1. Introduction
Originating in the tropical Pacific, El Niño–Southern Oscillation (ENSO) is the dominant signal of interannual variability and has tremendous impacts on global climate via teleconnections (e.g., Webster and Palmer 1997; Wallace et al. 1998; Timmermann et al. 2018; Fedorov et al. 2020; Wei et al. 2020; Wang et al. 2021; Yang and Huang 2021). Although ENSO has its strongest variability in boreal winter, it is often considered as a predictor of East Asian summer climate (e.g., Huang and Wu 1989; Zhang et al. 1996; Chang et al. 2000; Wang et al. 2000; Wu and Wang 2002; Wu et al. 2003; Xie et al. 2016; Zhang et al. 2016; Li et al. 2018). When ENSO is decaying, the sea surface temperature anomalies (SSTAs) in other ocean basins can respond through various interbasin interactions (e.g., Du et al. 2009; Cai et al. 2019; Wang 2019). The SSTAs in other basins can persist into the following summer and play an essential role in the close relationship between winter ENSO and East Asian summer climate (e.g., Wang et al. 2000; Yang et al. 2007; Wu et al. 2009; Xie et al. 2009; Rong et al. 2010; Li et al. 2018; Tang et al. 2022).
Although some typical SSTAs in other basins responding to ENSO SSTAs during the post-ENSO seasons have been identified, the evolutions of ENSO-related SSTAs show considerable case-to-case diversity (e.g., Feng et al. 2014; Xu et al. 2020). Some ENSO events maturing in boreal winter fast convert to the opposite phase in the postseason, whereas some other ENSO events slowly decay and remain the phase to the postseason (Chen et al. 2012; Feng et al. 2014; Jiang et al. 2019). From the perspective of summer ENSO, the summer ENSO events could newly emerge from the previous spring, but also could continue from the previous winter (Yang and Huang 2021). The diversity of ENSO decaying is a key factor disrupting the relationship between ENSO and the East Asian summer climate (Jiang et al. 2019; Zhou et al. 2019b; Wang et al. 2020).The low predictability of tropical SSTAs during the post-ENSO period, limited by the great diversity of ENSO evolution, is generally referred to as the spring predictability barrier (SPB) (e.g., Webster and Yang 1992; Latif et al. 1994).
Although the mechanism of the SPB, such as seasonal growth rate amplitude, seasonal thermodynamic damping, etc. (e.g., Jin et al. 2006; Jin et al. 2020, 2021), have been prospectively studied, the identification of the initial signal affecting the development of the SPB also plays a vital role in improving summer ENSO forecasting. Chen et al. (1995, 2004) emphasized the role of initial errors during boreal winter in ENSO forecasting. Several early noise signals in the tropical Pacific have been explored in previous studies. Duan et al. (2009) identified initial errors that are often located in the equatorial Pacific with a zonal dipolar pattern, which was also verified by Zhang et al. (2014) using the model outputs from phase 5 of the Coupled Model Intercomparison Project. Duan and Hu (2015) pointed out that spatial type errors in the central-eastern equatorial Pacific were caused by the Bjerknes feedback and eastward Kelvin waves through equatorial upwelling. Recently, a study by Hua and Su (2020) indicated that an initial cold signal from the southeastern Pacific could hinder the development of ENSO through local positive feedback, leading to a failed prediction of ENSO.
Other ocean basins—specifically, the Indian Ocean and Atlantic Ocean basins—can also modulate ENSO variations through atmospheric circulation and the ocean “conveyor belt” (Cai et al. 2019; Wang 2019; Fedorov et al. 2020; Exarchou et al. 2021; Ham et al. 2021; Wang and Wang 2021). Zhou et al. (2019a) argued that the Indian Ocean could cause the initial condition errors of the SPB through the tropical oceanic channel related to Indonesian throughflow and atmospheric bridge. Recently, Zhao et al. (2020) suggested that considering the leading modes of ENSO-related SSTAs in the whole of the tropics can markedly improve the predictability of tropical SSTAs and boreal summer climate.
Although numerous regional signals associated with the diversity of ENSO evolution have been reported, the key signals and processes among them remain unclear. Therefore, in the present study, we analyzed the leading modes of the evolutions of ENSO-related SSTAs from the perspective of interbasin interactions through the empirical orthogonal function (EOF) and extended EOF (EEOF) analyses. We identified that the ENSO decaying periods (from September to August) contains a steady ENSO decaying mode and an unsteady ENSO developing mode simultaneously, and two signals located in the South Atlantic and equatorial South Pacific from the ENSO developing mode are recognized as the main sources influencing the diversity of ENSO evolution. The organization of the paper is as follows. Section 2 briefly introduces datasets and methods. Section 3 identifies and verifies the key signals of the ENSO diversity from the lead–lag regressions of seasonal SSTAs and SSTA-residual EOF. Section 4 confirms the decisive role of the ENSO developing mode stability in the ENSO evolution by using a prediction model to predict the July–August (JJA) SSTAs. Finally, the conclusions and discussion are given in section 5.
2. Datasets and methods
a. Datasets
The SST data employed in this study are from the monthly mean of the Hadley Centre Sea Ice and Sea Surface Temperature (HadISST) dataset (Rayner et al. 2003). Moreover, all the results were verified by the Extended Reconstructed SST dataset, version 5 (ERSST.v5) (Huang et al. 2017), and Centennial Observation-Based Estimates of SST, version 2 (COBESST2) (Hirahara et al. 2014). The results from these three datasets were quite close, and thus this paper mainly shows the results of HadISST. After 10-yr high-pass filtering for SSTAs, we obtained the interannual anomalies by removing the long-term climatological mean and linear trend of the SSTAs in 1948–2014. The Niño-3 index was calculated by the averaged SSTAs over the region (5°S–5°N, 90°–150°W); while the Niño-4 index was calculated by the averaged SSTAs over the region (5°S–5°N, 160°E–150°W).
b. EEOF
EOF analysis is a traditional method used to extract the leading spatial mode from a linear consideration (e.g., Kawamura 1994; Deser et al. 2010). EEOF analysis, in which multiple timeslice anomalies are combined as the input of EOF analysis, was developed to extract the spatiotemporal coupled characteristics that could otherwise be ignored in EOF analysis (Weare and Nasstrom 1982; Hannachi et al. 2007). In the present application of EEOF analysis, the monthly SSTAs in 1948–2014 were divided into 66 segments, each of which included 12 months from September to August, i.e., September 1948–August 1949, September 1949–August 1950, …, September 2013–August 2014. Then, the four-dimensional [year] × [month] × [lat] × [lon] SSTAs were combined into two-dimensional [year] × [month × lat × lon] SSTAs to be put into a traditional EOF analysis. The EEOF analysis was able to obtain the leading modes with SSTAs evolving from September to August and corresponding 66-year principal components (PCs). The segments from September to August cover a part of typical developing periods and decaying periods of ENSO. We also performed the analyses using the segments from December to August (Fig. S1 in the online supplemental material). This selection does not influence our conclusions. Considering that the SSTAs in boreal autumn include important signals for the evolution of ENSO decaying, we select the segments starting from September in this study.
