1 Introduction

The El Niño–Southern Oscillation (ENSO), as an evident interannual fluctuation signal originating in the tropical Pacific, has great impacts on agriculture, ecosystems, fisheries, even financial markets and public health worldwide (e.g., McPhaden et al. 2006; Capotondi et al. 2013). Focusing on observation, theoretical model, and prediction, ENSO studies in the last decades have dramatically improved our understanding of ENSO dynamics and significantly increased the ENSO forecast skills up to more than one year (e.g., Rasmusson and Carpenter 1982; Zebiak and Cane 1987; Latif et al. 1994; Kirtman and Schopf 1998; Park et al. 2018). However, the prediction skills became relatively low for much of the first decade of the 2000s (Barnston et al. 2012). Some researchers suggest that the reduced ENSO predictability may associated with some unusual changes in the tropical Pacific (Zhao et al. 2016). Those changes in the average atmosphere/ocean state conditions, potentially driven by climate change, may also lead to diverse ENSO behaviors over the last few decades. (Yeh et al. 2009).

Indeed, El Niño events have behaved more diversely since the late 1990s (Lee and McPhaden 2010). The zonal locations of the maximum sea surface temperature (SST) anomaly (SSTA), as one of the important aspects of ENSO diversity, have drawn wide attention in the recent decades. Research in recent decades have primarily focused in separating ENSO events into two groups: the eastern Pacific (EP) El Niño event with the maximum SSTA located in the eastern equatorial Pacific, and the central Pacific (CP) El Niño event with the maximum SSTA located in the central equatorial Pacific (e.g., Ashok et al. 2007; Kao and Yu 2009; Kug et al. 2009). Although he classifications of different El Niño types are generally based on the assessment during boreal winter, the impacts of different types of El Niño in developing summer cannot be neglected. For example, the boreal summer El Niño Modoki could induce excessive rainfall in southern China and rainfall deficiencies in the low reaches of the Yangze River vally (Weng et al. 2007), but for the EP El Niño, the impacts on southern China rainfall are limited (Feng et al. 2016). During CP El Niño developing summer, there is a triangle-pattern of precipitation anomalies with wet conditions in both southern and northern China, but dry conditions in the north of the Yangtze River. But during EP El Niño developing summer, there is a dipole-pattern of precipitation anomalies with wet conditions in Southeast China and dry conditions in North China (Wen et al. 2020). Besides, different boreal summer El Niño types also have different impacts on tropical cyclone. For instance, tropical cyclone formation tends to be enhanced over the central to eastern Pacific and decreased (increased) in the northwestern (northeastern) part of the western (eastern) Pacific during EP El Niño developing summer. And because of the westward shift of positive SSTA during CP El Niño developing summer, a descending motion of dry air in the northeastern Pacific reduce the tropical cyclone activity in this region (e.g. Kim et al. 2011). Therefore, different types of El Niño in developing summer can also provide valuable information for seasonal climate prediction.

To predict the El Niño, different atmospheric/oceanic precursors or “triggers” which have a close relationship with the initiation of El Niño events cannot be ruled out (Capotondi et al. 2013). The relationship between the ENSO and the North Pacific precursors have been widely investigated in the last decades (e.g., Chang et al. 2007; Yu et al. 2010; Su et al. 2014; Ding et al. 2015). Focusing on the relationship between the ENSO and North Pacific atmospheric variability. Vimont et al. (2001, 2003) proposed a mechanism called “seasonal footprinting mechanism” (SFM) to explain the extratropical-to-tropical connection between the North Pacific Oscillation (NPO, Rogers 1981) and the ENSO. As a key source of ENSO complexity (Yu and Fang 2018), this mechanism highlights that fluctuations in the southern lobe of NPO can modulate the subtropical northeasterly tropical trade winds during boreal winter, which in turn force the ocean through surface heat flux anomalies and leave a SST “footprint”. The SST footprint extended from extratropical to tropical also show a meridional SST gradient, termed as the North Pacific Meridional Mode (NPMM) (Chiang and Vimont 2004; Chang et al. 2007). Through the NPMM conduit and the wind-evaporation-SST feedback (WES, Xie and Philander 1994), the extratropical surface energy flux and zonal wind stress anomalies can extend to the tropics and may initiate ENSO events in the following seasons. Some recent studies also found that the NPO preceded CP ENSO events (Yu et al. 2010; Yu and Kim 2011) and could modulate the amplitude of CP ENSO events (Xu et al. 2020). The NPO-forced victoria mode variability also closely related to the El Niño diversity (Shi et al. 2023). Besides the SFM, the subtropical lobe of the NPO can also modulate the off-equatorial wind stress curl, which in turn drives the equatorward upper ocean mass transport and charges the central equatorial Pacific with warm water. These warm water can be transported eastward and upward along the equatorial thermocline by zonal advection, promoting the development of positive SSTA in the eastern equatorial Pacific. This mechanism is known as “trade wind charging” (TWC, Anderson 2004, 2007; Anderson et al. 2013a, b). The fluctuations in the NPO-like atmospheric pattern, especially in its subtropical lobe, are pivotal to the development of El Niño through either the SFM or the TWC mechanism.

