1 Introduction

Although the evolutions of the warm and cold phases of El Niño-Southern Oscillation (ENSO) are symmetric in many aspects (e.g., Philander 1990; An and Jin 2004; Sarachik and Cane 2010), they also display some prominent asymmetric features (Kumar and Hoerling 1997; Yu and Kim 2011; Anderson et al. 2013; Choi et al. 2013). For example, sea surface temperature (SST) anomalies (SSTAs) in the tropical Pacific associated with El Niño (warm) and La Niña (cold) years are asymmetric in terms of their meridional extent (Zhang et al. 2009), amplitude (Hoerling et al. 1997), and zonal phase propagation (McPhaden and Zhang 2009). On the other hand, the atmospheric responses associated with ENSO also demonstrate asymmetries (Hoerling et al. 1997), which can partially be attributed to the asymmetry of the climatological state, upon which the warm and cold events evolve. As summarized in Hoerling et al. (1997), because of the zonal asymmetries of the climatology with SSTs being high in the west and low in the east, even small deviations of SST from its climatological condition can generate large rainfall anomalies in the western tropical Pacific, whereas warm SST anomalies with remarkable amplitudes are needed to induce convections over the cold tongue in the tropical eastern Pacific. Furthermore, negative SSTAs have no significant impact on the already climatologically dry conditions in the cold tongue region. The asymmetric response of the atmospheric convections associated with El Niño and La Niña may result in the asymmetric response of the extra-tropical atmosphere to El Niño and La Niña.

One of the prominent features of the ENSO SST evolution is that the temporal evolution of El Niño and La Niña events is asymmetric. After their mature phase, for example, El Niño events generally decay by the next summer, but La Niña events often persist through the following year (Fig. 1c, d). Further, minor La Niña instead of an El Niño remerges after major La Niña, clear departure from the concept of an ENSO cycle (Kessler 2002; Hu et al. 2014). Current ENSO theories, such as the recharge/discharge oscillator (e.g., Wyrtki 1985; Jin 1997a, b; Meinen and McPhaden 2000; Clarke et al. 2007), the delayed oscillator (e.g., Suarez and Schopf 1988; Battisti and Hirst 1989), as well as others (e.g., Wang 2001), all imply a cyclic nature for ENSO whereby the climate over the tropical Pacific oscillates periodically between El Niño and La Niña states. The frequent occurrence of a second-year La Niña is a notable departure from the cyclic behavior.

Fig. 1
figure 1

Niño3.4 index composite for a El Niño and d La Niña years. Time and longitude evolutions of SST (contour), zonal component of surface wind stress (vector) and HC300 (shading) anomalies averaged between 5°S–5°N for composite of b El Niño and c La Niña years. The contour interval is 0.2 °C, and the unit is °C for HC300 and N/m2 for wind stress. The green (purple) vectors represent westward (eastward) wind stress anomalies. The white horizontal lines represent January of Year(1) and Year(2), respectively. The composites range from January of 1 year before the peak of ENSO [Year(0)] to July of second year after the peak [Year(2)]

Full size image

The causes of temporal asymmetric evolution have been investigated. Some of the proposed causes include: (a) nonlinearity of the atmospheric responses between the warm and cold events (e.g., Ohba and Ueda 2009); (b) the asymmetries in the meridional extent of the wind shift and its interaction with annual cycle (e.g. McGregor et al. 2013; Stuecker et al. 2013; Ren et al. 2016); (c) the asymmetric evolution in the oceanic response to wind forcing in delayed-oscillator conceptual model (e.g., Choi et al. 2013; DiNezio and Deser 2014); (d) asymmetry of reflected oceanic Rossby wave induced modification of ocean meridional gradient in association with meridional heat transport (Hu et al. 2014; Chen et al. 2016); and (e) contrast impact of remote oceans, such as the North Pacific (e.g., Alexander et al. 2010; Kim et al. 2012) and Indian (Kug et al. 2005; Okumura et al. 2011) Oceans.