To identify the source of the key noise signals, we also analyzed two traditional EOFs, which are susceptible to noises, and the associated lead–lag regression to compare with the EEOF results. The first EOF was for the averaged tropical SSTAs in December–February (DJF) of the 66 segments, and the other was for the averaged tropical SSTAs in JJA of the 66 segments. Also, subsequently, the SSTAs in other months were regressed onto the PCs of the two EOFs, respectively. The two EOFs extracted the leading modes of tropical SSTAs in boreal winter and summer, respectively, and the associated lead–lag regressions described the SSTA evolutions related to these modes. To investigate the role of interbasin interaction in ENSO evolution, the SSTAs in the whole of the tropics were first input into the EEOF and EOF analyses, and then the tropical Pacific SSTAs were input into the DJF and JJA EOFs for comparison.
c. Prediction for the JJA SSTAs
The above coefficients were obtained using the “take-one-out” hindcast. First, the first year’s data were excluded and EOF was performed using the remaining 65 years’ DJF SSTAs within the whole tropics; then, Regpattern1 and Regpattern2 were obtained by regressing the 65 years’ JJA SSTAs onto PC1 and PC3 [DJF EOF1 and EOF3 denote the ENSO decaying mode and the ENSO developing mode, respectively; see Fig. 4 for details]. Next, Cpattern1 and Cpattern2 were calculated with the projection of the first year DJF SSTAs onto EOF1 and EOF3, and the JJA SSTAs of the first year were predicted using Eq. (1); and finally, the above step was repeated for the other years to obtain the projected JJA SSTAs for the period of 1949–2014. To verify the superiority of interbasin interaction with the continuous SSTAs in the connection between pre-SSTAs and JJA SSTAs, the JJA SSTAs were also predicted based on EEOF from September to February, September to March, and September to April. For comparison, the Pacific SSTAs were input into the EOF and EEOF for prediction. Besides, a similar independent forecast was performed using 70% training set (1948–94) to predict the remaining 30% test set (1995–2014).
3. Source of the ENSO diversity
a. Evolution of ENSO based on EEOF and EOF
The seasonal SSTA evolutions of the EEOF analysis from September to August are shown in Fig. 1. The spatial pattern of EEOF1, accounting for 35.53% of the annual explained variance during ENSO typical decaying periods, shows a typical winter El Niño decaying characteristic (Figs. 1a–d), peaking in DJF, declining in the following spring, and reversing to a La Niña in the subsequent summer. Hereafter, an evolution similar to EEOF1 is referred as the ENSO decaying mode, which is also termed in Shi and Wang (2018) to emphasize the evolution of the spatial distribution of climate systems. Since EEOF extracts a linear process, the opposite situation occurs during La Niña. The correlation between the EEOF1-related PC1 and DJF Niño-3 index reaches 0.97. With the tropical Pacific SSTA decaying, the SSTAs in the Indian Ocean evolve from an Indian Ocean dipole mode in the previous autumn to a basin warming mode in the subsequent winter and spring, which has been well documented in previous studies (e.g., Klein et al. 1999; Du et al. 2009). Meanwhile, the tropical North Atlantic warms in the boreal winter, peaks in spring, and dissipates in summer, which is also well known (Enfield and Mayer 1997; Chiang and Sobel 2002).
Seasonal SSTA evolution for (a)–(d) EEOF1, (e)–(h) EEOF2, and (i)–(l) EEOF3 within the whole tropics (20°S–20°N) from September to the following August using HadISST, explaining 35.52%, 14.03%, and 4.13% of the annual variance, respectively.
Citation: Journal of Climate 35, 12; 10.1175/JCLI-D-21-0892.1
Seasonal SSTA evolution for (a)–(d) EEOF1, (e)–(h) EEOF2, and (i)–(l) EEOF3 within the whole tropics (20°S–20°N) from September to the following August using HadISST, explaining 35.52%, 14.03%, and 4.13% of the annual variance, respectively.
Citation: Journal of Climate 35, 12; 10.1175/JCLI-D-21-0892.1
Seasonal SSTA evolution for (a)–(d) EEOF1, (e)–(h) EEOF2, and (i)–(l) EEOF3 within the whole tropics (20°S–20°N) from September to the following August using HadISST, explaining 35.52%, 14.03%, and 4.13% of the annual variance, respectively.
Citation: Journal of Climate 35, 12; 10.1175/JCLI-D-21-0892.1
In contrast to EEOF1, EEOF2, accounting for 14.03% of the annual explained variance from September to August (Figs. 1e–h). The tropical Pacific SSTAs emerge from the winter in the eastern equatorial Pacific and grow to an apparent El Niño in the following summer. Similarly, the SSTA evolution features similar to EEOF2 are referred as ENSO developing mode. In contrast to the terms of ENSO decaying/developing periods or phases emphasizing the independent time periods (Wu et al. 2014; Cao et al. 2017), the present terms of ENSO decaying/developing modes (EEOF1 and EEOF2) can coexist in the same period from September to August, which often recognized as the ENSO decaying periods. The correlation between the EEOF2-related PC2 and JJA Niño-3 index is 0.94. Meanwhile, the tropical Atlantic shows a typical Atlantic Niña mode, which peaks in the previous autumn and winter—one season before the development of ENSO. This advanced Atlantic Niña could be an essential trigger of El Niño through the Walker circulation and the oceanic Kelvin waves (Rodríguez-Fonseca et al. 2009; Ding et al. 2012; Ham et al. 2013a,b; Polo et al. 2014; Chikamoto et al. 2020; Ham et al. 2021). Unlike the tropical Atlantic, there is little change in the Indian Ocean until the following summer.
