Impacts of Different Types of ENSO Events on Thermocline Variability in the Southern Tropical Indian Ocean

Remarkable interannual variability in the thermocline depth in the southern tropical Indian Ocean (STIO) is analyzed using reanalysis data during 1980–2017. Previous studies have shown that the El Niño‐Southern Oscillation (ENSO) has a significant relationship with thermocline depth anomalies in this region. We find that both the eastern‐Pacific (EP) and central‐Pacific (CP) ENSO have important impacts on the STIO thermocline variation. The positive and negative phases of thermocline anomalies in the STIO are induced by asymmetric forcings from the two phases of ENSO. EP‐El Niño and CP‐La Niña events tend to induce larger thermocline depth anomalies in the STIO. Equatorial westerly and STIO anticyclonic winds during EP‐El Niño events can induce downwelling Rossby waves that extend far toward the western Indian Ocean, which dominates the westward propagation of thermocline anomalies, while upwelling Rossby waves during CP‐La Niña events cannot extend that far west.


Introduction
The tropical Indian Ocean has received intensive attention due to its significant climate impacts at regional and global scales, its low latitude connection with the tropical Pacific Ocean through the Indonesian throughflow (ITF), its basin-wide coupled air-sea mode the Indian Ocean Dipole (IOD; Saji et al., 1999;Webster et al., 1999), and the dynamic response to the remote forcing by El Niño-South Oscillation (ENSO). Unlike Pacific and Atlantic, Indian Ocean is the only ocean with the climatological equatorial westerlies (e.g., Webster et al., 1999;Xie et al., 2002Xie et al., , 2016, and the depth of the thermocline is used to identify the dynamic response of the tropical Indian Ocean to wind forcing (Schott et al., 2009). Compared to the amplitude in the tropical Pacific, the maximum variability in the 20°C isothermal depth (Z20 hereafter) in the tropical Indian Ocean occurs in the southern tropical Indian Ocean (STIO), indicating strong subsurface variations (Figure 1a). Vertical displacements of the thermocline depth affect sea surface temperature (SST) variability via vertical water exchange  and thereby modulate the atmospheric environment (Xie et al., 2002) and local air-sea interaction events (Jin & An, 1999). ©2019. The Authors. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. However, subsurface variability in the tropical Indian Ocean lacks sufficient understanding (e.g., Feng et al., 2001;Huang & Kinter, 2002;Jin et al., 2018;Keerthi et al., 2013). At the interannual time scale, several studies checked the thermocline variability related to the IOD (Liu et al., 2011(Liu et al., , 2014(Liu et al., , 2017Rao & Behera, 2005;Vinayachandran et al., 2002;Zhao et al., 2016), ITF (Annamalai et al., 2005;Du & Qu, 2010;Feng et al., 2018;Gordon et al., 2003;Hu et al., 2015;Lukas et al., 1996;Meyers, 1996;Meyers et al., 2007;Qiu et al., 1999;Sprintall et al., 2003;Wijffels et al., 2008;Zhang et al., 2018), or ENSO (Chambers et al., 1999;Shinoda et al., 2004;Tourre & White, 1995;Yu et al., 2002). Rao et al. (2002) and Xie et al. (2002) considered that ENSO-forced off-equatorial Rossby waves were prominent only in the southern Indian Ocean but not in  (Ren & Jin, 2011). The positive (negative) months indicate that the Z20A box lead (lag) the EP/CP index. The shadings and vectors in (b) and (c) and the gray line in (d) denote significance levels exceeding 95%. the northern part. Moreover, during El Niño events, anomalous easterlies in the eastern tropical Indian Ocean could raise the local thermocline, resulting in the enhancement of thermocline feedback on SST (Xie et al., 2002). Rao and Behera (2005) and Yu et al. (2005) noted that thermocline variations south of 10°S were dominantly induced by ENSO, whereas thermocline variations north of 10°S were more associated with the IOD.

