind (vectors in ms−1) and 500-hPa geopotential height anomalies (shadings in m) with respect to the DJF mean Niño-3.4 index for the a P1, b P2, and c P3. The wind anomaly is shown only when its zonal or meridional component is significant at the 90% confidence level. d Regression coefficients of the North Atlantic zonal mean (70°–0°W) daily SLP anomalies (shadings in hPa) and NAO index (green curve) from December to January with respect to DJF mean Niño-3.4 index. The dots denote the corresponding anomalies are significant at the 90% confidence level. The vertical black dashed line in (d) roughly denotes the timing of the NAO phase transition.

Full size image

Since the wave source associated with the convection anomalies in the tropical Indo-western Pacific is still there, this disappearance of the wave train may be a result of interference from other forcing. Indeed, we observe a distinct atmospheric wave train in the mid-latitudes of Eurasia with a strengthened Siberian high and East Asian trough (i.e., weakened Kuroshio anticyclone), which brings strong northerly wind anomalies to East Asia and strengthens the EAWM. In P3 (Fig. 4c), although no significant change over the Kuroshio region, the Philippine Sea anticyclone is slightly restored to its original position to exert a significant influence on the EAWM. Meanwhile, the atmospheric wave activity in the mid-latitudes is largely weakened, which may be associated with the receded NAO response.

To sum up, the main reason for the collapse of the EAWM response in P2 appear to be the strengthened Siberian high which may be embedded mid-latitude wave train adjustment associated with the NAO phase transition. We further show the daily evolution of the ENSO-regressed North Atlantic SLP anomalies in Fig. 4d. There is a clear abrupt NAO phase reversal from positive to negative in early January, which occurs about a week before the breakdown of the EAWM response. This lead-lag relationship is also evident during the two super El Niño winters, namely the winters of 1997–1998 and 2015–2016 (Supplementary Fig. 4). As demonstrated in our previous study34, this abrupt NAO phase reversal is primarily triggered by a climatological weakening of the Atlantic jet meridional shear in early January, which shifts the propagation direction of the ENSO-induced low-frequency Rossby waves before they enter the North Atlantic region. The changes over the North Atlantic are further amplified by the local intrinsic positive feedback between the eddy and the low-frequency flow. It is well known that the NAO has important downstream influences on the East Asian climate, mainly via quasi-stationary wave propagation of upper-tropospheric anomalies along the Asian jet, with a negative NAO generally associated with a strengthened Siberian high and EAWM26,27,28,29,30. Therefore, it appears that the Siberian high and East Asian trough strengthening, which are key reasons for the collapse of the EAWM response, are actually the results of the NAO phase transition. Considering that it usually takes about a week for the NAO signal to maximumly affect the East Asian atmospheric circulation on the subseasonal timescale35, the breakdown of the ENSO–EAWM teleconnection in the mid-January seems to be a consequence of the abrupt ENSO-induced NAO phase reversal that occurs about one week earlier.

To further probe the hypothesis, we focus on the spatiotemporal characteristics of the atmospheric circulation and wave activity related to the NAO. Figure 5a–c displays the horizontal wave activity flux (WAF) corresponding to the 250-hPa geopotential anomalies associated with the sign-reversed NAO index at different lag days, together with the anomalous 250-hPa stream function and 850-hPa meridional wind. When the NAO is in its negative phase (lag = 0), there is a strong wave flux emanating from the NAO anomaly center over the North Atlantic and directed southeastward across Europe and the Middle East to reach India (Fig. 5a). In addition to the southeastward branch, a high-latitude pathway also emanating from the North Atlantic region emerges at lag = 3 days: this wave flux propagates eastward across Siberia and turns progressively southeastward to affect the East Asian atmospheric circulation (Fig. 5b). Six days after the -NAO peak, we can still detect this high-latitude Rossby wave train (Fig. 5c). Correspondingly, along with the quasi-stationary wave propagation, we also see a strengthened Siberian high and East Asian trough at lag = 3 days and lag = 6 days. These suddenly appearing atmospheric circulation systems shift the anomalous WNP anticyclone to the south (Fig. 2b), and East Asia is taken over by the strong northerly wind anomalies, which strengthens the EAWM and thus leads to the breakdown or even reversal of the conventional ENSO–EAWM teleconnection mediated by the anomalous WNP anticyclones. In P3, the downstream effects of the NAO weaken as the negative NAO response to ENSO recedes (Fig. 4d). As a result, the Siberian high and the East Asian trough return to their normal states, and the anomalous Philippine Sea anticyclone regains its position to exert a weakening effect on the EAWM. These results are consistent with a previous study that focused on the subseasonal influence of the NAO on the East Asian winter climate35.