EEOF3 (Figs. 1i–l) shows a persistent dipole mode in the equatorial central-eastern Pacific. The correlation coefficient between the combination of PC1 + PC3 and the DJF Niño-4 index exceeds 0.8. Differing from EEOF1–2 reflecting the diversity of ENSO’s temporal evolution, EEOF3 resembles to the second mode of transitional monthly SSTAs EOF analysis, which was believed to reflect the diversity of ENSO spatial patterns, i.e., the nonlinear signal of eastern Pacific ENSO and central Pacific ENSO (Takahashi et al. 2011). Since EEOF1 and EEOF2 contain the complete development–decay process of ENSO, our attention will focus on EEOF1–2, and EEOF3 will be discussed in future studies.
To highlight the stability of the ENSO decaying/developing modes, the classic EOF is performed to compare with the EEOF results. We present the leading EOF modes within the DJF and JJA SSTAs over the whole tropics and the Pacific (20°S–20°N, 110°E–80°W) in Fig. 2. Both in DJF and JJA, the first mode captures a typical El Niño pattern but is accompanied by different responses in other ocean basins (Figs. 2a,d). The evolution of DJF EOF1 reflects the mature period of ENSO, while the JJA EOF1 reflects the development period of ENSO, which represents different ENSO evolutionary features (Figs. S2 and S3). With ENSO’s strongest variability concentrated in winter, the seasonal explained variance of DJF EOF1 (Fig. 2a) is greater than JJA EOF1 (Fig. 2d). EOF2 and EOF3 in DJF and JJA also exhibit the specific seasonal structure of the EEOF in Fig. 1. DJF EOF2 and JJA EOF3 present the ENSO asymmetric mode, and DJF EOF3 and JJA EOF2 present the developing and decaying modes, respectively (Figs. S2 and S3). The robustness of EOF analysis is confirmed by confining the domain to the Pacific (Figs. 2g–l). The results show no apparent change in the three leading EOF modes, except that the seasonal explained variance of the EOF1 in DJF (JJA) rises to 70.57% (59.97%). Compared to the EOF method, the EEOF method can divide the typical ENSO during decaying periods into a decaying and a developing mode and does not depend on seasonal changes.
(a)–(f) Leading EOF modes for the SSTAs in (a)–(c) the whole tropics (20°S–20°N) in DJF and (d)–(f) the whole tropics in JJA using HadISST. (g)–(l) As in (a)–(f), but for the EOF with the Pacific (20°S–20°N, 110°E–80°W) SSTAs. The seasonal explained variance is displayed at the upper right of each panel.
Citation: Journal of Climate 35, 12; 10.1175/JCLI-D-21-0892.1
(a)–(f) Leading EOF modes for the SSTAs in (a)–(c) the whole tropics (20°S–20°N) in DJF and (d)–(f) the whole tropics in JJA using HadISST. (g)–(l) As in (a)–(f), but for the EOF with the Pacific (20°S–20°N, 110°E–80°W) SSTAs. The seasonal explained variance is displayed at the upper right of each panel.
Citation: Journal of Climate 35, 12; 10.1175/JCLI-D-21-0892.1
(a)–(f) Leading EOF modes for the SSTAs in (a)–(c) the whole tropics (20°S–20°N) in DJF and (d)–(f) the whole tropics in JJA using HadISST. (g)–(l) As in (a)–(f), but for the EOF with the Pacific (20°S–20°N, 110°E–80°W) SSTAs. The seasonal explained variance is displayed at the upper right of each panel.
Citation: Journal of Climate 35, 12; 10.1175/JCLI-D-21-0892.1
b. Identify the key signals from lead–lag regression
To compare the evolution features between EOF and EEOF, Figs. 3 and 4 show the lead–lag regression of seasonal SSTAs onto the PCs of EOF and EEOF. The explained annual variance of the lead–lag SSTAs during ENSO typical decaying periods is displayed on the top of each column. Due to the dominance of the strongest variability of winter SSTAs within the equatorial central-eastern Pacific, it is apparent that DJF EOF1 (Figs. 3e–h) presents the ENSO decaying mode as EEOF1 (Figs. 1a–d, 3a–d) and has nearly equal annual explained variance reaching 35% from September to August. The result is stable and does not depend on the tropics or the pacific (the results of DJF EOF1 and EEOF1 within the pacific are not shown). However, the surprise is that the SSTA evolutions related to JJA EOF2 (Figs. 3i–p) show a highly consistent pattern with Figs. 1a–d and 3a–h. Although JJA EOF2 within the tropics or JJA EOF3 within the Pacific explains only 10.17% or 7.20% of the JJA SSTA variance and shows a relatively weak pattern, the annual explained variance for the whole pattern from September to August exceeds 25% and 18%, respectively. This result illustrates that the complete El Niño decaying phase can be well reconstructed from a weaker La Niña signal in boreal summer to the mature El Niño in the previous winter, with an intensity basically equivalent to that obtained directly from EEOF1 and DJF EOF1. The result indicates that the ENSO decaying mode is very steady and structured, and mainly dominated by the Pacific SSTAs.
The lead–lag regressions of seasonal SSTAs onto the (a)–(d) EEOF PC1 and (e)–(h) PC1 of the EOF in DJF, and the (i)–(l) PC2 of the EOF in JJA, with the SSTAs in the whole tropics during 1948–2014 using HadISST. (m)–(p) As in (i)–(l), but for the PC3 of the EOF with the SSTAs in Pacific. Hatching denotes statistical significance exceeding the 95% confidence level. The annual explained variances for the continuous SSTAs from September to the following August are shown on the top of each column.
Citation: Journal of Climate 35, 12; 10.1175/JCLI-D-21-0892.1
The lead–lag regressions of seasonal SSTAs onto the (a)–(d) EEOF PC1 and (e)–(h) PC1 of the EOF in DJF, and the (i)–(l) PC2 of the EOF in JJA, with the SSTAs in the whole tropics during 1948–2014 using HadISST. (m)–(p) As in (i)–(l), but for the PC3 of the EOF with the SSTAs in Pacific. Hatching denotes statistical significance exceeding the 95% confidence level. The annual explained variances for the continuous SSTAs from September to the following August are shown on the top of each column.
Citation: Journal of Climate 35, 12; 10.1175/JCLI-D-21-0892.1
The lead–lag regressions of seasonal SSTAs onto the (a)–(d) EEOF PC1 and (e)–(h) PC1 of the EOF in DJF, and the (i)–(l) PC2 of the EOF in JJA, with the SSTAs in the whole tropics during 1948–2014 using HadISST. (m)–(p) As in (i)–(l), but for the PC3 of the EOF with the SSTAs in Pacific. Hatching denotes statistical significance exceeding the 95% confidence level. The annual explained variances for the continuous SSTAs from September to the following August are shown on the top of each column.