10.1029/2019GL082818
During past decades, ENSO events have been divided into two different types: the central-Pacific (CP) and the eastern-Pacific (EP) El Niño/La Niña, with maximum positive/negative SST anomalies (SSTA) located in the central Pacific and the eastern equatorial Pacific, respectively (Ashok et al., 2007;Kao & Yu, 2009;Kug et al., 2009;Lee & McPhaden, 2010;Pan et al., 2018;Ren & Jin, 2011;Yu et al., 2011;Zhang, Wang, Xiang, et al., 2015, Zhang, Wang, Jin, et al., 2015. These two types of ENSO events are different in terms of their formation mechanisms and teleconnection signals (Ashok & Yamagata, 2009; and may induce different impacts on Z20 variability in the Indian Ocean, especially in the STIO region. In addition, the asymmetry between the polarities is an intrinsic nonlinear characteristic of ENSO. The amplitudes of EP-El Niño events, on average, tend to be stronger than those of La Niña events (Burgers & Stephenson, 1999;Hannachi et al., 2003;Kessler, 2002), whereas the asymmetry of CP ENSO events is opposite . Whether ENSO asymmetry impacts Z20 anomalies (Z20A) in the STIO has not been well discussed yet. The present study aims to answer this question by analyzing observation and reanalysis data.

Data
Monthly SST, subsurface temperature, and sea surface wind data from January 1980 to December 2017 are analyzed in this paper. Z20 is estimated using the National Centers for Environmental Prediction Global Ocean Data Assimilation System (Behringer & Xue, 2004). The SST data are acquired from the monthly National Oceanic and Atmospheric Administration Extended Reconstructed Sea Surface Temperature V5 data set (Huang et al., 2017). To explore the relationship between wind forcing/wind stress curl and Z20 changes, monthly 2.5°National Centers for Environmental Prediction version-1 reanalysis surface wind data (Kalnay et al., 1996) are also used. Anomaly variables are calculated by removing monthly climatologies and then smoothing with a 13-month low-pass digital Butterworth filter. Moreover, the linear trends of all data sets are removed before the analysis.

Methods
There are many commonly used ENSO indices for two types of ENSO (e.g., Ashok et al., 2007;Kao & Yu, 2009;Kug et al., 2009;Ren & Jin, 2011). Here the indices of Ren and Jin (2011) are used in correlation analysis because the EP and CP ENSO based on their indices can be better separated due to their low correlation. In addition, these previously mentioned indices could identify two types of El Niño but fail to distinguish different La Niña types (Zhang, Wang, Xiang, et al., 2015), including the indices of Ren and Jin (2011 1982, 1986, 1991, 1997, and 2015 (1994, 2002, 2004, and 2009), and the EP (CP) La Niña years are 1984, 1985, 1995, 1999, 2005, and 2007 (1983, 1988, 1998, 2000, 2008, 2010, and 2011). Please note that our qualitative conclusions remain the same if we use the other EP/CP index mentioned above.

