Sea Surface Temperature Anomalies in the Western Indian Ocean as a Trigger for Atlantic Niño Events

Abstract

Pacific El Niño and Atlantic Niño events represent prominent interannual climate fluctuations in tropical regions. Both of them have considerable impacts on the climate system and human livelihoods. The interaction between the Pacific El Niño and Atlantic Niño has received wide attention. Here, we use observations and numerical model experiments to show a new trigger from the western Indian Ocean (WIO) that can serve as a predictor for Atlantic Niño events. The warm sea surface temperature (SST) anomalies in the WIO alter the Walker circulation in the boreal winter, weaken the surface trade winds over the tropical Atlantic and favor an Atlantic Niño in the subsequent summer. These inter-basin processes also affect the interaction between the Pacific El Niño and Atlantic Niño. The results imply that a better simulation of the WIO SST and its teleconnection may help to forecast Atlantic Niño events.

Key Points

• Sea surface temperature (SST) anomalies in the western Indian Ocean (WIO) may trigger Atlantic Niño events through the Walker circulation

• SST anomalies in the WIO play an important role in the relationship between ENSO and Atlantic Niño events

• A better simulation of the WIO SST and its teleconnection may help to forecast Atlantic Niño events

Plain Language Summary

The Atlantic Niño is a prominent mode of climate variability in the equatorial Atlantic. It not only influences climate over the surrounding continents but also has an important contribution to variabilities in other tropical oceans. The ability to forecast the Atlantic Niño accurately is important to human livelihoods and requires a thorough understanding of the relevant physical mechanisms. Surface wind anomalies are important in the evolution of Atlantic Niño events, which can trigger the Atlantic Niño through the Bjerknes feedback. However, part of these surface wind anomalies can be induced remotely from Pacific El Niño events. In this study, we find the surface wind anomalies can also be induced by warm sea surface temperature (SST) anomalies in the western Indian Ocean (WIO). The warm WIO SST anomalies alter the Walker circulation in the boreal winter, weaken the surface trade winds over the tropical Atlantic and favor an Atlantic Niño in the subsequent summer. The WIO SST anomalies can also modulate the interaction between the Pacific El Niño and Atlantic Niño. A better simulation of the WIO SST may provide a predictor for Atlantic Niño events and may even provide deeper insights into tropical climate variability.

1 Introduction

The Atlantic Niño is a pronounced interannual variability mode in the tropical Atlantic (Philander, 1986; Zebiak, 1993), and it can significantly affect the climate systems of the surrounding continents (Li et al., 2016; Lübbecke et al., 2018; Wang, 2002; Xie & Carton, 2004). It is generally agreed that the summer-peaking Atlantic Niño has had a robust remote impact on Pacific El Niño-Southern Oscillation (ENSO) events since the 1970s (Ham et al., 2013; Jia et al., 2019; Rodríguez-Fonseca et al., 2009) and has also been responsible for the weakening of the ENSO-Indian monsoon relationship since 1980 (Kucharski et al., 2007; Wang et al., 2009). The ability to forecast the Atlantic Niño requires a thorough understanding of the relevant physical mechanisms. The remote forcing from other oceans can significantly influence sea surface temperature (SST) variability in the equatorial Atlantic (Latif & Grötzner, 2000; Münnich & Neelin, 2005; Wang, 2002). In particular, some studies have suggested a strong influence of ENSO on Atlantic Niño events, although there is a time lag of a few months (Latif & Grötzner, 2000; Münnich & Neelin, 2005).

However, the relationship between Pacific ENSO and the subsequent equatorial Atlantic interannual variability is weak and fragile due to the competing mechanisms (Chang et al., 2006; Lübbecke & Mcphaden, 2012; Lübbecke et al., 2018; Richter et al., 2013). El Niño, the positive phase of ENSO, can warm the tropical Atlantic through the ENSO-induced warm tropospheric temperature (Chang et al., 2006). In addition, studies have suggested that El Niño events have a robust impact on the north tropical Atlantic SST and surface wind curl, therefore warming the equatorial Atlantic through oceanic Kelvin waves and meridional heat advection (Lübbecke & Mcphaden, 2012; Richter et al., 2013). However, this warming may occasionally be counteracted by anomalous easterly winds over the equatorial Atlantic that respond to an anomalous Walker circulation during El Niño (Chang et al., 2006). Whether anomalous easterly winds appear is the key component driving the interaction between the equatorial Atlantic and Pacific. Several studies have proposed reasonable explanations for this inconsistent relationship, such as the duration of ENSO events (Tokinaga et al., 2019), but a more thorough comprehension of the corresponding mechanisms is needed.