Citation: Journal of Climate 35, 12; 10.1175/JCLI-D-21-0892.1
The lead–lag regressions of seasonal SSTAs onto the (a)–(d) EEOF PC2 and (e)–(h) PC1 of the EOF in JJA, and the (i)–(l) PC3 of the EOF in DJF, with the SSTAs in the whole tropics during 1948–2014 using HadISST. (m)–(p) As in (i)–(l), but for the PC3 of the EOF with the SSTAs in Pacific. Hatching denotes statistical significance exceeding the 95% confidence level. The annual explained variances for the continuous SSTAs from September to the following August are shown on the top of each column.
Citation: Journal of Climate 35, 12; 10.1175/JCLI-D-21-0892.1
The lead–lag regressions of seasonal SSTAs onto the (a)–(d) EEOF PC2 and (e)–(h) PC1 of the EOF in JJA, and the (i)–(l) PC3 of the EOF in DJF, with the SSTAs in the whole tropics during 1948–2014 using HadISST. (m)–(p) As in (i)–(l), but for the PC3 of the EOF with the SSTAs in Pacific. Hatching denotes statistical significance exceeding the 95% confidence level. The annual explained variances for the continuous SSTAs from September to the following August are shown on the top of each column.
Citation: Journal of Climate 35, 12; 10.1175/JCLI-D-21-0892.1
The lead–lag regressions of seasonal SSTAs onto the (a)–(d) EEOF PC2 and (e)–(h) PC1 of the EOF in JJA, and the (i)–(l) PC3 of the EOF in DJF, with the SSTAs in the whole tropics during 1948–2014 using HadISST. (m)–(p) As in (i)–(l), but for the PC3 of the EOF with the SSTAs in Pacific. Hatching denotes statistical significance exceeding the 95% confidence level. The annual explained variances for the continuous SSTAs from September to the following August are shown on the top of each column.
Citation: Journal of Climate 35, 12; 10.1175/JCLI-D-21-0892.1
For the ENSO developing mode, both the SSTAs and the annual explained variance associated with JJA EOF1 (Figs. 4e–h) are very close to those of EEOF2 (Figs. 1e–h, 4a,b), whereas the SSTAs related to DJF EOF3 (Figs. 4i–l) are much weaker than those of JJA EOF1 and EEOF2, and the annual explained variance for the whole pattern is only 6.03%. The discrepancies in the SSTAs related to DJF EOF3 can be traced back to the previous winter. In EEOF2 and JJA EOF1 during winter (Figs. 4b and f), there are significant SSTAs in the southeastern Pacific with the same sign as the SSTAs in the equatorial Pacific; however, in DJF EOF3 (Fig. 4j), the SSTAs in the equatorial Pacific and the southern flank of the basin show a meridional dipole pattern. Therefore, the equatorial South Pacific could be a key signal affecting the development of ENSO, which has been well observed in previous studies from other perspectives (Duan et al. 2009; Duan and Hu 2015; Hua and Su 2020).
However, DJF EOF3, with the Pacific SSTAs in Figs. 4m–p, totally fails to reconstruct the ENSO developing mode. Unlike the ENSO decaying mode, the explained variance for the ENSO developing mode from September to August related to DJF EOF3 is only about 2.78%, making the equatorial Pacific extremely vulnerable to the noise of the surrounding SSTA forcing and other ocean remote forcing. The result suggests the unreliable ENSO developing mode may be the main source of evolutionary diversity during ENSO typical decaying periods.
In DJF (Fig. 4n), the meridional dipole SSTA mode in the equatorial Pacific and southern flanks of the basin is even more apparent than that in Fig. 4j. Moreover, in contrast to the SSTAs in Figs. 4a,b,e,f,i,j, there are no negative South Atlantic SSTAs during the previous autumn and winter (Figs. 4m,n) due to the lack of interbasin interaction when only using the Pacific for EOF analysis. Hence, South Atlantic SSTAs should be another key signal determining the development of summer ENSO, the mechanisms of which have also been studied (Ding et al. 2012; Keenlyside et al. 2013; Polo et al. 2014; Chikamoto et al. 2020). The Atlantic Niña in spring and boreal summer can generate El Niño by driving westward wind anomalies related to Walker circulation (Ding et al. 2012) or equatorial Kelvin waves (Polo et al. 2014). Subsequently, Chikamoto et al. (2020) further indicated that the negative equatorial South Atlantic SST anomalies are the main contributor to equatorial Pacific warming. Similar lead–lag regression analyses were performed with ERSST and COBESST2 in Fig. 5. The DJF EOF3 for the whole tropics can evolve into summer ENSO in JJA (Figs. 5a–d,i–l), whereas the DJF EOF3 for only the Pacific (Figs. 5e–h,m–p) cannot develop into summer ENSO in JJA due to the lack of interbasin interaction as in Figs. 4m–p.
(a)–(p) As in Figs. 4i–p, but for the dataset of (a)–(h) ERSST and (i)–(p) COBESST2. Hatching denotes statistical significance exceeding the 95% confidence level. The annual explained variances for the continuous SSTAs from September to the following August are shown on the top of each column.
Citation: Journal of Climate 35, 12; 10.1175/JCLI-D-21-0892.1
(a)–(p) As in Figs. 4i–p, but for the dataset of (a)–(h) ERSST and (i)–(p) COBESST2. Hatching denotes statistical significance exceeding the 95% confidence level. The annual explained variances for the continuous SSTAs from September to the following August are shown on the top of each column.
Citation: Journal of Climate 35, 12; 10.1175/JCLI-D-21-0892.1
(a)–(p) As in Figs. 4i–p, but for the dataset of (a)–(h) ERSST and (i)–(p) COBESST2. Hatching denotes statistical significance exceeding the 95% confidence level. The annual explained variances for the continuous SSTAs from September to the following August are shown on the top of each column.
Citation: Journal of Climate 35, 12; 10.1175/JCLI-D-21-0892.1
c. The source and stability of key signals
The SSTA-residual EOF analysis (Choi and An 2013; Funk and Hoell 2015), was often performed to explore the key signals out of the leading EOFs, which performs the EOF analysis for the datasets with the three leading EEOF modes removed. In Figs. 6a–c, the first leading SSTA-residual EOF mode for the whole tropics shows the same pattern in the equatorial Pacific and the South Atlantic in Fig. 4n. Although there are strong positive North Atlantic SST anomalies in Figs. 6a–c, which may trigger ENSO through a subtropical teleconnection mechanism with a low-level anticyclonic flow according to previous studies (Ham et al. 2013b; Ham and Kug 2015; Wang et al. 2017), ENSO is still unable to develop in this situation (Figs. 6d–o). This result further confirms that the negative South Atlantic SSTAs and the absence of the equatorial South Pacific SSTAs are the key signals determining the development of ENSO. Specifically, Fig. 4n or Figs. 6a–c is not a predictable mode but a kind of noise mode from the residual data for the subsequent summer, suggesting more attention should be paid to this mode when connecting DJF and JJA SSTAs.