Recently, Yeh et al. (2018) examined the changes of both the NPO’s spatial structure and the NPO-ENSO relationship in recent decades. Their research indicated that the NPO’s subtropical lobe experienced a significant zonal shift after the mid-1990s. Moreover, after the significant change of the NPO’s spatial structure, the simultaneous NPO-ENSO relationship during boreal winter became weak, but the relationship between the winter NPO and the following winter ENSO became close through atmosphere-ocean coupled processes. This raises questions about whether the relationship between the boreal winter NPO and the following summertime El Niño had changed after the center of the NPO’s subtropical lobe shifted to the east, and if so, how? And what role did SFM or TWC mechanism play in different periods.

It is unclear whether the boreal winter NPO is associated with the El Niño development or decaying during following summer. The purpose of this study is to confirm the relationship between the winter NPO and following summertime El Niño and examine the interdecadal change in the relationship in recent decades. Specific sections of this paper is organized as follows. Descriptions of the data and methodology used in this study are provided in Sect. 2. Changes in the NPO’s spatial structure and the relationship with the following summer El Niño are examined in Sect. 3. Mechanisms linking the NPO to El Niño are discussed in Sect. 4. Finally, a summary and discussions are provided in Sect. 5.

2 Data and methods

2.1 Data

The monthly mean SLP, 10 m zonal and meridional wind and the surface net latent heat flux are taken from the NCEP-NCAR reanalysis for the period 1980–2020 (Kalnay et al. 1996). The monthly mean sea surface wind stress, subsurface ocean current and temperature are taken from GODAS for the period 1980–2020 (Saha et al. 2006). The observed SST field is taken from HadISST for the period 1948–2020 (Rayner et al. 2003). The precipitation data is taken from GPCP v2.3 (Adler et al. 2003) for the period 1979 to present.

2.2 NPO, WWV and indicators of summertime El Niño

The NPO pattern is defined as the second empirical orthogonal function (EOF2) of SLPA in the 120°E–120°W, 20°N–60°N region for the corresponding season. The corresponding principal component time series (PC2) is defined as the NPO index. We focus on the boreal winter (December–January–February, DJF) when the NPO is most variable to better elucidate the NPO-ENSO diversity linkage and use a 15-year running window to examine the changes on low frequency timescales. The depth of the main thermocline is estimated using the depth of the 20 °C isotherm (Z20). Water above the 20 °C isotherm (Z20) is defined as ‘‘warm water’’ in this study, and warm water volume (WWV) is determined by spatial integration of Z20 over the region 5°S–5°N, 120°E–80°W (Meinen and Mcphaden 2000).The E-index (indicating SSTA in the eastern equatoial Pacific) and C-index (indicating SSTA in the central equatoial Pacific) (Takahashi et al. 2011) along with the Nino1 + 2 and Nino4 index are introduced as descriptors of El Niño in developing summer.

2.3 Liang-kleeman information flow

In order to determine the causal structure from the boreal winter NPO to the following summer tropical equatorial Pacific SSTA, we introduce the Liang-Kleeman Information Flow (IF for short). The Detailed derivations about the IF can be found in Liang (2014). Unlike classic statistical methods, the IFs represent a true physical cause-and-effect association between two variables. It opens up a new approach to investigate the relation between different variables and interpret the related physical cause-and-effect association.