Significant SST anomalies over the Indian Ocean are known to have broad impacts on climate variability through the Indian monsoon. The Indian Ocean SST statistically parallels ENSO variability, but multiple studies have suggested that there are actually two-way interactions (Cai et al., 2019; Du et al., 2009; Kug & Kang, 2006; Wang, 2019). For example, the strong positive Indian Ocean dipole (IOD) mode (Saji et al., 1999) events during the boreal autumn may favor the development of El Niño (Hameed et al., 2018), while the warming of the Indian Ocean basin (IOB) mode (Klein et al., 1999) during the boreal spring hastens El Niño's demise (Kug & Kang, 2006). Thus, it is possible that the Indian Ocean SST may also modulate interannual climate variability in the tropical Atlantic (Kajtar et al., 2017; Wang, 2019). Studies suggested that Indian Ocean variability has a net damping effect on Atlantic Ocean variability (Kajtar et al., 2017), and can influence the Atlantic Ocean by the westward Agulhas current (Boebel et al., 2003). However, the influence through the atmospheric bridge has not been studied and documented (Wang, 2019). In this study, we show a new trigger from the Indian Ocean that can serve as a predictor for Atlantic Niño events.

2 Data Sets, Model Experiments, and Methods

2.1 Climate Indexes

The Atlantic Niño index (Atl3) was calculated as the area-averaged SST anomalies in the region of 20°W–0°, 3°S–3°N. The Niño3 index was calculated as the area-averaged SST anomalies in the region of 150°W–90°W, 5°S–5°N. The IOB index was calculated as the area-averaged SST anomalies in the region of 40°E–100°E, 20°S–20°N. The IOD index was calculated as the difference in SST anomalies between the western (50°E–70°E, 10°S–10°N) and southeastern (90°E–110°E, 10°S–0°) tropical Indian Ocean. To better illustrate the sequence of climate variations, the Atlantic Niño year is defined as year “0,” and the previous year is defined as year “−1.”

The western Indian Ocean (WIO) index was calculated as the area-averaged residual SST (RSST) anomalies in the region of 50°E–75°E, 5°S–10°N (Figure 1a and Figure S1), with the ENSO influence removed by using linear regression with respect to the Niño3 index during the previous winter of Atlantic Niño events (November to January; ND(−1)J(0)). The RSST was defined as

where β represents the linear regression coefficient of the SST anomalies regressed onto the winter ND(−1)J(0) Niño3 index.

Figure 1

Inter-basin connections between the Atlantic Niño and the WIO. (a) Correlation between the observed winter D(−1)JF(0) SST anomalies and the boreal summer MJJ(0) Atl3 index. The impact of ENSO on the SST anomalies has been excluded by removing the linear regression with respect to the Niño3 index during the winter ND(−1)J(0). The black square represents the WIO area. The black dots represent the regions in which the correlation between the SST and Atl3 indexes is statistically significant at the 95% confidence level. (b) Twenty-one-year sliding lead-lag correlations (e.g., the correlation coefficient in 2000 represents 1990–2010) between the observed summer MJJ(0) Atl3 index and the WIO index. The contour interval is 0.1, and the negative correlation has been dashed. Only those regions for which the correlation is 95% statistically significant under a Student's t-test for the effective degrees of freedom are shaded. (c) Time series of the observed summer MJJ(0) Atl3 standardized index (red bars) and the winter D(−1)JF(0) WIO standardized index (blue line) for the period of 1950–2019. The black lines represent the values of 0.75 and −0.75. (d–f) are the observed tropical Atlantic SST anomalies during winter D(−1)JF(0), spring MAM(0), and summer MJJ(0) regressed onto the winter D(−1)JF(0) WIO index for the period after the 1970s; (g–i) are the observed Atlantic equatorial subsurface temperature anomalies (shading, °C) during winter D(−1)JF(0), spring MAM(0), and summer MJJ(0) regressed onto the winter D(−1)JF(0) WIO index for the period after the 1970s. The green contours in (g–i) indicate the climatological mean of the 23 °C isothermal depth. The black dots represent the regions in which the correlation between the SST (subsurface temperature) and WIO indexes is statistically significant at the 95% confidence level. Only those regions for which the correlation between the anomalous wind and the WIO index is statistically significant at the 95% confidence level are shown. ENSO, El Niño-Southern Oscillation; SST, sea surface temperature; WIO, western Indian Ocean.

