Impacts of Tropical Indian Ocean Decadal Variability on ENSO Transition Asymmetry

1.   INTRODUCTION

El Niño-Southern Oscillation (ENSO) is the most prominent driver of interannual variability in tropical air-sea interactions, with profound influences on climate, ecosystems, and human activities worldwide (Alexander et al. 2002; McPhaden et al. 2006; Fang and Xie 2020). Broad theories exist on the phase transition of ENSO, involving the delayed oscillator (Suarez and Schopf 1988), the recharge and discharge oscillator (Jin 1997a, 1997b) or the zonal mean thermocline oscillator (Li 1997), the western Pacific oscillator (Weisberg and Wang 1997), the zonal advection oscillator (Picaut et al. 1997) and the unified oscillator (Wang 2001). The above collectively supports a traditional but canonical cyclic view that ENSO develops and decays between the warm phase (El Niño) and the cold phase (La Niña) continuously. Both warm and cold episodes tend to last 1–2 years and reoccur every 3–8 years approximately (Wu et al. 2019). Since nonlinearity has been proven to smash the simple mirror images (Sun and Liu 1996; An and Kim 2017), the complexity of ENSO (Timmermann et al. 2018; Chen et al. 2022; Chen and Fang 2023), especially asymmetry, has been investigated for many years (Kang and Kug 2002; An et al. 2020). Most of these studies focused on asymmetries in the amplitude (Burgers and Stephenson 1999) and duration (Okumura and Deser 2010; Okumura et al. 2011), while further research suggested that it is also exhibited in phase transition (Ohba and Ueda 2007; Fan et al. 2023). Specifically, (1) almost all El Niños decay and vanish rapidly after their maturity, while La Niñas tend to persist into the next year; (2) a strong El Niño is frequently followed by a multi-year La Niña, while the opposite scenario rarely happens.

Various mechanisms have been proposed to explain the asymmetry in the duration of ENSO. As a stronger tendency in the transition from one phase to another corresponds to a shorter duration of the event, while a weaker tendency in the transition harmonizes with a longer duration (An and Kim 2017), duration asymmetry is often associated with transition asymmetry (TA) (Choi et al. 2013). A regular consensus is that asymmetric atmospheric nonlinearity without invoking external forcing is key to asymmetric transition and evolution (An and Kim 2017; Ohba and Ueda 2009; Dommenget et al. 2013). This nonlinearity includes the feedback process from the wind stress and the surface heat fluxes (Bayr et al. 2021; Fang and Chen 2023; Fang et al. 2015; Chen et al. 2018; Chen et al. 2019). For example, positive sea surface temperature anomalies (SSTA) tend to generate stronger zonal wind stress anomalies as well as thermocline response than negative ones of equal magnitude (Frauen and Dommenget 2010), which then leads to a quicker termination of El Niño and a quicker transition to La Niña. On the contrary, Im et al. (2015) believed that these asymmetrical responses to SSTA are not the primary factor in such ENSO asymmetry. Therefore, the present study aims to explain the possible causes of TA from other perspectives. Furthermore, there is no consensus on the generation and evolution of multi-year La Niña events (Hu et al. 2014; DiNezio et al. 2017; Liu et al. 2023), which are also related to TA. Some believe that multi-year La Niña is the result of the strong negative phase interrupting the recharge process as well as the absence of oceanic downwelling Kelvin waves (Hu et al. 2014), but this does not apply to all events (Liu et al. 2023). Given the above, it is still worth exploring whether there exist other factors driving the asymmetric transition of El Niño and La Niña.

Moreover, great breakthroughs have been made in ENSO prediction due to more abundant observational data and more accurate numerical models (L'Heureux 2020). Given all these conditions, simulation and prediction of ENSO asymmetry seem to be easy, but actually, both Coupled Model Intercomparison Project Phase 5 (CMIP5) and Phase 6 (CMIP6) underestimate ENSO asymmetry in the SST field (Zhao and Sun 2022) in which the impact of neighbor ocean basins is rarely mentioned. Given that the global atmosphere is connected around, and ENSO is a highly complicated system, it could be deduced that externally driven changes like other ocean basins SSTA might be one of the potential influences of TA, such as the Indian Ocean (IO) (An and Kim 2018). It has been confirmed that inter-basin interactions between the Pacific Ocean (PO) and IO greatly affect the evolution pattern of ENSO (Kug and Kang 2006; Wang and Wang 2021) as well as multi-year ENSO (Kim et al. 2020; Lin and Yu 2023). With the increasing attention towards multi-year ENSO, especially multi-year La Niña in recent years (Wu et al. 2019; Fan et al. 2023), quantifying the extent of contributions from other ocean basins holds considerable significance for research.