\({T}_{2\to 1}\) represents the rate of IF from time series \({X}_{2}\) to \({X}_{1}\), which can be calculated with a statistical estimation of the interaction between the variances of \({X}_{2}\) and \({X}_{1}\) and their differentials. A formula easy to compute:

$${T}_{2\to 1}=\frac{{C}_{11}{C}_{12}{C}_{2,d1}-{C}_{12}^{2}{C}_{1,d1}}{{C}_{11}^{2}{C}_{22}-{C}_{11}{C}_{12}^{2}}$$
(1)

where \({C}_{ij}\) is the sample covariance between \({X}_{i}\) and \({X}_{j}\) and \({C}_{1,d1}\)is the covariance between \({X}_{i}\) and the differential of the time series \(\dot{{X}_{j,n}}=\frac{{X}_{j,n+1}-{X}_{j,n}}{\varDelta t}\) where \(\varDelta t\) is the time step, which in ourcase is 1 (1 year in this study). \({T}_{2\to 1}\) has the units of information or entropy (nats/unit time). If the IF \({T}_{2\to 1}\)is non-zero, \({X}_{2}\) is considered causal to \({X}_{1}\). A bootstrap resampling procedure is introduced to compute the statistical significance with replacement of all terms included in (1) using 1,000 realizations.

2.4 Sverdrup transport

In order to elucidate the linkage between the off-equatorial Pacific atmosphere and the equatorial Pacific ocean, we also computed the Sverdrup transport \(V\) in the tropical Pacific region as in Anderson and Perez (2015).

$$V\approx \frac{1}{\beta \rho }[\frac{\partial {\tau }_{y}}{\partial x}-\frac{\partial {\tau }_{x}}{\partial y}]$$
(2)

Where 𝛽 is the latitudinal gradient of the Coriolis parameter, 𝜌=1025 kg/m3 is the density of sea water, \({\tau }_{y}\) and \({\tau }_{x}\) are the northward and eastward components of the wind stress, respectively,.

2.5 Statistical methods

Both correlation and linear regression analysis are used in this study. A two-tailed student’s t-test with n-2 degrees (n denotes the number of years) of freedom is used for the statistical significance of correlation analysis. A F-test is used for the statistical significance of linear regressions. Nonparametric tests against trend are used for the statistical significance of linear trend (Mann 1945). All the anomalies are calculated relative to the 1991–2020 climatology, and linearly detrended prior to calculation.

3 The interdecadal change of the structure of NPO and its relationship with ENSO diversity

Figure 1a shows the DJF mean NPO index (NPOI) for 1980–2019. The corresponding EOF patterns are represented in Fig. 1b, we refer to this pattern as a typical positive NPO phase. To further characterize the NPO structural change, especially in the subtropics, the second EOF of North Pacific SLPA during each of overlapping 15-year periods from 1980 to 1994 to 2005–2019 are calculated. As shown by Yeh et al. (2018), there was a clear eastward shift of the center of the southern lobe of the NPO in recent decades. Before the period 1989–2003, the center shifted eastward and northwestward gradually. But after the period 1989–2003, the most significant eastward shift in recent decades occurred (Fig. 1c). The eastward shifts of NPO’s southern lobe SLPA from west to the east (Fig. 1d–h) are explicit.

Fig. 1
figure 1

a Time series of the PC2 of winter mean (December–January–February) SLPA in the 120°E–120°W, 20°N–60°N region for 1980–2019. b The EOF2 of winter mean SLPA in the North Pacific for 1980–2019. c Locations of the maximum anomaly centers of the NPO’s southern lobe with a 15-year running window for 1980–2019. The middle years of the calculated periods are shown above the locations. For example, 2000 indicates the anomaly center of the southern lobe in the NPO for 1993–2007. d–h are the same as in c except that they cover the periods of 1980–1994, 1985–1999, 1990–2004, 1995–2009, 2000–2014, and 2005–2019, respectively. The explained variances of NPO are also shown in the brackets behind the corresponding periods in b and d–h

Full size image

To examine the relationship between the boreal winter NPO and El Niño in developing summer, we calculated the correlation coefficients between the NPOI and the June–July–August (JJA) mean E-index (Niño1 + 2 index) and C-index (Niño4 index) in the following year during each of overlapping 15-year periods from 1980 to 1994 to 2005–2019. It is clear that the boreal winter NPO is associated with the El Niño during following summer. And it is interesting that the correlation coefficients between the NPOI and the JJA E-index (Niño1 + 2 index) show a significant increasing trend since the period 1991–2005, but a significant downward trend for the correlation coefficients between the NPOI and the JJA C-index since the period 1994–2008 (Fig. 2a, both trends are significant at the 95% level). We can also see similar trends over 1980–2019 in Fig. 2b and c (also significant at the 95% level). When the southern lobe of the winter NPO located east of its average position, it tends to promote the development of eastern equatorial Pacific SSTA in the following summer (Fig. 2b). But after the southern lobe shift to the west, it is more conducive to the development of central equatorial Pacific SSTA (Fig. 2c).