2.2 Observational Data Sets

We use the following data sets: the Hadley Centre Sea Ice and Sea Surface Temperature (HadISST) data set (Rayner et al., 2003), available since 1870 on a 1 ° × 1 ° grid; wind velocity and air temperature data from the European Centre for Medium-Range Weather Forecasts (ECMWF) Reanalysis 40 (Uppala et al., 2005) (ERA40) data set and ECMWF interim reanalysis (Dee et al., 2011) (ERA-Interim) data set, where the former is available from September 1957 to August 2002 and the latter is available since 1979 on a 1 ° × 1 ° grid; and the monthly mean ocean temperature data from the Simple Ocean Data Assimilation version 2.2.4 (Carton et al., 2000) (SODA2.2.4), available from January 1871 to January 2014 on a 0.5 ° × 0.5 ° grid. In addition, one satellite-based observational SST data set and two reanalyzes SST data sets are used in Figure S2: the Optimum Interpolation Sea Surface Temperature version 2.1 (OISSTv2.1) data are available from September 1981 to December 2015 on a 0.25 ° × 0.25 ° grid (Reynolds et al., 2007); the Extended Reconstructed Sea Surface Temperature version 5 (ERSSTv5) data are available from January 1854 to the present on a 2 ° × 2 ° grid (Huang et al., 2017); and the Global Sea Surface Temperature Data sets (COBE-SST) are available from January 1891 to the present on a 1 ° × 1 ° grid (Folland & Parker, 1995). All of the data are first detrended. The significance tests are based on two-tailed P-values using Student's t-test. We determine the number of effective degrees of freedom by using the methods introduced by Bretherton et al. (1999).

2.3 Categorization of Events

The observational composite analyses are based on the events in which the indexes exceed 0.75 times of their standard deviation during the mature phase (Figure S3a). Based on HadISST, El Niño events accompanied by a positive (negative) WIO index and an Atlantic Niño (Niña) event occurred in 1973, 1988, 1998, and 2010 (1977, 1983, 1992, and 2015); La Niña events accompanied by a positive (negative) WIO index and an Atlantic Niño (Niña) event occurred in 1968, 1996, and 2008 (1965, 1976, 1997, and 2012). All the years mentioned here denote the Atlantic Niño year of “0” (see Text S1 for details).

2.4 Model Experiments

The CESM1.2.2 model (Hurrell et al., 2013) is used for the coupled general circulation model (CGCM) experiments, which is developed by the National Center for Atmospheric Research (NCAR). The CESM1.2.2 has difficulty in simulating the tropical Atlantic mean state but can capture the basic features of tropical Atlantic interannual variabilities and the linkages to the Pacific. Four CGCM experiments are carried out by setting different SST fields. One is the control experiment (EXP_CTRL) that is applied with the climatological SST outside the tropical Atlantic (30°S–30°N). The EXP1 experiment uses warm WIO SST anomalies added to the climatological SST in the WIO region (Figures S3b and S3c). The remaining two experiments are designed with the same warm El Niño SST anomalies in the tropical Pacific (30°S–30°N, 120°E-eastern Pacific coastline), but with negative (EXP2) or positive (EXP3) WIO anomalies in the WIO region (Figures S3d–S3g). The Student's t-test is used to determine the significance of the ensemble-mean differences between EXP_CTRL and other experiments (see Text S2 for details).