Fig. 5: ENSO-driven abrupt NAO phase reversal leads to the midwinter breakdown of the ENSO–EAWM teleconnection.
figure 5

Forty-two-year-averaged regression coefficients of the anomalous daily 850-hPa meridional wind (shadings in ms−1), 250-hPa geostrophic stream function (contours in 106 m2 s−2), and the associated wave activity flux (WAF, vectors in m2 s−2) with respect to the sign-reversed daily NAO index from December to January at a 0-day, b 3-day, and c 6-day time lags. The dots denote the 850-hPa meridional wind anomalies that are significant at the 90% confidence level. The WAFs are masked out where the climatological DJF mean zonal wind speed is less than 5 ms−1. d Scatterplot of the ΔEAWM index as the function of the ΔNAO index for El Niño (red circle), La Niña (blue quadrate) and normal (gray rhombus) boreal winter. The ΔEAWM index is defined as the difference of the EAWM index between P2 and P1, and the ΔNAO index is defined as the difference of the NAO index between the periods of January 9–19 and November 1 to January 8. The correlation coefficient (R) between these two variables is also displayed.

Full size image

We further present the relationship between the changes in the NAO and EAWM indices before and after the reversal in Fig. 5d. They show a significant in-phase relationship, which means that when El Niño forces the NAO index to decrease from P1 to P2, the EAWM index will be also decreased subsequently. In other words, the winter monsoonal circulation will be intensified, further consolidating our aforementioned hypotheses that the ENSO-driven abrupt NAO phase reversal in early January is responsible for the midwinter breakdown of the ENSO influence on the EAWM. However, it should be noted that, in Fig. 5d, not all ENSO events are accompanied by significant NAO phase transition in early January, suggesting that this ENSO-driven NAO phase reversal bears a considerable extent of uncertainty. Although beyond the scope of the current study, we assume that this uncertainty is likely related to the spatiotemporal complexity of ENSO events36 and the nonlinear and nonstationary feature of the North Atlantic atmospheric response to ENSO37,38,39. Further details on this are still the subject of future investigation.

Discussion

The ENSO footprints in the East Asian winter climate have been widely investigated. While the bulk of previous studies has focused on the influence at seasonal-to-interannual timescale, more and more recent studies suggest that ENSO effects in East Asia may change at shorter timescales. For example, the early winter climate in East Asia has been shown to be more responsive to ENSO and more predictable in climate models, compared to the counterpart in late winter9,10,18,19,22. This motivates us to systematically explore the subseasonal evolution of the ENSO influence on the East Asian winter climate.

The daily observations and reanalysis datasets during the past several decades are utilized in this paper, which allows us to detailly identify the possible subseasonal changes of the ENSO influence. It is found that the ENSO teleconnection to the EAWM is generally stable throughout the whole boreal wintertime days, that is, El Niño events are typically accompanied by a weakened EAWM with enhanced precipitation and SAT in East Asia. However, it is interesting to find that there is a remarkable and distinct breakdown of the ENSO influence on the EAWM during the midwinter. This collapse lasts about 10 days and occurs in mid-January, when East Asia’s climate is at its coldest point of the year, posing a potential threat to local socio-economic activities40.

The possible mechanisms responsible for this midwinter breakdown are then analyzed. It is suggested that the break is largely caused by interference from the ENSO-driven NAO abrupt phase reversal in early January. This ENSO-induced abrupt NAO phase transition has been investigated in our previous work34, which suggests that it is primarily caused by the climatological change in the Atlantic jet meridional shear, which shifts the propagation direction of the ENSO-induced low-frequency Rossby waves, and is further amplified by the intrinsic positive feedback between the eddy and the low-frequency flow over the North Atlantic region. It is well known that the NAO has important downstream influences on the East Asian climate, with a negative NAO generally associated with a strengthened Siberian high and EAWM26,27,28,29,30. On the subseasonal timescale, the NAO exerts the strongest influence on the East Asian atmospheric circulation about a week after its peak phase35. In our case, the abrupt NAO phase transition from positive to negative during early January of the El Niño winter leads to a rapid adjustment of the mid-latitude atmospheric circulation via wave eastward propagation along the mid-latitude jet. This leads to a strengthened Siberian high, producing strong northerly wind anomalies over East Asia, and the anomalous Philippine Sea anticyclone is shifted southward and confined to the tropics, which eventually gives rise to the collapse of the teleconnection. During the La Niña winters, a generally opposite picture is detected.

Our findings provide key insights into the non-stationarity of the ENSO effects on the EAWM, and carry important implications for the subseasonal predictability of the ENSO influence on the East Asian winter climate. However, we note that this ENSO-NAO-EAWM linkage bears a considerable extent of uncertainty, possibly be related to the spatiotemporal complexity of the ENSO characteristics36, as well as the nonlinear and nonstationary NAO response to ENSO37,38,39. The underlying reasons for this uncertainty are interesting topics worthy of future study. In addition, as our current conclusions are based on observations and reanalysis datasets with a limited time record, it would be worthwhile to analyze the model representation of this midwinter breakdown of the ENSO influence based on model simulations, such as the simulations in the Coupled Model Intercomparison Project phase 6 (CMIP6)41.