Fig. 2
figure 2

a 15-year correlation between D(0)JF(+ 1) NPOI and JJA(0) E-index/C-index. The x-axis shows the middle year of the period. For example, 1990 indicates the period 1983–1997. ‘0’ donates the ENSO year, and ‘-1’ means the preceding year. The black, green and magenta dotted lines indicate 80%, 90%, 95% confidence level, respectively. b Scatter plots for the locations of the maximum anomaly centers of NPO’s southern lobe (x-axis, Units: °) and correlation coefficients between D(-1)JF(0) NPO index and JJA(0) E-index (y-axis) in each 15-year. c As for b, but the y-axis indicates correlation coefficients between D(-1)JF(0) NPO index and JJA(0) C-index. The solid lines in b and c are the best fits

Full size image

The results we show above are based on time-delayed correlation analysis, which is still the major tool in climate science. But it is worth noting that, without carrying the needed directedness or asymmetry, the correlation does not necessarily imply causality. Here we use the IF as a more robust method to determine the “causal structure” from the NPO to the tropical equatorial Pacific SSTA. To examine causality and the mechanisms linking the boreal winter NPO to the El Niño in developing summer, we will focus on two different periods before and after the significant eastward shift: One is 1989–2003 when the center of the NPO’s southern lobe moved westward with an averaged westernmost center at around 165ºE, 42.5ºN. Although it seems that 1997–2011 is another choice because the NPO’s southern lobe had moved easternmost an averaged center at around 140ºW, 40ºN, the overlap between the two period in time domain may affect the results about mechanisms linking NPO to El Niño. Therefore, we choose the period 2005–2019 when the center of the NPO’s southern lobe shifted eastward with an averaged center at around 152.5ºW, 35ºN to avoid the overlap in time domain.

The spatial distributions of the IF, or “causal structure” from the NPO to the tropical equatorial Pacific SSTA are shown in Fig. 3. Significant IFs for NPO are detected over the central equatorial Pacific (160ºW–120ºW) for the period 1989–2003 (Fig. 3a). But for the period 2005–2019, significant IFs for NPO are detected over the eastern equatorial Pacific (90ºW–80ºW) (Fig. 3b). The results indicate that the boreal winter NPO cause SSTA in central (eastern) equatorial Pacific as its southern lobe shift to the west (east) the following summer. It also distinguishes the role of NPO on the diversity of ENSO under two different climatological background states from another perspective. Recent studies have also found that the transfer of predictability between dynamical events is possible and can be rigorously derived from first principles in physics (Liang and Kleeman 2005; Liang 2016, 2018). As IF tells the transfer of practicability, the winter NPO forms a predictor for El Niño in the developing summer, consistent with the results of time-delayed correlation analysis.

Fig. 3
figure 3

a The information flows from delayed series of NPOI to the SSTA in the tropical equatorial Pacific during 1989–2003 (in nats/year). b As for a but for the period 2005–2019. The information flows significant at a 90% confidence level are shaded

Full size image

However, the linkage between NPO and El Niño cannot be well explained just by IF. To better elucidate the dynamic response of the equatorial Pacific Ocean to different NPO-like atmospheric forcing, time-delayed correlation and regression analysis are used to investigate the physical mechanisms behind the linkage in the following section, including discussions on the role of the SFM and the TWC mechanism in the development of El Niño events.