Statistical analyses have indicated that the changing surface conditions of IO are strongly related to ENSO (Alexander et al. 2002). A widely accepted argument is that the IO basin-wide SSTA contributes to the transition from El Niño to La Niña (Bayr et al. 2021; Ohba and Udea 2005), within which zonal wind stress anomalies over the equatorial north western Pacific (NWP) plays a crucial role (Weisberg and Wang 1997; Ohba 2013; Zhang and Huang 2011; Li et al. 2017). For example, the meridional shift of such zonal wind stress anomalies is considered favorable for ENSO termination (Harrison and Vecchi 1999; Gong and Li 2021). Numerous research documented that the IO basin-wide warming (IOBW) lasts till boreal spring or even summer after El Niño, enhancing the anticyclonic anomalies in the lower troposphere over the NWP (Annamalai et al. 2005; Wu et al. 2009; Xie et al. 2009). Subsequently, such anticyclonic anomalies generate anomalous equatorial easterlies that conduce to the eastward oceanic upwelling Kelvin waves (Watanabe and Jin 2002). Thus, IOBW could be used as an important element to investigate the process of El Niño-to-La Niña. Intriguingly, the feedback of IO on the opposite process of La Niña-to-El Niño has not been fully elucidated, notwithstanding observational studies intuitively showed IO warming and cooling appear several months after the mature phase of El Niño and La Niña respectively (Hong et al. 2010). This overlooking is largely because most coupled general circulation models (CGCMs) fail to reproduce the observed equatorial easterly anomalies over the western Pacific (WP) after La Niña, resulting in inaccurate simulations of the duration of La Niña (Ohba et al. 2010). Moreover, some studies stressed the importance of the position of atmospheric deep convection when investigating TA (Okumura and Deser 2010; Okumura et al. 2011), namely, the position shifts westward in La Niña compared to that in El Niño, making surface winds over WP more likely to be offset or even reversed under the influence by IO. Existing research focuses on the transition of El Niño as well as internal causes of TA; however, the transition of La Niña and the influence on ENSO TA due to IO is not well understood. Consequently, further research on the remote impact of IO is urgently needed to deepen the understanding of factors behind TA and improve the forecasting skills of coupled models.

It is reported that many characteristics of ENSO have changed notably after the 1970s (Wang 2019). Decadal changes in the relationship between IO and ENSO are also intriguing. Before the 1970s, the correlation coefficient was smaller than that after the 1970s (Yuan and Li 2008). Furthermore, the tropical Indian Ocean (TIO; 20°S–20°N, 40°–110°E) is inclined to be warming more significantly than the rest of the tropic all oceans since the 1950s (Hu and Fedorov 2019). The relation between TIO and ENSO may be modulated by decadal variation of the climatological background SST (Vibhute et al. 2020; Xiao et al. 2022). For example, Han and Wang showed that the feedback of IO warming on El Niño has been weakened since the early 1990s, which may be related to the mean state of IO SST (Han and Wang 2021). Moreover, McPhaden and Zhang indicated that there also exist decadal variations in ENSO asymmetry (McPhaden and Zhang 2009). It is widely accepted that a sudden TIO warming can arrest El Niño development (Dong and McPhaden 2014), but there is no clear consensus on whether ENSO would exhibit a greater or smaller TA under the TIO SST changing on a longer time scale. Some argue that a warming IO would strengthen Walker circulation and trade winds in PO, weakening El Niño through the Bjerknes feedback (Bjerknes 1969). On the other side, a warmer IO is often accompanied by delayed stronger El Niño (Wang and Wang 2021). The elusive nature of these factors complicates the ENSO response. The primary interest of the study is in the characteristics of the TIO SSTA and TA across different periods. Subsequently, we reveal their interrelationship through the background wind anomalies. This study aims to elucidate how TIO SSTA modulates the background wind anomalies over the Pacific, thereby influencing TA. This approach offers a novel perspective on inter-basin interactions and aims to enhance the accuracy of climate prediction models. Overall, how to reformulate the conventional TA framework by integrating the influence of the TIO decadal/interdecadal variability and offer a rationale for enhancing the accuracy of La Niña prediction is worthy of further investigation.