Methods

Reanalysis and observational products

The daily mean observational datasets used in this study are derived from the 699 weather stations in China (China Meteorological Administration, CMA) and 60 weather stations in South Korea (Korea Meteorological Administration, KMA). Two variables, precipitation and surface air temperature (SAT), are analyzed. For better reliability, we adopt the data during the recent period from Jan 1979 to Dec 2021. We also utilize the National Oceanic and Atmospheric Administration (NOAA) Extended Reconstructed global monthly SST V5 dataset42 for the same period with a horizontal resolution of 2° longitude × 2° latitude. In addition, the daily atmospheric circulations, along with the daily 2-m temperature (i.e., SAT) and precipitation, are also used from the fifth generation of the European Centre for Medium-Range Weather Forecasts (ECMWF) reanalysis (ERA5)43 for the same period with the horizontal resolution of 1° longitude × 1° latitude.

All datasets are analyzed for the boreal winter season (November–March, NDJFM) in a broad sense, and the winter of 1979 refers to the winter of 1979–1980. Anomalies were derived relative to the daily mean climatology over the entire period of record (1979–2020). Linear trends have been removed to avoid possible influences associated with global warming or long-term trends.

Climatic index

We employ several climatic indices to facilitate our analyses. The intensity of the EAWM is measured by the area-averaged boreal winter 850-hPa meridional wind anomalies (e.g., a positive EAWM index is associated with anomalous southerly winds and thus a weakened EAWM) within the domain of 20°–40°N and 100°–140°E44. According to the previous study45, the performance of the low-level wind indices is relatively better than the others in capturing the interannual ENSO–EAWM relationship. Other EAWM indices, such as the zonal SLP difference index46 and the 500-hPa East Asian trough index47, are also used and the results remain qualitatively unchanged (Supplementary Fig. 5). The Niño-3.4 index, defined as the area-averaged SST anomalies in the Niño-3.4 region (5°S–5°N, 120°–170°W), is used to describe the amplitude of ENSO. Following conventions, we utilize a threshold of ±0.5 standard deviations of the DJF mean Niño-3.4 index to define ENSO events. With this method, we can identify 14 El Niño winters: 1982, 1986, 1987, 1991, 1994, 1997, 2002, 2004, 2006, 2009, 2014, 2015, 2018, and 2019; and 15 La Niña winters: 1983, 1984, 1985, 1988, 1995, 1998, 1999, 2000, 2005, 2007, 2008, 2010, 2011, 2017, and 2020. The NAO index is defined by the difference in normalized SLP anomalies between 35°N and 65°N over the North Atlantic sector27.

To investigate the possible roles of ENSO-related SST and precipitation forcing in different regions, several indices are defined. The tropical CP SST index (SST_CP), western Pacific SST index (SST_WP) and Indian Ocean SST index (SST_IO) are defined by the SST anomaly averaged in the regions of 5°S–5°N/170°W–120°W, 5°S–20°N/120°–160°E and 20°S–20°N/40°–110°E, respectively (Supplementary Fig. 3a). The tropical CP precipitation index (Pr_CP), western Pacific precipitation index (Pr_WP) and western Indian Ocean precipitation index (Pr_WIO) are defined by the precipitation anomaly averaged in the regions of 5°S–5°N/160°E–180°–120°W, 5°S–20°N/110°–150°E and 5°S–10°N/40°–80°E, respectively (Supplementary Fig. 3b).

WAF

To analyze the source and direction of wave energy propagation, the wave activity flux (WAF) developed by Takaya and Nakamura48 is applied. The WAF is parallel to the local group velocity that corresponds to the stationary Rossby waves and is independent of the wave phase48. It has been considered a useful tool for supplying information about wave propagation and is defined as:

$$F=\frac{p\,\cos \phi }{2|U|}\left\{\begin{array}{l}\frac{U}{{a}^{2}{\cos }^{2}\phi }[{(\frac{\partial \psi ^{\prime} }{\partial \lambda })}^{2}-\psi ^{\prime} \frac{{\partial }^{2}\psi ^{\prime} }{\partial {\lambda }^{2}}]+\frac{V}{{a}^{2}\,\cos \phi }(\frac{\partial \psi ^{\prime} }{\partial \lambda }\frac{\partial \psi ^{\prime} }{\partial \phi }-\psi ^{\prime} \frac{{\partial }^{2}\psi ^{\prime} }{\partial \lambda \partial \phi })\\ \frac{U}{{a}^{2}\,\cos \phi }(\frac{\partial \psi ^{\prime} }{\partial \lambda }\frac{\partial \psi ^{\prime} }{\partial \phi }-\psi ^{\prime} \frac{{\partial }^{2}\psi ^{\prime} }{\partial \lambda \partial \phi })+\frac{V}{{a}^{2}}[{(\frac{\partial \psi ^{\prime} }{\partial \phi })}^{2}-\psi ^{\prime} \frac{{\partial }^{2}\psi ^{\prime} }{\partial {\phi }^{2}}]\end{array}\right.$$
(1)

where p is the pressure normalized by 1000 hPa, a is Earth’s radius, φ is the latitude, and λ is the longitude. The geostrophic stream function ψ is defined as z/f, where z is the geopotential, and f (= 2Ωsinφ) is the Coriolis parameter with the Earth’s rotation rate (Ω). Also, |U | , U, and V represent the basic states of wind speed and zonal and meridional wind, whereas ψ’ denotes the perturbed stream function.

Statistical significance test

All our results are tested based on the two-sided Student’s t test.