Most of these previous works had demonstrated the importance of various nonlinearities in the asymmetry of the ENSO duration. The main difference between these studies is on the processes that produce the nonlinear behavior. For instance, Choi et al. (2013) emphasized the importance of the nonlinear relation between surface wind stress and SST anomalies. They noted that the wind stress anomalies depend more strongly on SSTAs during warm year than during cold year. DiNezio and Deser (2014) argued that different from El Niño, the thermocline doesn’t show proportionally delayed response to La Niña intensity, and this nonlinearity determines the asymmetric duration of warm and cold events. These previous explanations are based on the nonlinear delayed oscillator framework of ENSO. However, there exists another framework for ENSO evolution—namely, the recharge/discharge oscillator and the question is whether the temporal asymmetry could be explained within this paradigm. According to the recharge/discharge theory (Jin 1997a, b), the discharge of equatorial heat content occurs during an El Niño, with a net heat transport from the equatorial to the off-equatorial regions. The reverse happens during a La Niña event, with heat being transported from the off-equatorial regions into the equatorial ocean. Within this framework, subsequent studies further elucidated the roles played by some specific equatorial physical processes, such as the zonal advection and the zonal wind-thermocline-SST feedback (e.g., An and Jin 2001; Zhang et al. 2007; Bejarano and Jin 2008; Ren and Jin 2013).

In this work, through composite analysis using observational and reanalysis data, the temporal evolution of oceanic and atmospheric anomalies in the equatorial Pacific is examined to understand the symmetric and asymmetric features between El Niño and La Niña. The differences and similarities in the recharge and discharge processes in association with the warm and cold events are illustrated. Furthermore, surface ocean zonal and meridional geostrophic currents and their association with meridional gradient of SSH across the equator are examined. The remainder of this paper is organized as follows. Section 2 descripts the data used in this work. The symmetric and asymmetric characteristics of El Niño and La Niña evolutions are displayed in Sect. 3. In particular, we examine the contrast of recharge and discharge processes connected with El Niño and La Niña and the role of the meridional gradient of SSHAs across the equator in the ENSO asymmetric evolution. Section 4 summaries the results with some discussion.

2 Data

Monthly means of sea surface height (SSH), the 20 °C isotherm depth (D20), ocean heat content between the surface and the 300 m depth (HC300), and surface wind stress on a 1° × 1° grid are from the Global Ocean Data Assimilation System (GODAS) (Behringer and Xue 2004). In addition, we utilize monthly mean SST of the version 2 of the optimum interpolation (OIv2) on a 1° × 1° grid (Reynolds et al. 2002). All data used in this work cover the period January 1982-December 2015. Due to the fact that OIv2 does not have data before November, 1981, here the anomalies are referred to monthly climatologies in 1982–2011.

Two indices are used in the analysis. First one is the warm water volume (WWV) index which is defined as anomalous D20 depth averaged in (5°N–5°S, 120°E–80°W) (Meinen and McPhaden 2000). According to the recharge and discharge theory (see Eq. (4.2) of Jin 1997a), the WWV tendency is an indicator for the heat recharging and discharging strengths in the whole equatorial Pacific, which is negatively correlated with SSTA (Jin 1997a, b), and is associated with the phase turnaround of ENSO (Meinen and McPhaden 2000; Clarke et al. 2007; Kumar and Hu 2014). Another one is the Niño3.4 index which is defined as SSTA averaged in (5°S–5°N, 170°W–120°W) to measure the evolution of ENSO (Barnston et al. 1997). The ENSO year definition follows that of the NOAA Climate Prediction Center (CPC) based on 3-month running mean Niño3.4 index (see http://www.cpc.ncep.noaa.gov/). There are total 10 El Niño or 9 La Niña events during 1982–2015 (Table 1). Here, second year El Niño or La Niña is treated as a new event (see Table 1). Due to the fact that OIv2 SST starts from November 1981, we focus on ENSO events during 1982–2015. The composites start from January [January (0)] of the year of El Niño and La Niña development [Year(0)], and end by July [July(2)] of the year [Year(2)] after their peak year. The Year(1) is referred to as the year of ENSO from mature to decay.