This paper is organized as follows. The data and methods are introduced in Section 2. Then, in Section 3, the decadal variations of TIO SSTA and its relationship with ENSO are investigated. Section 4 delineates the main results, in which quantitative analyses are utilized to measure the asymmetrical transition tendency of the ENSO. Conclusions and some further discussions are presented in Section 5.

2.   DATA AND METHODS

2.1   Data

The monthly mean SST data utilized in this study are obtained from the National Oceanic and Atmospheric Administration's (NOAA) Extended Reconstructed SST Version 5 (ERSSTv5; Huang et al. 2017). Ocean reanalysis data are from the European Centre for Medium-Range Weather Forecasts (ECMWF) Ocean Reanalysis System 5 (ORAS5; Zuo et al. 2019). Moreover, monthly zonal wind flux data are provided by the National Centers for Environmental Prediction/National Center for Atmospheric Research (NCEP/NCAR) Reanalysis project (Kalnay et al. 1996). These data cover the period from March 1958 to February 2022. The 20℃ isotherm is used to identify the thermocline depth. Variables anomalies are defined as departures from the climatological calendar mean over the entire data period after removing the linear trend. The ENSO signal in the tropical PO (TPO; 20°S–20°N, 110°E–70°W) is represented by a three-month running mean of Niño3.4 region (5°S–5°N, 170°–120°W) SSTA (i.e., Niño3.4 index). An event is defined as El Niño (La Niña) when the Niño3.4 index is greater than or equal to 0.5℃ (less than or equal to −0.5℃) for a minimum of five consecutive months. Two-tailed Student's t tests at different confidence levels (90%, 95%, and 99%) are utilized in the study.

2.2   Methods

The Season-reliant Empirical Orthogonal Function (S-EOF; Wang and An 2005) analysis is a statistical approach to detect major modes of season-dependent evolution regulated by the annual cycle. A basic consideration is that interannual variability of variables strongly interacts with the seasonal cycle, while traditional EOF used in specific season would neglect the potential impact of the seasonal cycle and might miss a more complete comprehension of ENSO asymmetry. S-EOF obtains a series of seasonal evolution patterns corresponding to a single principal component (PC) at the expense of temporal resolution. Commonly, the division of a year into four seasons is represented as follows: winter (December–February, DJF), spring (March–May, MAM), summer (June–August, JJA), and fall (September–November, SON). Therefore, the matrix for the S-EOF is $S_P\left[\vec{x}^{\mathrm{MAM}}(t), \vec{x}^{\mathrm{JJA}}(t), \vec{x}^{\mathrm{SON}}(t), \vec{x}^{\mathrm{DJF}}(t)\right]$, where S_N is a spatial state vector, t is a time interval of one year, superscripts indicate the seasonal mean of corresponding seasons, and the covariance matrix for the eigenvalue analysis is $X(\vec{x}, t) X(\vec{x}, t)^T$. In this study, some changes to the original S-EOF analysis are applied to analyze the dominant modes of the year-to-year variability of multiple variables simultaneously. When constructing the covariance matrix, the spatial state vector is $\overrightarrow{x(\text {SSTA, TCDA})}$ instead of a single variable, considering the co-evolution process of SSTA and thermocline depth anomalies (TCDA) of ENSO. Therefore, an eight-pattern (4 SSTA + 4 TCDA) mode shares the same yearly PC. By this, the corresponding PC becomes a quarter of the original time series length after applying seasonal mean treatment since the matrix is built in a seasonal sequence. Such a method thus produces a multi-variable, seasonal evolution pattern with annual variation.