The amplitude of the WIO SST anomalies in the CGCM experiments is enhanced to 1.5°C to avoid the influence of model biases in simulating the mean SSTs of the tropical oceans. The 1.5°C is much stronger than the observed WIO SST anomalies. We also test the idea of the influence of the WIO on the tropical Atlantic by using atmospheric general circulation model (AGCM) experiments. The AGCM experiments show that the weaker amplitude of the observed WIO SST anomalies is able to induce the wind response in the equatorial Atlantic (see Text S2 and Figure S9).

3 Results

3.1 Defining the WIO

We start by examining the inter-basin connections associated with the Atlantic Niño based on observational data. Since the Atlantic Niño peaks in June, the value of the Atl3 index during the boreal summer (May to July; MJJ(0)) is selected to represent the Atlantic Niño intensity (Okumura & Xie, 2006; Zebiak, 1993). Since ENSO has a large impact on the Indian Ocean and Atlantic variability (Klein et al., 1999; Wang, 2019), the potential interaction between the tropical Indian Ocean and Atlantic may be obscured. To isolate the impact of the Indian Ocean SST on tropical Atlantic climate variability, we first remove the ENSO influence on Indian Ocean SST by using linear regression with respect to the Niño3 index during the previous winter ND(−1)J(0) of the Atlantic Niño. Based on a 5-month lagged correlation analysis, we find a robust connection between the summer MJJ(0) Atlantic Niño index and the previous winter (December to February; D(−1)JF(0)) SST anomalies over the WIO (Figure 1a and Figure S1). We define the regional average of the SST anomalies in the corresponding area as the WIO index, with the ENSO influence removed beforehand. The connection between the Atlantic Niño index and the WIO SST anomalies peaks in the boreal winter D(−1)JF(0) (Figures S4a and S4e), indicating that the WIO may play an effective role in tropical Atlantic variability.

To further elucidate the Indian Ocean-Atlantic connection, we then perform a 21-year sliding of lead-lag correlation analysis between the observed summer Atl3 index and the WIO index (Figure 1b). After the 1970s, the WIO index is significantly correlated with the Atl3 index, with the WIO leading the Atl3 by 5 months. The correlation coefficient can reach 0.6. The time series also show a relatively close connection between the winter WIO index and the subsequent summer Atl3 index after the 1970s (Figure 1c). We verify this relation in several data sets to exclude systematic biases (Figure S2). The WIO warming emerges in the boreal autumn and decays rapidly in the spring of the following year, with the maximum occurring in January (Figure S1). Shortly after the emergence of WIO warming, a weak warm signal takes place in the equatorial Atlantic. However, this signal enhances rapidly only after the peak in WIO warming (Figure S4a).

Note that the correlations of the WIO index with the IOB and IOD indexes are relatively weak (Figures S4c and S4d). The WIO index peaks in the boreal winter, whereas the IOB mode maximizes in the boreal spring and the IOD mode peaks in the boreal autumn. Since the linear effect of ENSO is already removed beforehand, the WIO index may represent an internal variability in the tropical Indian Ocean or the nonlinear effect of ENSO. If the ENSO influence is not removed beforehand, the correlation between the raw WIO index and the ATL3 index is weaker (Figure S4f). This is because the raw WIO index includes the ENSO signal (two indexes are correlated) and the ENSO index is not correlated with the Atl3 index.

3.2 Mechanism of Influencing the Atlantic Niño

The horizontal maps of the observed SST regressed onto the WIO index reveal the evolution process of Atlantic Niño events associated with Indian Ocean warming. In response to WIO warming, slight warming occurs in the equatorial Atlantic during the boreal winter (Figure 1d) and is accompanied by both anomalous westerly winds over the western part of the equatorial Atlantic (Figure 1d) and positive upper ocean temperature anomalies in the east (Figure 1g). The SST, surface wind, and upper ocean temperature anomalies further develop in the boreal spring and finally reach the peak in the following summer (Figures 1d–1i), when the Atlantic cold tongue reaches its maximum (Grodsky & Carton, 2003; Tokinaga & Xie, 2011).