4 Mechanisms linking NPO to El Niño

4.1 The TWC mechanism and El Niño

The lagged-regression maps of the anomalous vertically-integrated meridional oceanic mass transport and ocean heat content onto the standardized NPOI for the period 1989–2003 and 2005–2019 are examined to elucidate the relationship between the eastward shift of the NPO’s southern lobe and the development of summertime El Niño. From winter (DJF) to early spring (February–March–April, FMA) for the period 1989–2003, the anomalous wind stresses associated with the NPO promote the anomalous equatorial convergence of warm water over 160ºE–160ºW in both hemispheres (Fig. 4a, b), charging the central and eastern equatorial Pacific subsequently in accordance with the TWC mechanism, as shown in Fig. 4c, d. For the period 2005–2019, by contrast, the anomalous equatorial convergence of warm water over 160ºE–160ºW in both hemispheres, especially in southern hemisphere, are not as significant as the period 1989–2003 during winter and early spring (Fig. 4e, f). Moreover, the positive ocean heat content built in the eastern equatorial Pacific earlier than that during the period 1989–2003 (Fig. 4b, f). As shown in Fig. 5, for both periods, significant positive subsurface temperature anomaly has formed in equatorial central pacific (5ºS–5ºN, 180º–150ºW) during FMA. But for the period 2005–2019, the eastward propagation of the subsurface warm water is faster. The eastern equatorial Pacific warming is evident in late spring and summer (Fig. 5e, f). While for the period 1989–2003, the eastward extension of warm water are relatively slow and confined in the area 5ºS–5ºN, 180º–120ºW (Fig. 5b, c). Focusing on the build up of positive ocean heat content, the ocean temperature anomalies and climatologic mean meridional ocean currents (averaged in the 180º–150ºW) before spring are also calculated. For the period 1989–2003, it is clear that the subsurface positve temperature anomalies in the equatorial region changed gradually from DJF to FMA. The equatorward migration of subsurface warm water from 2ºN–7ºN and 5ºS–10ºS to the equator can be seen explicitly (Fig. 6a–c, similar to Anderson et al. 2013a, Fig. 3). Characterized by equatorward currents at the subsurface and poleward currents at the surface, the subtropical cells (STCs, e.g., Capotondi et al. 2005) control the equatorward movement of the off-equatorial warm water. Its asymmetric structure can also explain why warm waters in Southern Hemisphere play a more important role. By contrast, the equatorward migration of off-equatorial warm water is not significant for the period 2005–2019 (Fig. 6d–f). The above results indicate that it may favor the development of moderate El Niño in developing summer if the NPO’s southern lobe located west of its average position, and provide evidences for the variation of the equatorial Pacific warm water volume associated with the TWC mechanism. And after the NPO’s southern lobe shift to the east of its average position, the TWC mechanism associated with the NPO may not significantly promote the development of El Niño. Considering subsurface warm water have already extended to the eastern equatorial Pacific in late spring and early summer, we suppose the SFM and related waveguide warming maybe more important to the development of subsurface temperature anomalies for the period 2005–2019.

Fig. 4
figure 4

Lag regression of the anomalous vertically integrated meridional oceanic mass transport (shading, with intervals of 0.1 m2 s− 1) and ocean heat content (vertically averaged temperature integrated from 0 to 300 m; contours, with intervals of 0.1 °C) onto the standardized NPOI for a December–February (DJF), b February–April (FMA), c April–June (AMJ), and d June–August (JJA) during 1989–2003. e–h are the same as a–d but for the period 2005–2019. ‘0’ donates the ENSO year, and ‘-1’ means the preceding year. Only positive ocean heat content anomalies are shown. Zero contour omitted. Stippling indicates statistically significant anomalous V regression coefficients at the 90% level

Full size image
Fig. 5
figure 5

a–c Mean temperature anomalies averaged within 5°N–5°S (contours, with intervals of 0.25 °C) regressed onto the standardized NPOI for the period 1989–2003. d–f are the same as a–c but for the period 2005–2019. Red (blue) line contours indicate positive (negative) values. The values that significant at the 90% level are shaded

Full size image
Fig. 6
figure 6

a–c Mean ocean temperature anomalies (contours, with intervals of 0.25 °C) regressed onto the standardized NPOI for the period 1989–2003 and climatological mean meridional ocean currents (vectors) averaged in the central Pacific (180º–150ºW). d–f are the ame as a–c but for the period 2005–2019. ‘0’ donates the ENSO year, and ‘-1’ means the preceding year. Red (blue) line contours indicate positive (negative) values. The values that significant at the 90% level are shaded