Moreover, the possible physical processes and dynamic linkages are explored by regression analyses. To measure the effect of one single ocean basin, the observation and reanalysis data are firstly preprocessed by removing the linear influence of a certain region. Data without the influence of M can be obtained by using the following equation:

where PC(M) represents PC obtained by EOF analysis on M, D₀ and D₁ refer to the original data and the data after removing the linear influence of M, and s represents the linear regression coefficient. Thus, TIO-inclusive and TIO-exclusive stand for the results analyzed from the original data and the data after removing the linear influence of TIO respectively.

All results are obtained after a 31-yr sliding process to illustrate the decadal variability.

3.   TIO AND ITS CHANGING FEEDBACK ON ENSO

This section analyzes some features of TIO SSTA, including its spatial and temporal pattern, and seasonal phase-locking. EOF and a 31-yr sliding correlation are applied to investigate the characteristics of decadal variability of TIO SSTA, revealing its changing feedback on ENSO and laying a theoretical foundation for subsequent research.

3.1   TIO pattern and seasonal phase locking

To illustrate the decadal variations of the TIO SSTA pattern, we conduct the EOF analysis over the entire TIO area during 1958–2022. As shown in Fig, the leading mode EOF1 explains about 40% of the total variance of TIO SSTA (Fig. 1a). It displays a basin-scale pattern that exhibits the uniform warming or cooling phenomenon, i.e., the well-known Indian Ocean Basin (IOB; Klein et al. 1999; Guo et al. 2018). The relevant IOB index is frequently defined as the area-averaged SSTA over the whole TIO to describe such a pattern. The EOF2 explains about 12% of the total variation of TIO SSTA (Fig. 1d). This mode is called India Ocean Dipole (IOD; Saji et al. 1999; Han et al. 2014) that shows a seesaw pattern characterized by warming in the west and cooling in the east in its positive phase. The sign reverses during the negative phase and could be well quantified by a simple dipole mode index (DMI), i.e., the difference in SSTA between the tropical western IO (10°S–10°N, 50°–70°E; IODW) and the tropical south-eastern IO (10°S–0°, 90°–110°E; IODE). As mentioned before, several investigations have examined the decadal modulation of TIO SSTA and its evolving influence on the climate in surrounding regions (Behera and Yamagata 2003). However, there is a lack of research on signals at longer time scales. PC1 (Fig. 1b) crudely illustrates that the IOB exhibits uniform oscillations between positive and negative phases from the 1960s to the 2010s, with a slightly larger amplitude in the positive phase. In contrast, PC2 (Fig. 1e) exhibits significant interdecadal variations, with a significantly positive IOD during the 1960s–1980s. A robust oscillation between positive and negative phases followed for the rest of the period.

Figure  1.  (a) and (d) are the first and second EOF modes of TIO SSTA (units: ℃; shading) from March 1958 to February 2022. The number at the top-right of each panel corresponds to the variance explained by the mode. (b) and (e) are the original (blue) and 31-yr lowpass filtered (red) PC series. (c) and (f) are the lead-lag correlation between the Niño3.4 index and PC1 and PC2, respectively. The red dots indicate the results are significant at the 95% confidence level. The green shading indicates the peak value.

DownLoad: Full-Size Img PowerPoint

It is worth mentioning that El Niño usually occurs during boreal spring and summer, peaks in boreal winter, and decays over the following year (Kim and An 2021; Xing et al. 2024). IOB typically exhibits a significant development after the ENSO mature phase, reaching its maximum amplitude in the following boreal spring (Zhao et al. 2023). IOD generally undergoes development in boreal summer, maturity in boreal fall, and decline in boreal winter (Jo et al. 2022). Similarly, the lead-lag correlation analysis between PCs and the Niño3.4 index demonstrates that IOB and IOD exhibit the strongest correlations with the Niño3.4 index at a 3-month lag and 3-month lead, respectively (Figs. 1c and 1f). It should be pointed out that such lead-lag correlation may not be true in the physical sense, only for calculating the series in the following 31-yr sliding correlation mathematically. The pronounced seasonal phase-locking characteristic of SSTA in the TIO basin is a foundation of the subsequent application of S-EOF analysis.