To support the results and shed further light on the dynamical processes of how the WIO SST anomalies force equatorial Atlantic variability, we perform two CGCM experiments based on CESM1.2.2 (Hurrell et al., 2013). To isolate the impacts of the WIO warming (Figures S3b and S3c), the only model design difference between the two CGCM experiments is whether they are carried out with or without the positive WIO SST anomalies (hereafter EXP1 and EXP_CTRL, respectively).

The model results show that the WIO warming triggers the Atlantic Niño in the subsequent summer through atmospheric and oceanic processes. During the boreal winter, the WIO warming triggers Walker circulation anomalies, with an ascending branch over the WIO and a descending branch over the tropical Atlantic (Figure 2a). The descending air draft leads to an anomalous convergence inflow at the top of the tropical Atlantic troposphere (Figure S5a), and it is accompanied by anomalous divergence centered over the surface of the western tropical Atlantic (Figure S5d). In response to the low-level divergence, westerly wind anomalies appear over the surface of the tropical Atlantic (Figure 2d).

Figure 2

The impact of the WIO SST in the CGCM experiments. Shown are the ensemble-mean differences between the EXP1 and EXP_CTRL experiments during the winter D(−1)JF(0), spring MAM(0), and summer JAS(0). (a–c) are the ensemble-mean differences of Walker circulation anomalies by averaging divergent wind (vector, m/s) and vertical velocity anomalies (vector and shading, −0.02 Pa/s) between 2°S and 2°N; (d–f) are the ensemble-mean differences of SST anomalies (shading, °C) and surface wind anomalies (vector, m/s) at 925 hPa; (g–i) are the ensemble-mean differences of equatorial subsurface temperature anomalies (shading, °C). The green contours in (g–i) indicate the climatological mean of the 23°C isothermal depth. Wind anomalies that exceed the 95% confidence level are black in color; temperature anomalies that exceed the 95% confidence level are dotted. CGCM, coupled general circulation model; SST, sea surface temperature; WIO, western Indian Ocean.

The westerly wind anomalies give rise to significant oceanic responses along the equatorial Atlantic. During the boreal winter, D(−1)JF(0), the westerly wind anomalies generate eastward-propagating downwelling Kelvin waves and rapidly thicken the equatorial thermocline depth (Philander, 1986) (Figure 2g). The forcing from the Indian Ocean sustains the anomalous Walker circulation and westerly surface wind anomalies through the subsequent spring (Figures 2b and 2e). In response to the persistent westerly wind anomalies, the eastward-propagating downwelling Kelvin waves induce a large warm signal to the east, therefore dramatically deepening the thermocline in the eastern tropical Atlantic (Figure 2h). As the climatological thermocline depth in the eastern equatorial Atlantic is shallow during the boreal summer, the deepened thermocline cuts off the cold heat advection from the deep ocean and rapidly warms the SST in the eastern Atlantic cold tongue area as a consequence (Dippe et al., 2019) (Figure 2i). Thus, an Atlantic Niño finally emerges under remote forcing from the WIO (Figure 2f).

3.3 Contributions of the WIO During El Niño Events

Thus far, we have shown the entire dynamical inter-basin interaction process in which the WIO warming during the boreal winter triggers an Atlantic Niño event in the subsequent summer. However, since both the WIO and ENSO vary on interannual time scales and usually peak in the boreal winter, some WIO events may occur with El Niño or La Niña events. The atmospheric responses to WIO events are dramatically disturbed by these ENSO events since the amplitude of WIO events is weaker than ENSO events (Figure S6, see Text S3 for details). Nevertheless, based on composites of the selected observational events (Section 2), we find that the SST anomalies in the WIO still play an important role in the relationship between ENSO and Atlantic Niño events. The SST anomalies in the WIO change El Niño's influences on the tropical Atlantic, hence producing two types of entirely different initiation conditions during the previous winter of Atlantic Niño events.