Table 1 El Niño and La Niña events selected in this work according to the definition of NOAA CPC based on 3-month running mean Niño3.4 index (see http://www.cpc.ncep.noaa.gov/)
Full size table

3 Symmetric and asymmetric evolution between El Niño and La Niña

3.1 SST, surface wind stress, and ocean heat content along the equator

From the Niño3.4 index composites (Fig. 1a, d), the asymmetric features between El Niño and La Niña evolutions are evident. The composite shows that, on average, both El Niño and La Niña years are followed by negative Niño3.4 index, indicating a non-cyclic nature of the ENSO evolution. To further compare the similarities and differences in the temporal evolutions of the warm and cold years, Fig. 1b, c show the corresponding composites of HC300, SST, and surface zonal wind stress anomalies on the equator (averaged between 5°S–5°N). For the HC300 anomaly composites (shading), it is noted that the anomaly starts to develop in the western Pacific in Year(0) in both the events, then propagates eastward. Later, through the Bjerknes air–sea interaction (Bjerknes 1969), the anomalies are further amplified, and peak around the winter of Year(0)–Year(1) in the central and eastern equatorial Pacific.

During the development phase, the anomalous patterns are quite symmetric between the warm and cold year composites. After their peaks, however, the asymmetric features are seen. For example, for the El Niño composite (Fig. 1b), in the early summer of Year(0), negative HC300 anomaly is observed in the western, while the positive anomaly in the central and eastern Pacific Ocean. For the negative HC300 anomalies to the west of the dateline, they are nearly stationary from the summer of Year(0) to the spring of Year(1). In the 2nd half of Year(1), they propagate eastward into the central Pacific, which is associated with the SST cooling in Year(1) as shown Fig. 1a. However, for the corresponding La Niña composite (Fig. 1c), throughout Year(0)–Year(1), the positive HC300 anomalies in the western equatorial ocean do not have corresponding eastward propagation. Instead, the negative anomalies of HC300 emerge locally in the central and eastern Pacific in the 2nd half of Year(1), suggesting that asymmetric features between El Niño and La Niña become evident in the 2nd half of Year(1) and in Year(2).

The composite pattern of zonal surface wind stress anomalies (vectors in Fig. 1b, c) is physically consistent with the corresponding anomalous composites of SST and HC300. For example, anomalous westerly (easterly) wind stress is associated with reduction (enhancement) of anomalous zonal gradients of SST and HC300 (Bjerknes 1969; Wyrtki 1985; Philander 1990; Sarachik and Cane 2010). Corresponding to the evolutions in SSTAs, westerly anomalies switch to easterly anomalies over the western and central equatorial ocean in late winter and spring for the El Niño composite, while the easterly anomalies persist for 2 years with no sign changes for the La Niña composite, consistent with the findings of Ohba and Ueda (2009).

Overall, the composite evolution analyses suggest that the anomalies on the equator are basically symmetric in the El Niño and La Niña composites prior to and in their mature phase. The temporal evolution becomes asymmetric after the mature phase. Since the recharge and discharge of water mass on and off the equator play a critical role in the ENSO evolutions and its phase turnaround, the asymmetry in the evolutions between El Niño and La Niña can be associated with differences in the recharge and discharge processes, which are addressed below.

3.2 Symmetry and asymmetry of recharge and discharge processes

The symmetric and asymmetric characteristics between the El Niño and La Niña composites are also reflected in the recharge-discharge process. The ocean heat accumulation or disappearance is an indicator for the ENSO phase turnaround (Jin 1997a, b). In essence, the recharge-discharge theory argues that the convergence (divergence) of meridional geostrophic current recharges (discharges) the WWV in the equatorial Pacific during La Niña (El Niño), leading to a zonally integrated deepening (shoaling) of equatorial thermocline, meaning recharging (discharging) mass and heat into (from) the upper equatorial Pacific. After the equatorial Pacific is recharged (or discharged), some feedbacks (such as the thermocline feedback and the zonal advective feedback) tend to work constructively (Suarez and Schopf 1988; Battisti and Hirst 1989; Picaut et al. 1997; Jin and An 1999; Ren and Jin 2013), leading to a phase transition of ENSO.