Note that the maximum correlation coefficient in Fig. 1c (0.74) is greater than that in Fig. 1f (0.30). This can be attributed to the fact that the impact of ENSO on the TIO is stronger than the reverse process (Wang 2019; Zhao et al. 2023).

3.2   Changes in relationship between TIO SSTA and ENSO

The long-term climate variations and more details of the decadal modulation of TIO SSTA on ENSO can be examined via a 31-yr sliding correlation between SSTA of the IOB (February(1)–April(1), MAM(1)), IOD (SON(0)) and Niño3.4 (D(0)JF(1)) (Fig. 2a). Here we mark months in the development year as 0 and those in the succeeding year as 1. In general, IOB was significantly correlated with ENSO over the period, displaying a stronger relationship with ENSO compared to the IOD, although the correlation was slightly weaker in the late 1990s. Moreover, the relationship between IOD and ENSO decreased rapidly after the 2000s. Given this result, TIO is divided into IODW and IODE to analyze the possible reasons for the existence of this decadal variability. Fig. 2b suggests the decadal variations in the coupling relationship between IODE and ENSO during autumn might be the primary factors leading to these results. The positive correlation between the IODW and ENSO has strengthened since the 1980s. As for IODE, the notable negative correlation observed in the early 1990s became insignificant after the late 1990s. Previous studies indicated that the feedback of the IOD and the IOB on the subsequent year's ENSO has significantly weakened since the early 1990s (Han and Wang 2021; Jo et al. 2022). From above, it can be inferred that the decadal changes in the spatial pattern of TIO, especially IODE's SSTA, may affect its feedback on ENSO, thereby further affecting TA.

Figure  2.  (a) 31-yr sliding correlation coefficients between the IOB during MAM(1), the IOD during SON(0), and the Niño3.4 during D(0)JF(1) (red line: IOB and Niño3.4; blue line: IOD and Niño3.4) from March 1958 to February 2022. The x-axis re-presents the midpoint year within the 31-yr moving window (e.g., 1980 indicates the correlation coefficient from March 1965 to February 1996). The solid line and dashed line in green denote correlation coefficients that are statistically significant at the 95% and 90% confidence levels. The shaded area represents the error range of 1 standard deviation about the mean. (b) As in (a), but for the IODW during SON(0), the IODE during SON(0), and the Niño3.4 during D(0)JF(1) (red line: IODW and Niño3.4; blue line: IODE and Niño3.4).

DownLoad: Full-Size Img PowerPoint

4.   TIO IMPACTS TA ON DECADAL TIMESCALE

The decadal variability in the spatial pattern of TIO SSTA has provided a novel perspective on the TA of ENSO. As highlighted in the introduction, prior research predominantly focused on the impact of the IO on the phase transition of El Niño, yet there exists no consensus on TIO SSTA's contributions to TA. Moerover, there are no indicators to quantify the impact of the decadal variability of TIO SSTA. This section aims to delineate the relationship between TIO SSTA and ENSO through a set of indices and to conduct an analysis specifically focusing on decades with significant effects.

4.1   Quantitative index for TA

In this study, the S-EOF analysis is applied to obtain the evolution modes of year-to-year combined SSTA and TCDA from boreal spring to winter in the TPO and TIO, as shown in the Appendix. To better explain the method in Section 3.2, the observational data from March 1982 to February 2013 is used as an example. As shown in Fig. A2, the S-EOF1, and the S-EOF2 account for 32.6% and 25.9% explaining variance against the total, respectively, and are well separated from each other after examining the North test (North et al. 1982). The S-EOF1 illustrates the evolution of ENSO from onset to maturity, characterized by the gradual meridional range expansion and amplitude intensification of SSTA throughout the season and an enhancement of the zonal contrast in TCDA (Fig. A2a). The S-EOF2 signifies the decay and transition of ENSO, in which SSTA mode reveals the attenuation of SSTA in the equatorial eastern PO and subsequent development of the opposite phase after the boreal autumn. Concurrently, the TCDA mode distinctly captures the eastward propagation of the equatorial Kelvin wave, generating the phase transition via thermocline feedback, zonal advection feedback, and other possible processes, which is consistent with Jin and An (1999). Furthermore, PC1 and PC2 exhibit the strongest correlations (0.86, at the 99% confidence level) when PC1 leads PC2 for one year. Therefore, the S-EOF1 can be perceived as an ENSO development mode, while the S-EOF2 is the phase transition mode. The S-EOF analysis in SSTA and TCDA thus efficiently depicts the comprehensive evolution cycle of ENSO.