Full size image

4.2 The SFM and El Niño

The SFM hypothesis suggests that the net surface heat flux are pivotal to the formation of SST footprint. Analysis of the individual heat flux components performed by Vimont et al. (2003) indicated that the latent heat flux is the dominate component over most the North Pacific. Here we further examine the regression maps of 10-m wind, SST and latent heat flux anomalies onto the standardized NPOI for the period 1989–2003 and 2005–2019 respectively. For the period 1989–2003, there is no obvious NPO-like variability. Although the surface wind and SST anomalies during winter and early spring are significant in the central Pacific 0º–20ºN, 160ºE–150ºW (Fig. 7a, b). the latent heat flux anomalies are not significant in the same area (Fig. 7e, f). This is because the wind speed change are too weak to significantly influence local evaporation. As a result, during late spring and summer, the westerly wind and related positive latent heat flux anomalies are barely presented in the equatorial Pacific region east of 160ºE (Fig. 7c, d, g, h). Only weak positive SSTA emerged in the central equatorial Pacific during summer (Fig. 7d). Consequently, the development of weak SSTA in central equatorial Pacific cannot be explained by the SFM. For the period 2005–2019, in contrast, the mid-latitude NPO forcing induces a significant NPMM-like pattern during winter (Fig. 8a). Because of surface wind anomalies associated with the NPO weaken the climatological winds, the upward latent heat flux from the ocean surface is reduced then and further lead to local ocean surface warming through a negative WES feedback (Fig. 8e). Unlike the period 1989–2003, the NPO induced equatorial westerly wind anomalies around the dateline are more evident during early spring (Fig. 8b). The anomalous westerlies further generate downwelling equatorial Kelvin waves, increasing the heat content in the equatorial eastern Pacific and promoting the occurrence of positive SSTA in the eastern equatorial Pacific (Fig. 8c, d). These processes can explain the quick eastward moving of subsurface warm water (Fig. 5d, e) which is consistent with the SFM proposed by Vimont et al. (2001).

Fig. 7
figure 7

10-m wind (vectors), SST (shading, with intervals of 0.2ºC) and downward latent heat flux (shading, with intervals of 2.5 W m− 2) anomalies regressed onto the standardized NPOI for the period 1989–2003. ‘0’ donates the ENSO year, and ‘-1’ means the preceding year. Only surface wind vectors significant at the 80% level are shown. Stippling indicates statistically significant SST/latent heat flux anomaly regression coefficients at the 90% significance level

Full size image
Fig. 8
figure 8

As in Fig. 7. But for the period 2005–2019

Full size image

It is noteworthy that the positions of La Niña SST forcing during the previous winter are able to modify the spatial structures of the NPO (Park et al. 2013). Whether the state of equatorial SSTA in previous winter makes a difference in the relationship between the NPO and the following El Niño? To answer this question, we subtract the linear least squares fit to the DJF mean cold tongue index (SSTA averaged over 6°S–6°N and 180°–90°W) from all the all fields in the following seasons prior to analysis and reexamine the regression maps of 10-m wind, SST and latent heat flux anomalies onto the standardized NPOI for the period 1989–2003 and 2005–2019 respectively (not shown). We notice that the relationships of the NPO and the following El Niño is nearly same as that in Figs. 7 and 8, which indicates that the state of equatorial SSTA in previous winter makes little difference in the relationship of the NPO and the following El Niño.

4.3 Mechanisms altered by the NPO

Focusing on the NPO-induced zonal wind and latent heat flux anomalies over the 10°N–20°N, 160°E–150°W region south of the NPO’s subtropical lobe, the time evolution of 15-year correlation coefficients of the NPOI with the boreal spring (March–April–May, MAM) 10 m zonal wind anomaly and the latent heat flux anomaly are calculated (Fig. 9). Considering an effective SFM, in response to anomalous south-westerlies associated with the positive phase of the NPO, the northeasterly trade winds would weaken and subsequently enhance the downward latent heat flux, warming the subtropical northeastern Pacific and leading to a negative wind-evaporation-SST (WES) feedback there. The thermodynamic coupling over this region favors the southwestward development of positive SSTA in the subtropical northeastern Pacific and equatorward shift of surface zonal wind anomalies. The westerly component of the anomalous surface winds along the equator would further promote the development of positive SSTA in the eastern equatorial Pacific. The significant (at 95% level) increasing trends after the period 1994–2008 for both the red and the blue lines in Fig. 9 indicate that the eastward shift of the NPO’s southern lobe can alter the response of zonal wind and related latent heat flux anomalies over that region. The SFM also become more effective after the significant eastward shift of the NPO’s southern lobe.

Fig. 9
figure 9

Time evolution of 15-year correlation coefficients of the D(-1)JF(0) NPO index with the MAM(0) 10 m U anomaly (red line), the MAM(0) latent heat flux anomaly (blue line), the D(-1)JFMAM(0) vertically integrated meridional oceanic mass transport anomaly (green line), the JJA(0) WWV anomaly (magenta line). The black, green and magenta dotted lines indicate 80%, 90%, 95% confidence level, respectively. The year marked in x-axis shows the middle year representing each 15-year period. For example, 1996 indicates the period 1989–2003

Full size image

The subsurface heat content transport associated with the variation of North Pacific trade winds are pivotal to the effectiveness of TWC mechanism. For the period 1989–2003, as shown in Figs. 4a and b and 6a–c, the anomalous equatorward mass transport in the region of 5°S–15°S, 180°W–150°W were pivotal to the accumulation of warm water in central equatorial Pacific (Fig. 5a–c). We further calculate the time evolution of 15-year correlation coefficients of the NPOI with the December–May(DJFMAM) mean vertically integrated meridional oceanic mass transport anomaly in this region (Fig. 9, green line), the correlations were insignificant at 90% level after the period 1989–2003. Its decreasing trend is also consistent the correlation between boreal winter NPO and summer WWV anomaly (Fig. 9, magenta line), which indicates that the eastward shift of the NPO’s southern lobe had also altered the TWC mechanism after the period 1989–2003.