Here, the tendency of the WWV index is used to represent the recharge and discharge processes. The composites show that the overall WWV tendency (shading in Fig. 2a, b) is out of phase with Niño3.4 index (solid curve in Fig. 2a, b), consistent with the recharge and discharge theory (Jin 1997a, b): El Niño (La Niña) is overall associated with the negative (positive) WWV tendency due to discharge (recharge) processes, although there are outliers (negative (positive) WWV tendency in some months of La Niña (El Niño) years, see Fig. 3a). Despite the overall synchronization between the WWV tendency and Niño3.4 index, some differences are noted for the relationship between the El Niño and La Niña composites (Fig. 2c). For example, the overall discharge intensity during El Niño [October(0)–April(1)] is stronger than its recharge counterpart during La Niña (Figs. 2, 3). As a result, the sum of El Niño and La Niña composites (Fig. 2c), which reflects the asymmetric evolution, is mainly discharge during November (0)–October(1) and recharge during November(1)–June(2). The latter probably reflects the fact that the La Niña events linger longer. Specifically, the maximum discharge for individual El Niño event reaches about −8 m per month for WWV tendency, while the maximum recharge for individual La Niña event is only about 5 m per month for WWV tendency (Fig. 3a).

Fig. 2
figure 2

Time evolutions of the WWV index anomalous tendency (shading), and Niño3.4 index (solid curve) for composite of a El Niño and b La Niña years, and c El Niño + La Niña years ((a)+(b)). The WWV index is defined as the D20 anomaly averaged in (5°S–5°N, 120°E–80°W). The unit is m/month for the WWV anomalous tendency and °C for Niño3.4 index

Full size image
Fig. 3
figure 3

Scatter plots of monthly mean anomalies of a Niño3.4 index (x-axis) and WWV tendency index (y-axis) and b Niño3.4 index (x-axis) and zonal wind stress (Taux) averaged in (5°S–5°N, 150°E–150°W) (y-axis) for positive Niño3.4 index (red squares) and negative Niño3.4 index (green squares). The regression coefficient of WWV anomalous tendency (zonal wind stress) onto Niño3.4 index is −1.3 m/(month °C) (0.012 N/m2) for positive Niño3.4 index and −1.1 m/(month °C) (0.010 N/m2) for negative Niño3.4 index. The correlation is −0.43 (−0.30) between Niño3.4 index and WWV tendency index, and 0.70 (0.47) between Niño3.4 index and zonal wind stress for positive (negative) Niño3.4 index

Full size image

Interestingly, the connection of the recharge and discharge processes with Niño3.4 SSTA, which can be approximated by linear regression coefficient of WWV tendency onto Niño3.4 SSTA, seems to also have some differences (Fig. 3a). Specifically, the regression coefficient of WWV tendency onto Niño3.4 SSTA is −1.3 m/(month °C) for the positive Niño3.4 months and it is −1.1 m/(month °C) for the negative Niño3.4 months (Fig. 3a). Consistently, the correlation between WWV tendency and Niño3.4 SSTA is higher for positive Niño3.4 months (−0.43) than for negative Niño3.4 months (−0.30) (Fig. 3a).

Furthermore, the relationship between Niño3.4 index and zonal wind stress averaged in (5°S–5°N, 150°E–150°W; it is so-called Niño4 region) also has some differences between positive and negative Niño3.4 index (Fig. 3b). For instance, the regression coefficient of zonal wind stress onto Niño3.4 index is 0.012 N/(m2 °C) for positive Niño3.4 index and 0.010 N/(m2 °C) for negative Niño3.4 index. Their corresponding correlation is 0.70 (0.47) for positive (negative) Niño3.4 index (Fig. 3b). The relationship between Niño3.4 SSTA and zonal wind stress in Niño4 region mainly reflects the intensity of the so-called Bjerknes feedback (Bjerknes 1969; Lloyd et al. 2009). The connections among Niño4 SST, WWV and zonal wind stress anomalies shown in Fig. 3 are the feature of air-sea coupling in the equatorial Pacific.