As mentioned, the asymmetric atmospheric response is argued as the key to driving the asymmetrical persistence of El Niño and La Niña (An and Jin 2004; Liang et al. 2017). To compare the inherent nonlinearities in each decadal process quantitatively, we employ the normalized asymmetry index λ, as proposed by An and Kim (2017) to measure the degree of ENSO phase transition asymmetry. It is defined as:

where S_P and S_N represent the slopes corresponding to the positive and negative cases on the scatter diagram between PC1(t) and PC2 (t + 1); the absolute value sign is to avoid the opposite pattern in the first two modes during the application of the S-EOF analysis. Larger (smaller) λ indicates the larger (smaller) difference between the probability of El Niño and La Niña transition, and larger (smaller) ENSO TA.

4.2   TIO contribution to decadal variation of TA

This section quantifies the impact of TIO on TA on the decadal timescale. For this, a 31-yr sliding ∆λ over the TPO considering both TIO-inclusive and TIO-exclusive scenarios is conducted (Fig. 3a), and the subtraction ∆λ of TIO-inclusive minus TIO-exclusive is presented in Fig. 3b. As anticipated, the contributions of the TIO to TA vary dramatically across different periods. Before the 1980s, the TIO amplified TA, while in the post-1980s, its influence manifested in a diametrically opposite manner, reaching its maximum negative feedback in the mid-1990s.

Figure  3.  (a) 31-yr sliding TA over TPO together with TIO (orange line) and only TPO (blue line) from March 1958 to February 2022. The x-axis represents the midpoint year within the 31-yr moving window. Green points indicate insignificant values at the 90% confidence level, according to a Student's t test. (b) Change of λ from the TIO-inclusive to the only TPO one in (a). The black curve is the fitted polynomial curve. The red and blue dots indicate the positive and negative impact of TIO, respectively. Specially, stars indicate the maximum/minimum value.

DownLoad: Full-Size Img PowerPoint

By contrasting the opposite impacts of TIO SSTA, two distinct periods when positive and negative TIO SSTA contributions to asymmetry are maximized have been identified. Specifically, Period 1 spans from March 1960 to February 1991, while Period 2 spans from March 1982 to February 2013, representing the period of the mid-1970s and the late 1990s respectively. These selected periods serve as a basis for in-depth exploration into the dynamical structures associated with TIO-induced asymmetry (stars in Fig. 3b).

4.3   Case analysis

In the above subsection, it has been identified that there exist considerable discrepancies in TIO contribution to TA in two specific periods. To delve deeper into the underlying physical processes, this part assesses the disparity in the S-EOFs under the presence and absence of a positive phase of TIO in each period (Figs. 4 and 5). Here, the positive phase means the spatial pattern when the PC of S-EOF is larger than 0. As shown in Figs. 4 and 5, the positive phase of TIO SSTA tends to generate positive SSTA in eastern TPO, except in MAM of the developing year. Typically, SSTA responses associated with the S-EOFs are more pronounced in SON and DJF, especially in the transition phase of ENSO. The positive anomaly generated by TCDA consistently extends eastward in ENSO development, ultimately concentrating in the eastern TPO (Figs. 4a and 5a). During the ENSO transition, this concentration of TCDA in the eastern Pacific persists and strengthens again in DJF, enhancing the zonal contrast. These indicate that the SSTA of TIO enhances the ENSO amplitude during the developmental phase and continues to reinforce the mature phase amplitude, impeding ENSO decaying during the transition phase.

Figure  4.  The disparity of the S-EOFs under TIO-inclusive and TIO-exclusive (TIO-inclusive minus TIO-exclusive) over (a) the development and (b) transition phase in Period 1 (from March 1960 to February 1991), including SSTA (℃; shading) and TCDA (m; shading) response

DownLoad: Full-Size Img PowerPoint

Figure  5.  As in Fig. 4, but for Period 2 (from March 1982 to February 2013).