The first mode for the SSTA-residual EOF within the whole tropics in DJF from (a) HadISST, (b) ERSST, and (c) COBESST2, respectively, after removing the top three EEOF modes from SST anomalies data. The lead–lag regression of seasonal SSTAs onto the PC1 of DJF SSTA-residual EOF from (d)–(g) HadISST, (h)–(k) ERSST, and (l)–(o) COBESST2. Hatching denotes statistical significance exceeding the 95% confidence level.
Citation: Journal of Climate 35, 12; 10.1175/JCLI-D-21-0892.1
The first mode for the SSTA-residual EOF within the whole tropics in DJF from (a) HadISST, (b) ERSST, and (c) COBESST2, respectively, after removing the top three EEOF modes from SST anomalies data. The lead–lag regression of seasonal SSTAs onto the PC1 of DJF SSTA-residual EOF from (d)–(g) HadISST, (h)–(k) ERSST, and (l)–(o) COBESST2. Hatching denotes statistical significance exceeding the 95% confidence level.
Citation: Journal of Climate 35, 12; 10.1175/JCLI-D-21-0892.1
The first mode for the SSTA-residual EOF within the whole tropics in DJF from (a) HadISST, (b) ERSST, and (c) COBESST2, respectively, after removing the top three EEOF modes from SST anomalies data. The lead–lag regression of seasonal SSTAs onto the PC1 of DJF SSTA-residual EOF from (d)–(g) HadISST, (h)–(k) ERSST, and (l)–(o) COBESST2. Hatching denotes statistical significance exceeding the 95% confidence level.
Citation: Journal of Climate 35, 12; 10.1175/JCLI-D-21-0892.1
Moreover, it has been well recognized that the regime of ENSO shifts around the 1970s and 2000s (Fedorov and Philander 2000; Wang and An 2001; Fedorov et al. 2020; Hu et al. 2020), exerting remarkable interdecadal changes in global climate systems (Yang and Huang 2021). Besides, ENSO-related interbasin interaction also changed in the 1970s and mid-1990s (Rodríguez-Fonseca et al. 2009; Martin-Rey et al. 2014; Wang et al. 2017), impacting ENSO predictability (Martin-Rey et al. 2015). Similar analyses for different time periods, 1948–75, 1975–2000, and 2000–20, were performed (Figs. S4–S9) to check the stability of ENSO’s evolution patterns and key signals revealed in the whole period of 1948–2014. The interdecadal filtering was not applied to the segmented data due to the limitation of its length. The main conclusions in the Pacific about the ENSO diversity are very robust and independent on the selection of time periods. Furthermore, we also can find some differences in the Atlantic in the analyses using different time periods. The anomaly center shifts from the equatorial Atlantic in the period 1975–2000 to the North Atlantic in the period 2000–20, which is in agreement with the previous studies (Rodríguez-Fonseca et al. 2009; Ding et al. 2012; Ham et al. 2013a; Cai et al. 2019). This result further highlights the role of the tropical Atlantic in the development of ENSO, especially for the decadal variation of ENSO’s evolution (Wang et al. 2017; Yang and Huang 2021).
4. Prediction for summer ENSO with EOF and EEOF
A steady ENSO decaying mode indicates that the SSTAs in boreal summer associated with it can be well predicted, whereas an unsteady developing mode of ENSO brings a great challenge to ENSO prediction. In the EEOF and EOF analyses, EEOF2 was much closer to the JJA EOF1-related SSTA evolution than that of DJF EOF3 (Fig. 4). Therefore, the structured SSTAs from the EEOF pattern in boreal winter and spring could be helpful toward summer SSTA prediction (Zhao et al. 2020). Therefore, we predicted JJA SSTAs based on the decaying and developing mode from the DJF EOF and EEOF analysis.
The hindcasted JJA SSTAs over the Niño-3 (5°S–5°N, 150°–90°W) region using the prediction model introduced in section 2c are compared to the observation in Fig. 7. For the prediction with DJF EOF mode (Fig. 7a), when only the decaying mode with the largest variance is used (the light blue curve), the predicted JJA Niño-3 does not reproduce the amplitude variation of the observed and the temporal correlation coefficient (TCC) with the observation is less than 0.1, indicating no predictive skill for the JJA Niño-3 only considering the decaying mode. In contrast, when the developing mode is also considered (the red curve), the TCC between the predicted and the observed Niño-3 has been dramatically improved to 0.44. Similar results can be seen in the prediction with EEOF mode using the continuous SSTAs in Figs. 7b–d, where the red curve simulates JJA Niño-3 much better than the light blue curve. This result indicates that the developing mode rather than the decaying mode plays a decisive role in the JJA SSTA prediction, which is consistent with the much higher seasonal explained variance for JJA SSTAs by the developing mode (around 34%) than the decaying mode (Table 1).
The time series of the observed JJA Niño-3 index (black) and the predicted JJA Niño-3 index made with “take-one-out” hindcast from (a) the DJF EOF, the EEOF from (b) September to February, (c) September to March, and (d) September to April with the whole tropical SSTAs. Light blue curve shows the prediction with only the decaying mode used, and red curve shows the prediction with both the decaying and developing modes used. (e)–(h) As in (a)–(d), but the Pacific SSTAs were used for prediction. The temporal correlation coefficients of the predictions in the light blue and red curves with the observation in the black curve are shown at the upper right of each subfigure.
Citation: Journal of Climate 35, 12; 10.1175/JCLI-D-21-0892.1
The time series of the observed JJA Niño-3 index (black) and the predicted JJA Niño-3 index made with “take-one-out” hindcast from (a) the DJF EOF, the EEOF from (b) September to February, (c) September to March, and (d) September to April with the whole tropical SSTAs. Light blue curve shows the prediction with only the decaying mode used, and red curve shows the prediction with both the decaying and developing modes used. (e)–(h) As in (a)–(d), but the Pacific SSTAs were used for prediction. The temporal correlation coefficients of the predictions in the light blue and red curves with the observation in the black curve are shown at the upper right of each subfigure.