STIO Thermocline Variation Associated With ENSO Events
In the tropical Indian Ocean, the thermocline depth presented large variability in the southern region, with the maximum in the region 70-85°E, 8-17°S (the black box in Figure 1a). The correlation pattern of the thermocline anomalies ( Figure 1b) and SSTA ( Figure 1c) associated with the box-averaged Z20A (Z20A box ) displayed a dipole mode in the tropical Indian Ocean, a negative center in the Sumatra-Java coastal region and a positive center over the south-central STIO. The correlation pattern is similar to the seesaw oscillation observed between the eastern and western Z20A identified by Masumoto and Meyers (1998), Murtugudde and Busalacchi (1999), and Shinoda et al. (2004) and the first mode of the empirical orthogonal function of sea level anomalies in Rao et al. (2002). Note that the tropical dipole pattern in Figure 1c is not the same as the subtropical dipole (SDI in supporting information Figure S1) presented by Behera and Yamagata (2001).
The thermocline in the tropical Indian Ocean is normally influenced by the ITF and local wind forcing. Via the ITF, coastal Kelvin waves from the Pacific could enter the Indian Ocean and propagate as westward Rossby waves. Through this process, the remote forcing effect from the Pacific is involved. However, previous studies have argued that the ITF has little impact on the interior STIO and that local wind forcing plays an important role in Z20 variation in the STIO (e.g., Xie et al., 2002). As shown in Figure S2, signals originating from the eastern Indian Ocean and propagating westward toward the STIO are rather weak, except for the 1997/1998 El Niño event. Compared with forcing from the ITF, local wind forcing plays a key role. Strong equatorial easterlies and the off-equatorial anticyclone in the tropical Indian Ocean can deepen the thermocline in the STIO (Figure 1c). Furthermore, the forcing of local wind on the tropical Indian Ocean is influenced by ENSO (Alexander et al., 2002). Through atmospheric teleconnections, ENSO modulates surface wind variations over the tropical Indian Ocean and thereby evokes Z20 variability through baroclinic Rossby waves and Ekman pumping. It is interesting to notice that the spatial pattern in the Pacific Ocean associated with the Z20A box is very similar to the CP ENSO (Figure 1c and CP in Figure  S1). Although the lead-lag correlations of Z20A box with the EP and CP index are both remarkable when the CP (EP) index leads up to 18 (7) months (Figure 1d), the CP index has a higher correlation than the EP one. Thus, spatial pattern in the tropical Pacific Ocean in Figure 1c is dominant by the CP ENSO, which indicates that the CP ENSO may have a greater impact on Z20 variation in the STIO.
Thermocline of Z20A box is seasonally phase-locked ( Figure S3), with a minimum depth existing from summer to early fall (June-September) and peaking in winter (DJF). Furthermore, the seasonal evolutions of oceanic and atmospheric parameters associated with Z20A box in DJF are shown in Figure 2. In the tropical Pacific Ocean, the evolution of SSTA changes from a La Niña pattern into an El Niño pattern. During January-March, the negative correlation with SSTA is significantly high in the central-eastern Pacific, which indicates the mature phase of a La Niña event. Then, the La Niña pattern gradually weakens and disappears in April-May due to downwelling Kelvin waves induced by westerly wind anomalies in the western Pacific. During the monsoon transition period (April-May; Susanto et al., 2001), southeasterly anomalies emerge in May, and the dipole mode of the Z20A appears and strengthens in the STIO. This result is coincident with that of Xie et al. (2002;their Figures 9b and 11), indicating significant correlations of the wind and Z20 variations in the STIO with ENSO.
The following summer is the transition period. Positive SSTA first appear in the central tropical Pacific in June and propagate eastward during July-August, which is in agreement with the CP-El Niño. In June-August, with the strengthening of anomalous anticyclonic winds in the STIO, a dipole mode of the thermocline emerges, with a negative Z20A near the Sumatra-Java coast and a positive Z20A in the STIO. Positive SSTA in the central STIO appear starting in August.
During September-October, westerly wind anomalies propagate toward the central Pacific and strengthen, which induces positive Z20A and SSTA that propagate eastward toward the eastern Pacific. However, the positive SSTA center still remains in the central Pacific until October. During November-December, the SSTA center propagates toward the eastern Pacific. But this pattern is not fully consistent with the mature phase of the EP-El Niño because the maximum SSTA do not emerge from the coast of South America. It indicates that the impact of ENSO is a combination of the EP and CP signals, and the latter one is stronger. This characteristic is accompanied by the enhancement of the dipole mode of the anomalous thermocline and SST in the STIO due to the strengthening of the anomalous anticyclone wind near the Java coast.
The distinct propagation of the positive centers of Z20A induced by wind curl in the STIO is another important feature. Under the background of the anticyclonic wind formed in July in the STIO, positive Z20A enhance and extend from the central into the western STIO since September, which indicates the response of Rossby waves to wind forcing (e.g., Chakravorty et al., 2014;Xie et al., 2002). Meanwhile, due to the westward propagation of the Rossby wave, the positive SSTA also show westward and northward propagations, but the SSTA center is still located over the STIO. In general, the formation and evolution of the dipole mode of the thermocline in the STIO are closely related to ENSO events through wind forcing.