While, the amplitude differences of the numbers shown in Fig. 3a, b suggest that statistically zonal wind stress anomaly is more closely connected with Niño3.4 than WWV tendency does. On the other hand, the different regression and correlation coefficient values between positive and negative Niño3.4 index (Fig. 3) suggest that on average, air-coupling is stronger in El Niño years than in La Niña years. That is generally consistent with some previous works. For example, both McGregor et al. (2013) and DiNezio and Deser (2014) noted that the discharge of oceanic heat content increase as an El Niño becomes intensified in its amplitude, but the recharge is insensitive to the amplitude of La Niña. However, if from the view of the linear regression and correlation, the differences seem not as significant as these previous works indicated. Overall, these results suggest that, consistent with the fact that, on average, the SSTA amplitude in El Niño is larger than that in La Niña, and the anomalous discharge during El Niño is stronger than the anomalous recharge during La Niña. Also, the linear relationship between discharge/recharge and Niño3.4 SSTA has some differences between warm and cold events.

3.3 Connection with zonal and meridional currents and role of off-equatorial SSHA

According to Jin and An (1999), recharge/discharge process is associated with the meridional (Vg) convergence and zonal gradient of thermocline depth on the equator that is tied up with the meridional concavity of the thermocline anomalies and zonal (Ug) component of the geostrophic current anomalies of upper ocean. Here, geostrophic current (Ug and Vg) is computed at latitudes of 4.5°S, 3.5°S, 2.5°S, 1.5°S, 0.5°S, 0.5°N, 1.5°N, 2.5°N, 3.5°N, and 4.5°N, then do the average to get the mean between 5°S and 5°N. For the calculation of the meridional geostrophic current, SSHA averaged in 156°E–140°W is used. According to Clarke et al. (2007; see their Figs. 3 and 7), the ocean heat recharge/discharge process takes place mainly in the western and central equatorial Pacific (156°E–140°W). In the other words, the SSHA in this longitude zonal is largely associated with ocean heat exchange between the equator and the off-equator, while the SSTA in the region east of 140°E may be due to the zonal propagation of the ocean heat along the equator.

The composites shown in Figs. 4, 5, and 6 are consistent with the schematic evolution of Jin and An (1999; see their Fig. 2). Specifically, both pronounced anomalous eastward zonal current and anomalous divergence of meridional current on the equator and positive SSHA in the central and eastern tropical Pacific are present during the El Niño phase (Figs. 4a, 5a, 6a; July(0)-January(1)). Meanwhile, both notable anomalous westward zonal current and anomalous convergence of meridional current on the equator and negative SSHA in the central and eastern tropical Pacific are present during the La Niña phase (Figs. 4b, 5b, 6b; July(0)-January(1)). During the transition phase [both from the warm to the cold phase and from the cold to the warm phase during April (1)-July(1)], the zonal current and SSHAs reverse their sign, and anomalous meridional current on the equator becomes smaller.

Fig. 4
figure 4

Time and longitude evolutions of zonal geostrophic current anomaly at the ocean surface (vector) and SSHA (shading) averaged in 5°S–5°N for composites of a El Niño and b La Niña years. The green (golden) vectors are for westward (eastward) current. The white lines represent January(1) and January(2), respectively

Full size image
Fig. 5
figure 5

Time and latitude evolutions of meridional geostrophic current anomaly at the ocean surface (vector) and SSHA (shading) averaged in 156°E–140°W for composites of a El Niño and b La Niña years. The green (golden) vectors are for southward (northward) current. The white lines represent the equator, January(1), and January(2), respectively