Citation: Journal of Climate 35, 12; 10.1175/JCLI-D-21-0892.1
The time series of the observed JJA Niño-3 index (black) and the predicted JJA Niño-3 index made with “take-one-out” hindcast from (a) the DJF EOF, the EEOF from (b) September to February, (c) September to March, and (d) September to April with the whole tropical SSTAs. Light blue curve shows the prediction with only the decaying mode used, and red curve shows the prediction with both the decaying and developing modes used. (e)–(h) As in (a)–(d), but the Pacific SSTAs were used for prediction. The temporal correlation coefficients of the predictions in the light blue and red curves with the observation in the black curve are shown at the upper right of each subfigure.
Citation: Journal of Climate 35, 12; 10.1175/JCLI-D-21-0892.1
Overall annual explained variance and JJA seasonal explained variance of EEOF1 and EEOF2 within the whole tropical SSTAs for HadISST, ERSST, and COBESST2. Parentheses indicate results within the Pacific.
In addition to the above findings, since the continuous SSTAs contain more processes of interbasin interactions, the prediction of JJA Niño-3 with the decaying and developing modes from the EEOF mode in Figs. 7b–d is more accurate than Fig. 7a, with the TCC increasing from 0.44 to 0.60 in Fig. 7b. This indicates that the prediction skill of JJA Niño-3 based on the EEOF mode of continuous SSTAs can be improved by 36% compared to the DJF EOF mode of seasonal SSTAs under the same previous data. As the prediction deepens and the lead month decreases (Figs. 7c,d), the prediction skill of JJA Niño-3 keeps improving, especially for the prediction of El Niño and La Niña events. Based on 30 selected ENSO cases with 0.75 standard deviation (13 El Niños, 17 La Niñas), Table 2 further confirms the above findings by the increased correlation coefficient and the reduced root mean squared error between the predicted and observed JJA Niño-3 with these ENSO cases. The prediction of JJA Niño-3 and ENSO cases with the EEOF mode demonstrates a great advantage compared to the DJF EOF mode, and improves greatly as the lead time decreases.
The selected years of 13 El Niños and 17 La Niñas for the period of 1948–2014, and the correlation coefficients and the root-mean-square errors between the observed and the predicted JJA Niño-3 for the ENSO cases. The ENSO cases are defined when the JJA HadISST Niño-3 is greater or below than 0.75 standard deviations. The predicted results are obtained by the model constructed with the decaying and developing modes analyzed from the whole tropical SSTAs.
The prediction is also performed based on the tropical Pacific SSTAs, in which the EOF and EEOF analysis were both used with the tropical Pacific SSTAs. When both the decaying and developing modes are considered, the prediction of JJA Niño-3 with DJF EOF mode is still not skillful in Fig. 7e due to the influence of two key signals, which is consistent with Figs. 4m–p and 5m–p. However, for the EEOF mode in Figs. 7f–h, a high forecasting skill is maintained despite the decrease in TCC scores compared to Figs. 7b–d. Table 3 further shows the TCC between the predicted and the observed JJA Niño-3 using ERSST and COBESST2, in good agreement with Fig. 7. Another prediction is performed that the datasets in the period of 1948–94 are used to construct the model for predicting the SSTAs during 1995–2014 (Fig. S10). Only when the developing mode is included in the model, the prediction for the JJA Niño-3 index is skillful. Moreover, the model including continuous SSTAs has higher prediction skill than the model including only winter SSTAs. These results verify the instability of the ENSO developing mode impacts the summer ENSO development and indicate that considering interocean basin interactions with the EEOF method can reduce uncertainty to some extent and improve the prediction skill of JJA Niño-3.
Correlation of the observed with the predicted JJA Niño-3 using the decaying and developing modes from the DJF EOF, the EEOF from September to February, September to March, and September to April within the whole tropical SSTAs and the Pacific during 1948–2014. The values before and after the slash are results of ERSST and COBESST2 datasets, respectively.
Figure 8 shows the spatial distribution of TCC scores for the JJA SSTAs with DJF EOF mode and EEOF mode using the decaying and developing mode. For comparison, the TCC exceeding 0.5 are also shown with hatching in Figs. 8a–h, and the differences within the tropics and the Pacific are also shown in Figs. 8i–l. The TCC in Figs. 8a–d exhibit high consistency and exceed 0.7 in the North Indian Ocean and the North Atlantic, but a large difference in the Pacific. The TCC scores in the tropical eastern Pacific of the EEOF mode are higher than the DJF mode, and are further improved with shorter lead times. When only the Pacific SSTAs were used in prediction, the TCC in Figs. 8e–h is essentially unchanged in other oceans but weakened in the Pacific (Figs. 8i–j). These results indicate that the stability of the ENSO developing mode directly determines the prediction of JJA SSTAs in the Pacific rather than other oceans.
Correlation coefficients between the observations and the predicted JJA SSTAs using both the ENSO decaying and ENSO developing mode from (a) the DJF EOF, the EEOF from (b) September to February, (c) September to March, and (d) September to April within the whole tropical SSTAs. (e)–(h) As in (a)–(d), but the Pacific SSTAs were used for prediction. (i)–(l) The differences between (a)–(d) and (e)–(h). Hatching denotes the correlation coefficient exceeding 0.5.
Citation: Journal of Climate 35, 12; 10.1175/JCLI-D-21-0892.1
Correlation coefficients between the observations and the predicted JJA SSTAs using both the ENSO decaying and ENSO developing mode from (a) the DJF EOF, the EEOF from (b) September to February, (c) September to March, and (d) September to April within the whole tropical SSTAs. (e)–(h) As in (a)–(d), but the Pacific SSTAs were used for prediction. (i)–(l) The differences between (a)–(d) and (e)–(h). Hatching denotes the correlation coefficient exceeding 0.5.
Citation: Journal of Climate 35, 12; 10.1175/JCLI-D-21-0892.1
Correlation coefficients between the observations and the predicted JJA SSTAs using both the ENSO decaying and ENSO developing mode from (a) the DJF EOF, the EEOF from (b) September to February, (c) September to March, and (d) September to April within the whole tropical SSTAs. (e)–(h) As in (a)–(d), but the Pacific SSTAs were used for prediction. (i)–(l) The differences between (a)–(d) and (e)–(h). Hatching denotes the correlation coefficient exceeding 0.5.
Citation: Journal of Climate 35, 12; 10.1175/JCLI-D-21-0892.1
5. Summary and discussion
In this study, we analyzed the leading modes of the SSTA evolutions and identified the key signals for ENSO diversity from the perspective of interbasin interactions through EEOF and EOF analyses. The EEOF analysis from September to August reveals there coexists a decaying mode and a developing mode in the typical ENSO decaying periods. The decaying mode is steady, whereas the developing mode is unsteady and thus determines the diversity of ENSO evolution during the decaying periods. Using multiple datasets, segmented datasets, and SSTA-residual EOF, we identified and verified that the two most prominent factors related to the diversity of ENSO evolution are concentrated in the South Atlantic and equatorial South Pacific from numerous precursors proposed by previous studies.