Influence From Positive and Negative Phases of EP and CP ENSO on STIO
The correlation analysis applied in Figure 2 assumes that the positive and negative phases of the Z20A in the Indian Ocean have the same spatial patterns and temporal evolutions. We conduct composite analyses for further examination of the common and different features between the two phases of the Z20A. The years in the composition are based on the normalized Z20A box with the absolute value ≥1 standard deviation. Figure 3 shows that the positive and negative events look very much alike in the tropical Indian Ocean except with opposite signs, indicating that the positive and negative phases is symmetrical in the Indian Ocean. However, the patterns in the tropical Pacific Ocean are different during different phases. The positive events show the EP-El Niño pattern, but the negative events display the CP-La Niña pattern. It is suggested that pattern of the positive/negative phase of Z20A in the STIO is influenced by the asymmetry of the forcing from different phases of CP/EP ENSO, due to strong intensity for EP-El Niño and CP-La Niña and weak intensity for CP-El Niño and EP-La Niña. It is interesting to notice that the westward propagation of the Z20 signal only appears during positive phase and is associated with EP-El Niño.
Furthermore, the composite during strong EP/CP ENSO years ( Figure 4) is analyzed. The following results are similar if using other EP/CP indices ( Figure S4), including Kao and Yu (2009), Kug et al. (2009), and Ren and Jin (2011). For the EP ENSO, the magnitude of the Z20A in the tropical Pacific Ocean is greater during El Niño events than La Niña events. The Z20A are suggested to deepen in the STIO and display a westward propagation structure during EP-El Niño events. However, the situation is exactly opposite for the CP ENSO. The dipole mode of Z20A is stronger during CP-La Niña events, which induces larger amplitudes of Z20A than the CP-El Niño events. Different from EP-La Niña events, the centers of the negative Z20A are always located in the central STIO without an evident westward propagation tendency during CP-La Niña events. This result supports the previous conclusion in Figure 3. Forcings from the positive and negative phases of ENSO are asymmetric. EP-El Niño events could induce an obvious deepening of the thermocline in the STIO and its westward signal propagation, whereas CP-La Niña events could result in strong shoaling of the thermocline. The asymmetry of the forcing during different phases of CP/EP ENSO can also be found in Figure S5. The scatter plot demonstrates that all the Z20A box with amplitude less than or equal to −10 m (≥12 m) are accompanied with CP-La Niña (EP-El Niño) events.
The roles of equatorial and off-equatorial wind anomalies are clearly shown in Figure 5. During CP-La Niña, the equatorial westerly and off-equatorial cyclonic winds south of the equator are weak and do not extend as far westward during fall and winter, thereby giving rise to shoaling Z20 without westward propagation. During EP-El Niño, strong equatorial easterlies appear from east to west, and the center of the off-equatorial anticyclone extends westward, which deepens the thermocline and propagates westward.

Discussions and Summary
Although the global warming hiatus has been increasingly mentioned (England et al., 2014;Kosaka & Xie, 2013;Lee et al., 2015;Li et al., 2017Li et al., , 2018, ENSO events are happening more frequently and extremely, especially CP ENSO events Han et al., 2006;Kao & Yu, 2009;Lee & McPhaden, 2010;Yeh et al., 2009  This study shows that wind forcing plays a key role in the communication between the STIO and ENSO through atmospheric teleconnections. ENSO impacts the wind field in the STIO, which influences Z20 variation through Rossby waves. The thermocline variation in the STIO is impacted by a combination of the EP and CP ENSO. Compared to the EP ENSO, Z20 is more strongly related to the CP ENSO. Further analysis shows that the impact of ENSO on the positive and negative Z20A events has significantly asymmetric signatures. EP-El Niño events could induce an obvious deepening of the thermocline in the STIO and its westward signal propagation, whereas CP-La Niña events could result in strong shoaling of the thermocline. By contrast, EP-La Niña and CP-El Niño events could hardly induce thermocline displacements. The major conclusion of Xie et al. (2002) is that much of the SST variability in the western STIO is due to Rossby waves that propagate from the east, and ENSO provides major forcing for these Rossby waves. This research further emphasizes that the EP-El Niño dominates the westward propagation of these Rossby waves. Whether the Z20 signal propagates westward depends on the westward extension of equatorial winds and off-equatorial winds south of the equator. Wijffels and Meyers (2004) and Cai et al. (2005) considered the transmission of ENSO-induced thermocline anomalies from the Pacific to be important forcings that cause off-equatorial STIO thermocline variation. Cai et al. (2008) found that shoaling of the off-equatorial STIO thermocline is caused by anthropogenic forcing via numerical models. This research indicates that the change in wind forcing may induce Z20 variation. Considering that more EP-El Niño and CP-La Niña events are occurring (Kao & Yu, 2009) under global warming scenarios, large Z20 variation may take place in the STIO frequently in the future. About the two types of La Niña, some studies pointed out that the separation of the cold states of the two types of ENSO is much less clear due to most of La Niña is classified into CP events Ren & Jin, 2011). Therefore, more work will be conducted in the future based on numerical experiments to verify this research.