Full size image

The equatorial surface zonal current reversal in January(1) originates from the reversal of meridional gradient of SSHA. In general, the zonally averaged SSHA and the meridional gradient are largely symmetric with respect to the equator before July(1). Using El Niño composite as an example (Figs. 4a, 5a), SSHA has its maximum on the equator during Year(0). After January(1), because the equatorial SSHA becomes the minimum in meridional direction due to the decrease of SSHA on the equator and the increase of SSHAs on both sides of the equator. The decrease on the equator and increase in the off-equatorial region in SSHA may be associated with the discharge due to poleward heat transports along the equator and the reflected equatorial Rossby waves along the off-equator at the eastern boundary (Hu et al. 2014). These changes reverse the concavity of the SSHA meridional profile and thus lead to the reversal in surface zonal current through geostrophic balance (Jin and An 1999; Chen et al. 2016). Therefore, the establishment of the off-equatorial SSHAs and the weakening of the equatorial SSHA may be a cause of the surface zonal current reversal and its subsequent sustenance from January(1) to July(1) (Figs. 4a, 5a). The La Niña event (Figs. 4b, 5b) is largely a mirror image of the El Niño process described above and the two are symmetric in their evolution before July(1).

After July(1), when the off-equatorial SSHAs move outside the equatorial wave-guide, the asymmetry of the evolutions between El Niño and La Niña emerges. In particular, the direction of the surface zonal current reverses again in the cold event but sustained in the warm one (Fig. 4). The differences in surface zonal current evolutions of the warm and cold events can be explained by the changes in the meridional gradient of SSHA and the different rates of the SSHA changes between the equatorial and the off-equatorial Pacific. For the warm event (Figs. 4a, 5a), the equatorial SSHA quickly switches from positive to negative around January(1)–April(1), meanwhile the off-equator is occupied by positive SSHAs in the north and by relatively smaller negative SSHAs in the south. Such changes on the equator and in the off-equator after January(1) make the concavity of the meridional SSHA profile favorable for growth and continuation of westward surface zonal current anomaly in the equatorial Pacific. For the cold event (Figs. 4b, 5b), the decay of the cold equatorial anomaly from January(1) to July(1) is much slower and the warm anomalies do not get well established by July(1). The overall weak recharge process seems associated with the slower demise of the cold SSTA during La Niña. After the off-equatorial negative SSHA disappears, the concavity is easily reversed and the westward surface zonal current anomaly, once again, starts to enhance the cold signals on the equator. Therefore, the relatively different rates of the equatorial and off-equatorial SSHA changes from January(1) to July(1) are a key feature of the ENSO asymmetry during Year(1).

The evolutions of the zonal and meridional current as well as the SSH anomalous composites (Figs. 4, 5) are consistent with the symmetric and asymmetric features of discharge/recharge processes associated with the ENSO evolution. These anomalies are symmetric for El Niño and La Niña prior to and in their mature phase, and become asymmetric afterwards. Post the peak phases of ENSO [January(1)–January(2)], anomalous westward zonal current and anomalous convergence of meridional current are pronounced and persistent for the El Niño composite (Figs. 4a, 5a, 6a), while for the La Niña composite, the notable anomalous eastward zonal current is disrupted by weak westward anomaly during August(1)–December(1) (Fig. 4b) and the anomalous convergence of meridional current persists (Figs. 5b, 6b). The lack of persistence of an anomalous eastward current and thus the lack of eastward advection of warm ocean temperature prevent the eastward surge of the warm water. This is consistent with the fact that instead of evolving to an El Niño, a La Niña often has “double-dip”, meaning persistent into second year or reemergence of a second year La Niña (e.g., Kessler 2002; Hu et al. 2014).

Fig. 6
figure 6

Same as Fig. 4, but for divergence and convergence of meridional geostrophic current anomaly at the ocean surface, which is defined as the difference of mean in 0°–5°N minus mean in 5°S–0°

Full size image

4 Summary and discussion

In this work, the evolutions of oceanic and atmospheric anomalies in the tropical Pacific are analyzed to show the symmetric and asymmetric features between El Niño and La Niña through composite analysis using observation-based reanalysis data. The differences and similarities in the recharge and discharge processes associated with El Niño and La Niña evolution are examined, and the possible influence of the off-equatorial sea surface height is investigated.