During the winter of El Niño with negative WIO events, the El Niño-forced Walker circulation anomalies are not affected by the Indian Ocean (Figure 3a). The anomalous Walker circulation induced by El Niño tends to trigger an anomalous descending branch over the eastern tropical Atlantic (Figures 3b and 3c). Thus, easterly surface wind anomalies are generated over the tropical Atlantic and consequently cooled the SST (Figure 3d).

Figure 3

Composites of the atmospheric responses to negative and positive WIO events in the winter D(−1)JF(0). (a–d) are the composites of the observed atmospheric and SST responses during El Niño and negative WIO events; (e–h) are the composites of the observed atmospheric and SST responses during El Niño and positive WIO events. Walker circulation anomalies are calculated by averaging wind (vector, m/s) and vertical velocity anomalies (vector and shading, −0.002 Pa/s) between 2°S and 2°N in (a and e); the contours are 0, ±1, ±2, ±5, ±10, ±20, and ±30. Velocity potential anomalies (shading, 10⁶ m²/s) and divergent wind anomalies (vector, m/s) at 200 hPa are shown in (b and f). Velocity potential anomalies (shading, 10⁶ m²/s) and divergent wind anomalies (vector, m/s) at 925 hPa are shown in (c and g). SST anomalies (shading, °C) and surface wind anomalies (vector, m/s) at 925 hPa are shown in (d and h). Wind anomalies that exceed the 90% confidence level are black in color; velocity potential anomalies in (b, c, f, and g), and SST anomalies in (d and h) that exceed the 90% confidence level are dotted. All composites are during the winter D(−1)JF(0) in which “0” denotes the Atlantic Niño year as well as the El Niño decaying year. SST, sea surface temperature; WIO, western Indian Ocean.

In contrast, during the winter of El Niño with positive WIO events, both El Niño and WIO warming induce anomalous regional ascending air drafts, together with altering the Walker circulation over the global tropical region (Figure 3e). The anomalous Walker circulation induced by the WIO warming tends to trigger an anomalous descending branch over the tropical Atlantic and westerly wind anomalies at the surface (Figure 2). Therefore, the anomalous Walker circulation that forms in response to the WIO warming competes with the effect from the Pacific over the tropical Atlantic (Figures 3e–3g), counteracts the easterly surface wind anomalies induced by El Niño (Figure 3h) and eliminates the cooling effect from El Niño. As a consequence of the counteracting effect between the WIO and El Niño warmings, the tropical Atlantic shows no significant wind anomalies (Figure 3h). However, the tropical Atlantic warms through the tropospheric temperature mechanism, which does not necessarily require a wind change (see Text S4 for details). The contributions of the WIO are more obvious during certain extreme El Niño events, which also verify the importance of WIO events (see Text S5 for details).

We further prove the above observational results by two additional CGCM experiments. Both experiments are designed with the same warm El Niño SST anomalies in the tropical Pacific but with negative (hereafter as EXP2) or positive (hereafter as EXP3) WIO anomalies in the Indian Ocean (Figures S3d–S3g). Thus, the different tropical Atlantic conditions after El Niño are caused by only the WIO SST, since it is the only difference in the experimental designs of EXP2 and EXP3. The EXP2 (EXP3) result confirms that the negative (positive) WIO SST anomalies during El Niño induce easterly (westerly) surface wind anomalies over the equatorial Atlantic through the Walker circulation (Figure S7). These easterly (westerly) surface wind anomalies uplift (deepened) the eastern tropical Atlantic thermocline, thus creating cold (warm) SST anomalies during the initiating period of Atlantic Niño events. Therefore, during the subsequent summer, the tropical Atlantic tends to undergo an Atlantic Niña (Niño) event in the EXP2 (EXP3) experiment (Figure S7).

4 Discussion and Conclusions

Our study shows that the WIO plays an important role in the development of tropical Atlantic variability, and is a key factor that modulates the interaction between the Pacific and Atlantic (Figure 4). We find that the WIO SST anomalies can induce an Atlantic Niño, and determine whether an Atlantic Niño or Atlantic Niña follows a Pacific El Niño event. The new finding here explains why the relationship between the Pacific El Niño and Atlantic Niño is weak and fragile in addition to the competing mechanisms reported previously (e.g., Chang et al., 2006; Lübbecke & Mcphaden, 2012; Lübbecke et al., 2018; Richter et al., 2013). The result also fills in a gap in the interaction among the three tropical oceans.