The altered SFM and TWC mechanism are closely linked with the interdecadal changes of the spatial structure of the boreal winter NPO in recent decades, which would further contribute to different types of El Niño in developing summer.

5 Summary and discussion

This study confirms the relationship between the boreal winter NPO and the El Niño in the following summer. Our analysis also reveals an interdecadal change in this relationship. It tends to promote the development of SSTA in the central equatorial Pacific in the following summer when the subtropical lobe of the NPO located west of its average position. But after a significant eastward shift of NPO’s subtropical lobe, it is more conducive to the development of SSTA in the eastern equatorial Pacific. Both the SFM and the TWC associated with the NPO were altered with the eastward shift of the NPO’s subtropical lobe. The changes in the SFM and the TWC mechanism could further influence the development of summertime El Niño.

The development of SSTA in the central equatorial Pacific can be well explained by the TWC mechanism. The meridional heat transport processes are relatively slow but pivotal to the equatorial Pacific warming in this case. Although the subsurface warm water already exist in central equatorial Pacific during early spring, it takes about 3 ~ 4 months for the warm water to come to the surface because of the relatively weak equatorial zonal wind anomalies associated with the anomalous westward movements of the NPO’s southern lobe. When the southern lobe located east of its average position, the NPO related SLP variations tend to favor the development of SSTA in the eastern equatorial Pacific in summer. The SFM plays a more crucial role in the development of westerly momentum in western and central equatorial Pacific. The westerly momentum in turn excite eastward propagating downwelling Kelvin waves and promote the eastward propagation of warm water. The eastward propagating Kelvin waves and related waveguide warming are important to the development of positive SSTA over the eastern equatorial Pacific (e.g. Gebbie et al. 2007).

One complicated and important question may arise concerning what causes such a significant zonal shift in the winter NPO’s subtropical lobe. The meridional movement of the mid-latitude jet stream is one of the potential explanations. Park and An (2014) investigated the relationship between the western tropical Pacific (100ºE–145ºE, 0º–20ºN) atmospheric convection (WTPC) and the North Pacific atmosphere circulation (NPAC) in boreal winter and found that the mid-latitude jet stream as an intermediary was pivotal to the WTPC-NPAC relationship. A strong WTPC can lead to significant meridional movement of the mid-latitude jet stream. The regression maps of the seasonal-mean DJF zonal wind at 200 hPa against the NPOI for period 1989–2003 and 2005–2019 are shown in Fig. 10a and b. It is oblivious that there is a significant northward and eastward extension of the westerly wind anomaly for the period 2005–2019. The difference of DJF mean upper-level zonal wind (200 hPa) between the two periods (Fig. 10c) also shows a similar pattern as Fig. 10b. Such extension is mainly located over the Northeast Pacific. Based on results proposed by Park and An (2014), the extended westerly wind anomalies at 200 hPa for the period 2005–2019 are associated with precipitation perturbations and heating anomalies over the western north Pacific. Difference in DJF mean precipitation between the period 2005–2019 and 1989–2003 is showed in Fig. 10d. For the period 2005–2019, there were significant positive precipitation anomaly over the western north Pacific. It has been found by Guo et al. (2015) that this kind of precipitation anomaly as a heat source facilitate the northward propagation of eddy momentum in the upper troposphere. The northward eddy momentum flux favor the strengthen and northward displacement of the subtropical westerly jet. The equivalent barotropic structure is dominated in boreal winter and following spring over the mid-latitude North Pacific. The positive vorticity anomalies (Fig. 10e) associated with the northward and eastward extended jet in the upper atmosphere contribute to the development of negative SLPA over Northeastern Pacific near California, which in turn cause a eastward shift of NPO’s subtropical lobe. Moreover, the WTPC can also regulate the tropical Pacific Walker circulation and reduce the eastern equatorial Pacific SSTA, which favor the maintenance of a La Niña-like mean state (Park and An 2014). The Rossby wave responses to this state cause significant changes in atmospheric baroclinicity over the extratropical North Pacific and regulates the rate of available potential energy conversion that contributes to the zonal displacement of NPO (Sung et al. 2019). There also exist other potential causes for the zonal shift of NPO’s subtropical lobe. For example, Chen et al. (2013) highlighted the significant modulation effect of springtime Arctic Oscillation on the NPO-ENSO relationship. A positive phase NPO in boreal winter favor the development of El Niño if the following spring Arctic Oscillation is positive. But after the early 1990s, the sea ice in the Arctic is significantly reduced. The rate of loss in the extent of the ice cover is widespread and has accelerated after 2000 (Perovich and Richter-Menge 2009). It remains unclear that to what extent the loss of sea ice cover can modulate the AO-NPO-ENSO relationship.