It is noted that the evolutions of ocean and atmospheric anomalies along the equator in the El Niño and La Niña composites are symmetric before and during their mature phase. However, the evolutions become asymmetric afterwards. It is also noted that, on average, the discharge intensity during El Niño is stronger than its recharge counterpart associated with La Niña. The symmetric and asymmetric evolution for different phases of El Niño-Southern Oscillation (ENSO) is consistent with the associated recharge and discharge processes that are linked with the evolution of anomalous ocean surface current and gradient of sea surface height anomalies (SSHAs). It is suggested that the evolution of the meridional gradient of SSHAs in the equatorial Pacific is different between El Niño and La Niña years due to different rates of the relative SSHA changes on and off the equator. During the decaying phase of a warm event, the equatorial SSHA quickly switches from positive to negative, meanwhile the off-equatorial regions are occupied by positive SSHAs. Such changes on the equator and in the off-equator make the concavity of the meridional SSHA profile favorable for the westward surface zonal current anomaly in the equatorial Pacific and for the ENSO phase transition. During the decaying phase of a cold event, the slow decrease of the negative SSHA on the equator combined with the disappearance of the negative SSHA in the off-equator make the meridional concavity easily reverse and is unfavorable for the steady growth of the eastward surface zonal current and for the phase transition. The persistently weaker recharge process seems to be the major reason for the slower demise of the cold SSTA during La Niña and its revival around July(1). Thus, the asymmetric evolution of ENSO is associated with both the atmosphere and ocean anomalies on the equator and also in the off-equatorial regions.

What causes the asymmetry between recharge and discharge processes associated with ENSO after their peak phases? We speculate that the answer may lie in the specific water properties transported from the off-equatorial to the equatorial ocean. For example, Fig. 5b shows that, from April(1) to July(1) after La Niña peak, the convergence of Vg brings negative SSHA from the off-equatorial oceans of both hemispheres into the equatorial ocean, which may cause the termination of the recharge process and cool the equator. That terminates the transition from La Niña to El Niño and leads to asymmetric evolution between El Niño and La Niña. Hu et al. (2014) argued that the cold ocean temperature anomaly or the negative SSHA over the off-equatorial region may be associated with the Rossby wave signal reflected at the American coast of the Pacific.

Our current work proposes the possible role of the meridional gradient of thermocline and the off-equatorial SSHA resulting in the asymmetric evolution of recharge and discharge of ocean and ENSO. In fact, the asymmetric evolution of ENSO may also be partially due to other nonlinear interactions. Jin et al. (2003), Stuecker et al. (2013), and Ren et al. (2016) argued that the interaction between the ENSO time scale variation and climatological annual cycle can generate near-annual time scale variation, which may turn out to affect the recharge and discharge processes associated with ENSO and then the evolution of ENSO. Furthermore, as discussed in the Sect. 1, the crucial role of climatological state of the tropical Pacific atmosphere played in the asymmetric evolution of ENSO had been proposed (Hoerling et al. 1997; Ohba and Ueda 2009).

Last, in this work, WWV tendency is used as a proxy to measure the charge/discharge processes. WWV includes all processes that alter heat content, which are dominated by changes in thermocline depth on interannual timescales. Hence the WWV can be seen as a surrogate for thermocline depth. This relationship might not necessarily work on seasonal timescales, because horizontal advection is more dominant. In addition, the results in this work are based on composite of ENSO events since 1982 and largely rely on products of one ocean data assimilation system (GODAS). The impact of biases in GODAS, particularly artificial ocean internal sources and sinks of heat introduced during assimilation process (Huang et al. 2010; Hu et al. 2016) should be identified and estimated through comparison by analyzing multiple oceanic reanalyzed datasets. Furthermore, it should be pointed out that the robustness and possible interdecadal dependence of the results should be verified by using independent and longer observational or modeling data (Zhang et al. 1999; Hu et al. 2017).