Figure 4

Schematic diagram of inter-basin interactions in tropical oceans. (a) Pacific El Niño warming generates an anomalous Walker circulation and forces surface easterly wind anomaly over the tropical Atlantic which enhances the zonal thermocline slope and cools the SST in the tropical eastern Atlantic. (b) The WIO warming generates an anomalous Walker circulation that competes with that from the Pacific. The anomalous Walker circulation forces surface westerly wind anomaly over the tropical Atlantic which weakens the zonal thermocline slope and warms the SST in the tropical eastern Atlantic. Positive (negative) SST and thermocline anomalies are shaded in red (blue). SST, sea surface temperature; WIO, western Indian Ocean.

The interaction between ENSO and Atlantic Niño events has received wide attention. A recent study shows that the duration of ENSO events is a key factor for the ENSO-Atlantic Niño relationship (Tokinaga et al., 2019). The multiyear ENSO events sustain easterly wind anomalies over the tropical Atlantic, while the single-year ENSO events have no robust relationship with the Atlantic Niño. Our study suggests that the remote influences from the WIO via the Walker circulation are important. In the sensitivity CGCM EXP2 and EXP3 experiments, the influence from the duration of El Niño was completely removed since both the experiments used the same El Niño-type SST anomalies (Figure S3). Moreover, the extreme El Niño events that occurred in 1982/1983 and 1997/1998 induced completely opposite responses in the tropical Atlantic (Figure S8, see Text S5 for details). Both the extreme El Niño events decayed rapidly and transferred to La Niña in the following winter (i.e., the single-year El Niño). Thus, the different Atlantic responses to the single-year El Niño can be easily explained by the WIO influence and the corresponding Walker circulation anomalies.

In addition, previous studies suggest that anomalous Indian Ocean warming helps rapidly terminate El Niño and induces a faster transition to La Niña (Kug & Kang, 2006), which is also related to the duration of ENSO events. These suggest that the formation and influence of the WIO should receive further investigation in future studies. A better simulation of the WIO SST and its teleconnection may provide a reliable predictor for Atlantic Niño events (Text S6) and may even provide deeper insights into tropical climate variability.

Acknowledgments

The authors thank two reviewers whose comments and suggestions helped improve the manuscript, Dr. Yao Fu and others in our group for the help and discussion, and Dr. Sheng Chen and Dr. Zhenya Song for the help in the model experiments. This study is supported by the National Key R&D Program of China (2019YFA0606701), the National Natural Science Foundation of China (41731173), the Strategic Priority Research Program of the Chinese Academy of Sciences (XDB42000000 and XDA20060502), Key Special Project for Introduced Talents Team of Southern Marine Science and Engineering Guangdong Laboratory (Guangzhou) (GML2019ZD0306), the Innovation Academy of South China Sea Ecology and Environmental Engineering, the Chinese Academy of Sciences (ISEE2018PY06), and the Leading Talents of Guangdong Province Program. The numerical simulation is supported by the High-Performance Computing Division in the South China Sea Institute of Oceanology.

Open Research

Data Availability Statement

The HadISST data are available at https://www.metoffice.gov.uk/hadobs/hadisst/data/download.html. The ERA40 and ERA-Interim data are available at https://apps.ecmwf.int/datasets/data/interim-full-moda/levtype=pl/. The SODA2.2.4 data are available at https://iridl.ldeo.columbia.edu/SOURCES/.CARTON-GIESE/.SODA/.v2p2p4/?Set-Language=en. The OISSTv2.1 data are available at https://www.ncdc.noaa.gov/oisst/optimum-interpolation-sea-surface-temperature-oisst-v21, the ERSSTv5 data are available at https://www.ncdc.noaa.gov/data-access/marineocean-data/extended-reconstructed-sea-surface-temperature-ersst-v5, and the COBE-SST data are available at http://ds.data.jma.go.jp/tcc/tcc/products/elnino/cobesst/cobe-sst.html