Fig. 10
figure 10

The regression map of DJF mean zonal wind anomalies (contours, with intervals of 1 m s− 1) at 200 hPa against the standardized NPOI for a the period 1989–2003 b the period 2005–2019. c Difference in DJF mean zonal wind at 200 hPa between the period 2005–2019 and 1989–2003. d Difference in DJF mean precipitation between the period 2005–2019 and 1989–2003 e The regression map of DJF mean vorticity anomalies (contours, with intervals of 2 × 10− 7 Pa m− 1) at 200 hPa against the standardized NPOI for the period 2005–2019. 200-hPa zonal wind climatology are indicated by black contours in a–c with intervals of 10 m s− 1. Red (blue) line contours indicate positive (negative) values, and the values that significant at the 90% level are shaded in a, b and e

Full size image

The results obtained in the present study also provide implications for continued research on the relationship between extratropical precursors and ENSO diversity. For instance, the relationship between the NPMM and ENSO has experienced an interdecadal change in 1980 (Lu et al. 2022), but to what extent the variation of the NPO’s spatial structure can influence that relationship remains unclear. Moreover, through both the SFM and the TWC mechanism, the NPO forced Victoria mode can contribute to the development of ENSO (Ding et al. 2015; Ji et al. 2023). Further research is also needed to clarify the relationship between the Victoria mode and the diversity of ENSO before/after the significant change of the NPO’s spatial structure. Besides, it is also noteworthy that there has a trend to use long-term time series of EOF principal component or other indexes when exploring the relationship between extratropical precursors and ENSO. The existence of inter-decadal change in the NPO-ENSO diversity relationship remind us that it is pivotal to examine the underlying physical processes which may influence the behavior of ENSO before using long-term time series to investigate the relationship between extratropical precursors and ENSO diversity.

After the 1990s, the CP El Niño, in contrast to the EP El Niño, has been suggested increased intensity and frequency during the past few decades (Lee and Macphaden 2010). It is also reported that the center of the southern lobe SLP in the NPO during boreal winter is shifted to the east after the mid-1990s compared to before the mid-1990s (Yeh et al. 2018). This study further argued that after the NPO’s southern pole moves eastward, it is more likely to cause SSTA in the eastern equatorial Pacific. Our argument seems to reveal a contradiction at first glance. However, there is no contradiction because ENSO is not defined during summer when either ENSO decays or develops. We cannot determine the ENSO diversity just by assessing the early development phase of ENSO in summer. Similarly, we cannot determine the ENSO diversity just by assessing the impact of the NPO. There are also other precursors in different ocean basins may favor the development of different ENSO types. For example, two of the strongest recent El Niño events (i.e. 1982–1983, 1997–1998) were preceded by boreal summer Atlantic Niña events, suggesting Atlantic Niñas tend to favor eastern EP El Niños (Martin-Rey et al., 2015). The positive phase of the South Pacific Oscillation in boreal summer promotes anomalous equatorward ocean mass transport in the eastern equatorial Pacific, facilitating the growth of warm SST anomaly there and thus promoting EP-type ENSO events (You and Furtado 2017). The South Pacific meridional mode (Zhang et al. 2014; Min et al. 2017), as robust air-sea coupled modes in South Pacific, have been found to play a role in the development of EP ENSO events. These precursors are not fully understood yet. Moreover, there are extra-tropical predictors in both hemispheres, how do the predictors in South Pacific synergistic with North Pacific predictors and tropical Pacific antecedent conditions in the development of ENSO remains unclear. The short-term forced response of predictors in both hemispheres under different background states and the joint impact on ENSO diversity also need further investigation based on idealized model experiments.