Based on the decaying and developing mode extracted from DJF EOF and EEOF analysis, a prediction model for predicting the JJA SSTAs was constructed. We confirmed that the ENSO developing mode, with the small annual explained variance, played a decisive role in ENSO evolution, suggesting that the ENSO developing mode will directly determine the prediction of the JJA SSTAs in the Pacific. Moreover, the forecasting skill of JJA ENSO can be improved when continuous SSTAs and interbasin interactions in the whole tropics are considered, because the negative South Atlantic SSTAs and the symmetric Pacific SSTAs are included in the previous winter. Therefore, it will be beneficial to improve the ENSO prediction skill by detecting the precursory disturbance signals correctly and considering the interaction of three oceans.
The two key factors in the South Atlantic and equatorial South Pacific influencing the diversity of ENSO’s evolution revealed here are well consistent with the mechanisms reported in previous studies. Hua and Su (2020) emphasized that the cold anomalies in the southeastern Pacific can hinder the development of ENSO through local positive feedback. Chikamoto et al. (2020) revealed the critical role of the negative equatorial South Atlantic SSTAs for equatorial Pacific warming through weakening equatorial zonal winds in the Indo-Pacific regions.
It is worth noting that although our prediction for JJA Niño-3 is slightly inferior to the currently ENSO prediction (e.g., Ham et al. 2019; Yan et al. 2020), the TCC can score 0.6 at the 5-month lead prediction by only considering two linear modes of EEOF analysis. This indicates that the prediction result of this minimalist statistical model has the ability to match the most advanced model, which further emphasizes the importance of EEOF mode in capturing the key processes affecting JJA Niño-3 and highlights the decisive role of the previously neglected ENSO developing mode.
Unlike the ENSO developing mode, the ENSO decaying mode plays an essential role for the SSTA evolution in the north Indian Ocean and the North Atlantic rather than the equatorial Pacific. The TCC between the predicted and the observed JJA SSTAs shows higher prediction skill in these two regions and is not affected by the DJF EOF and EEOF analysis (results not shown), consistent with previous evidence that the north Indian Ocean and the North Atlantic are the main response areas for the ENSO decaying phase (e.g., Enfield and Mayer 1997; Chang et al. 2000; Du et al. 2009). Therefore, it will be a fascinating issue worth clarifying the impact of the ENSO decaying mode and the ENSO developing mode in the future, expecting helpful information will be provided in identifying and predicting the ENSO evolution.
The present study highlights the essential linear patterns of the developing and decaying modes first based on the two leading modes of the EEOF analyses, extracting the common signals of El Niño and La Niña rather than the asymmetry signals. Besides, we can find that the EEOF3 representing the spatial asymmetry between eastern Pacific ENSO and central Pacific ENSO contributes considerable variance during ENSO’s evolution (Takahashi et al. 2011; Zhao et al. 2020). This result will facilitate the subsequent study of the characteristics of ENSO phase asymmetry and spatial asymmetry in ENSO’s evolution.
In this study, we just presented the characteristics of the decaying and developing modes obtained from the mathematical EEOF method, and discussed their role in the ENSO decaying paces and in ENSO prediction. In our preliminary study, the combination of the two modes can well describe the fast and slow decaying ENSO events (Chen et al. 2012; Feng et al. 2014; Jiang et al. 2019), or called continuing and emerging ENSOs (Yang and Huang 2021). We are also investigating the physical processes related to these two modes and comparing with the processes in the fast and slow decaying ENSO events revealed in previous studies (Chen et al. 2012; Feng et al. 2014; Jiang et al. 2019). These results will be presented in future to further support that the decaying and developing modes are realistic modes in the tropical Pacific SSTAs.
Acknowledgments.
This work was supported by the National Key R&D Program of China (2019YFA0606703), the National Natural Science Foundation of China (Grant 41975116), and the Youth Innovation Promotion Association of the Chinese Academy of Sciences (Y202025).
Data availability statement.
The HadISST data can be downloaded from https://www.metoffice.gov.uk/hadobs/hadisst/data/download.html. The ERSST.v5 data can be downloaded from https://psl.noaa.gov/data/gridded/data.noaa.ersst.v5.html. The COBESST2 data can be downloaded from https://psl.noaa.gov/data/gridded/data.cobe2.html.
REFERENCES
Cai, W., and Coauthors, 2019: Pantropical climate interactions. Science, 363, eaav4236, https://doi.org/10.1126/science.aav4236.
Cao, Q., Z. Hao, F. Yuan, Z. Su, R. Berndtsson, J. Hao, and T. Nyima, 2017: Impact of ENSO regimes on developing- and decaying-phase precipitation during rainy season in China. Hydrol. Earth Syst. Sci., 21, 5415–5426, https://doi.org/10.5194/hess-21-5415-2017.
Chang, C. P., Y. Zhang, and T. Li, 2000: Interannual and interdecadal variations of the East Asian summer monsoon and tropical Pacific SSTs. Part I: Roles of the subtropical ridge. J. Climate, 13, 4310–4325, https://doi.org/10.1175/1520-0442(2000)013<4310:IAIVOT>2.0.CO;2.
Chen, D., S. E. Zebiak, A. J. Busalacchi, and M. A. Cane, 1995: An improved procedure for El Niño forecasting: Implications for predictability. Science, 269, 1699–1702, https://doi.org/10.1126/science.269.5231.1699.
Chen, D., M. A. Cane, A. Kaplan, S. E. Zebiak, and D. J. Huang, 2004: Predictability of El Niño over the past 148 years. Nature, 428, 733–736, https://doi.org/10.1038/nature02439.
Chen, W., J.-K. Park, B. Dong, R. Lu, and W.-S. Jung, 2012: The relationship between El Niño and the western North Pacific summer climate in a coupled GCM: Role of the transition of El Niño decaying phases. J. Geophys. Res., 117, D12111, https://doi.org/10.1029/2011JD017385.
Chiang, J. C. H., and A. H. Sobel, 2002: Tropical tropospheric temperature variations caused by ENSO and their influence on the remote tropical climate. J. Climate, 15, 2616–2631, https://doi.org/10.1175/1520-0442(2002)015<2616:TTTVCB>2.0.CO;2.