DownLoad: Full-Size Img PowerPoint

Furthermore, the signal response to TIO in Period 2 is stronger than that in Period 1, and shows westward displacement in SSTA. The above analysis indicates that the role of the TIO remains consistent in the two periods, both serving to inhibit the transition phase. However, such analysis fails to indicate which ENSO phase is the main cause of the result. Thus, the assessment of asymmetry requires distinct consideration of its impact on El Niño and La Niña, respectively.

To investigate the extent of impact on the asymmetric property, we compute the process parameters in Eq. 2. As mentioned above, the steeper (flatter) slope of the positive and negative events in scatter diagrams between PC1(t) and PC2 (t + 1) depicts the larger (smaller) transition efficiency (Fig. 6). For example, it is shown that S_P is much steeper than S_N in four scatter diagrams, implying a more efficient transition from El Niño to La Niña rather than the other way around. Table 1 illustrates that the impact on the transition efficiency of El Niño to La Niña in the response of TIO in Period 1 is nearly five times larger than that in Period 2, while for La Niña, the influence of the latter period is greater than that, of the former. Furthermore, by comparing the magnitudes of the TIO influence in different phases, it can be observed that TIO primarily affects TA by influencing the La Niña phase transition process in Period 2. In Period 1, however, ∆S_N is a relatively small negative value. Therefore, the decadal variabilities in TA are primarily attributed to the phase transition process of La Niña, with the influence of El Niño events being secondary.

Figure  6.  Scatter diagram between PC1(t) and PC2 (t + 1). The slopes for positive PC1 (S_P; red) and negative PC1 (S_N; blue) are obtained by the least squares fitted lines. (a) TIO-exclusive in Period 1, (b) TIO-inclusive in Period 1, (c) TIO-exclusive in Period 2, and (d) TIO-inclusive in Period 2. A single asterisk indicates statistical significance at the 90% confidence level, and two asterisks indicate statistical significance at the 95% confidence level.

DownLoad: Full-Size Img PowerPoint

Table  1.  TIO SSTA effects on the transition efficiency of El Niño and La Niña.

Period	∆SP	∆SN	∆λ
1960–1991	0.106	–0.095	0.112
1982–2013	0.022	0.324	–0.236

 | Show Table

DownLoad: CSV

4.4   Wind variability for TA

To reveal the dynamical structure of TIO effects on TA, Fig. 7 presents the anomalous zonal wind stress (τ_a) regressed against the PCs corresponding to S-EOFs of TIO. Due to the concentration of the influence on La Niña mentioned in the last section, the analysis is focused on the result derived from the regression in the negative phase of TIO. As illustrated in Figs. 7a and 7c, during the Period spanning the development to the mature phase of ENSO, the wind response on the western and central TPO in Period 1 is nearly insignificant. In contrast, during Period 2, there are two prominent high-value regions in SON and DJF, corresponding to the anomalous eastward winds in the northwest-southeast direction over the TIO and the anomalous westerlies in the central TPO for each. As discussed in Sections 2 and 3, there exists a stronger correlation between the IOD and ENSO than IOB in the late 1990s, leading to a stronger anomalous eastward wind response over TIO. Moreover, the TIO exerts remote control over the Pacific through its influence on the Walker Circulation and other atmospheric bridge process (Cai et al. 2019; Kim and Yu 2022). The decadal variations of TIO SSTA may induce changes in the circulation pattern due to the shift of convective coupling region related to SSTA, ultimately resulting in the superposition of westerlies anomalies caused by the TIO onto the existing easterlies anomalies associated with La Niña in the tropical central Pacific. This accelerates the decay of La Niña and expedites the phase transition. Additionally, during the transition in Period 2, the TIO again exhibits a westward wind anomaly in SON and DJF in the following year over the central TPO (Fig. 7d). This fosters conditions conducive to El Niño development, whereas such an impact is relatively subdued in DJF during Period 1 (Fig. 7b). This result may contribute to the prediction of multi-year La Niña events, since such decadal impact from TIO SSTA regulates the persistence of La Niña.

Figure  7.  Regression of τ_a (units: N m^–2; shading) onto (a) PC1 in Period 1, (b) PC2 in Period 1, (c) PC2 in Period 2, and (d) PC2 in Period 2. The black dots indicate regression values with statistical significance at the 95% confidence level.

DownLoad: Full-Size Img PowerPoint

5.   CONCLUSIONS AND DISCUSSION

The present study has provided a comprehensive examination of the TIO decadal variability and its consequential impact on TA of ENSO. The previous studies on TA are mostly focused on the intrinsic nature of ENSO itself and atmospheric nonlinearity. TIO plays a crucial role in influencing the complexity and diversity of ENSO as a neighboring basin. This study comprehensively examined the decadal variations of the TIO and its impact on TA of ENSO. Following an examination of the spatial and temporal distribution of SSTA in the TIO and the seasonal phase-locking characteristics of IOB and IOD, a 31-yr sliding correlation analysis based on the strongest seasonal correlations was employed to explore the decadal variations in the relationship between TIO SSTA and ENSO.

The S-EOF illustrates the seasonal evolution process of ENSO, in which the first mode corresponds to the ENSO development, and the second mode represents the ENSO transition. Quantitative analysis of TA could be achieved by examining the slopes of positive and negative phase transitions displayed in scatter plots based on PC1(t) and PC2(t+1). By comparing the efficiency of phase transitions for different sliding windows, it shows that TIO SSTA strengthened TA before the 1980s, while assuming opposite effects after the 1980s. After that two distinct periods with maximum and opposite influences of TIO SSTA on TA were identified to represent the period of the mid-1970s and the late 1990s. After comprehensively analyzing the characteristics of SSTA and TCDA, it is found that the TIO SSTA tends to promote the phase transition of El Niño. Moreover, the transition of La Niña is more likely to occur in the late 1990s, and the contribution of the TIO to TA is more significant during this time. This suggests that during the late 1990s, TIO SSTA tends to affect TA mainly by influencing the transition efficiency of La Niña. Furthermore, to better analyze the effects of the TIO SSTA during La Niña, the differences in τ_a responding to negative phase transitions of ENSO were investigated. The results revealed that the TIO SSTA in the later period induce anomalous westerlies in the central TPO, accelerating the termination of La Niña and promoting the development of El Niño in the following year. However, this could not be well captured in the mid-1970s. It can be seen that wind over TPO exhibits a bridge-like effect between TIO SST and statistical features of ENSO. Specifically, TIO modulates climatological background of wind on the decadal scale, thus providing an environment for the decadal variations of TA. The TA modulated by the decadal variability of TIO SSTA exhibits corresponding characteristics in different decades, being more likely attribute to the decadal variability of ENSO nature. The role of the TIO is regulatory, since its effect is more easily observed when ENSO is less active. This also explains why wind anomalies over the Pacific become more prominent during the ENSO decay phase.

Despite these significant insights, it is essential to acknowledge the inherent challenges and limitations of the study. The intricate nature of climate systems necessitates a holistic consideration of various contributing factors, isolating the singular influence of TIO may not capture the entirety of the complex interactions involved. Hence future research may address through more sophisticated modeling approaches. Additionally, as these results are based on linear correlations, the possibility of complex and nonlinear effects of TIO on TA cannot be ruled out. The nonlinear relationships and specific physical processes will be discussed in future studies.

APPENDIX

  A1.  TIO-exclusive from March 1960 to February 1991. (a) The first S-EOF mode (S-EOF1) of combined SSTA and TCDA. Time goes on descending order with a 3-monthly interval from boreal early spring to winter. (b) As in (a) except for the second S-EOF mode (S-EOF2). (c) Normalized corresponding PC time series at a yearly interval (PC1 is indicated by blue line; PC2 by red line).

DownLoad: Full-Size Img PowerPoint

  A2.  As in Fig. A1, but for the period from March 1982 to February 2013.

DownLoad: Full-Size Img PowerPoint

  A3.  As in Fig. A1, but for TIO-inclusive.

DownLoad: Full-Size Img PowerPoint

  A4.  As in Fig. A2, but for TIO-inclusive.

DownLoad: Full-Size Img PowerPoint