Chikamoto, Y., Z. F. Johnson, S. Y. S. Wang, M. J. McPhaden, and T. Mochizuki, 2020: El Niño–Southern Oscillation evolution modulated by Atlantic forcing. J. Geophys. Res. Oceans, 125, e2020JC016318, https://doi.org/10.1029/2020JC016318.
Choi, J., and S.-I. An, 2013: Quantifying the residual effects of ENSO on low-frequency variability in the tropical Pacific. Int. J. Climatol., 33, 1047–1052, https://doi.org/10.1002/joc.3470.
Deser, C., M. A. Alexander, S. P. Xie, and A. S. Phillips, 2010: Sea surface temperature variability: Patterns and mechanisms. Annu. Rev. Mar. Sci., 2, 115–143, https://doi.org/10.1146/annurev-marine-120408-151453.
Ding, H., N. S. Keenlyside, and M. Latif, 2012: Impact of the equatorial Atlantic on the El Niño–Southern Oscillation. Climate Dyn., 38, 1965–1972, https://doi.org/10.1007/s00382-011-1097-y.
Du, Y., S.-P. Xie, G. Huang, and K. Hu, 2009: Role of air–sea interaction in the long persistence of El Niño–induced North Indian Ocean warming. J. Climate, 22, 2023–2038, https://doi.org/10.1175/2008JCLI2590.1.
Duan, W., and J. Hu, 2015: The initial errors that induce a significant “spring predictability barrier” for El Niño events and their implications for target observation: Results from an earth system model. Climate Dyn., 46, 3599–3615, https://doi.org/10.1007/s00382-015-2789-5.
Duan, W., X. Liu, K. Zhu, and M. Mu, 2009: Exploring the initial errors that cause a significant “spring predictability barrier” for El Niño events. J. Geophys. Res., 114, C04022, https://doi.org/10.1029/2008JC004925.
Enfield, D. B., and D. A. Mayer, 1997: Tropical Atlantic sea surface temperature variability and its relation to El Niño–Southern Oscillation. J. Geophys. Res., 102, 929–945, https://doi.org/10.1029/96JC03296.
Exarchou, E., P. Ortega, B. Rodriguez-Fonseca, T. Losada, I. Polo, and C. Prodhomme, 2021: Impact of equatorial Atlantic variability on ENSO predictive skill. Nat. Commun., 12, 1612, https://doi.org/10.1038/s41467-021-21857-2.
Fedorov, A. V., and S. G. Philander, 2000: Is El Niño changing? Science, 288, 1997–2002, https://doi.org/10.1126/science.288.5473.1997.
Fedorov, A. V., S. Hu, A. T. Wittenberg, A. F. Z. Levine, and C. Deser, 2020: ENSO low-frequency modulation and mean state interactions. El Niño Southern Oscillation in a Changing Climate, M. J. McPhaden, A. Santoso, and W. Cai, Eds., Wiley, 173–198, https://doi.org/10.1002/9781119548164.ch8.
Feng, J., L. Wang, and W. Chen, 2014: How does the East Asian summer monsoon behave in the decaying phase of El Niño during different PDO phases? J. Climate, 27, 2682–2698, https://doi.org/10.1175/JCLI-D-13-00015.1.
Funk, C. C., and A. Hoell, 2015: The leading mode of observed and CMIP5 ENSO-residual sea surface temperatures and associated changes in Indo-Pacific climate. J. Climate, 28, 4309–4329, https://doi.org/10.1175/JCLI-D-14-00334.1.
Ham, Y.-G., and J.-S. Kug, 2015: Role of north tropical Atlantic SST on the ENSO simulated using CMIP3 and CMIP5 models. Climate Dyn., 45, 3103–3117, https://doi.org/10.1007/s00382-015-2527-z.
Ham, Y.-G., J.-S. Kug, and J.-Y. Park, 2013a: Two distinct roles of Atlantic SSTs in ENSO variability: North tropical Atlantic SST and Atlantic Niño. Geophys. Res. Lett., 40, 4012–4017, https://doi.org/10.1002/grl.50729.
Ham, Y.-G., J.-S. Kug, J.-Y. Park, and F.-F. Jin, 2013b: Sea surface temperature in the north tropical Atlantic as a trigger for El Niño/Southern Oscillation events. Nat. Geosci., 6, 112–116, https://doi.org/10.1038/ngeo1686.
Ham, Y.-G., J. H. Kim, and J. J. Luo, 2019: Deep learning for multi-year ENSO forecasts. Nature, 573, 568–572, https://doi.org/10.1038/s41586-019-1559-7.
Ham, Y.-G., H.-J. Lee, H.-S. Jo, S.-G. Lee, W. Cai, and R. R. Rodrigues, 2021: Inter-basin interaction between variability in the South Atlantic Ocean and the El Niño/Southern Oscillation. Geophys. Res. Lett., 48, e2021GL093338, https://doi.org/10.1029/2021GL093338.
Hannachi, A., I. T. Jolliffe, and D. B. Stephenson, 2007: Empirical orthogonal functions and related techniques in atmospheric science: A review. Int. J. Climatol., 27, 1119–1152, https://doi.org/10.1002/joc.1499.
Hirahara, S., M. Ishii, and Y. Fukuda, 2014: Centennial-scale sea surface temperature analysis and its uncertainty. J. Climate, 27, 57–75, https://doi.org/10.1175/JCLI-D-12-00837.1.
Hu, Z. Z., A. Kumar, B. H. Huang, J. H. Zhu, M. L’Heureux, M. J. McPhaden, and J. Y. Yu, 2020: The interdecadal shift of ENSO properties in 1999/2000: A review. J. Climate, 33, 4441–4462, https://doi.org/10.1175/JCLI-D-19-0316.1.
Hua, L., and J. Su, 2020: Southeastern Pacific error leads to failed El Niño forecasts. Geophys. Res. Lett., 47, e2020GL088764, https://doi.org/10.1029/2020GL088764.
Huang, R., and Y. Wu, 1989: The influence of ENSO on the summer climate change in China and its mechanism. Adv. Atmos. Sci., 6, 21–32, https://doi.org/10.1007/BF02656915.
Huang, B., and Coauthors, 2017: Extended reconstructed sea surface temperature, version 5 (ERSSTv5): Upgrades, validations, and intercomparisons. J. Climate, 30, 8179–8205, https://doi.org/10.1175/JCLI-D-16-0836.1.
Jiang, W., G. Huang, P. Huang, R. Wu, K. Hu, and W. Chen, 2019: Northwest Pacific anticyclonic anomalies during post–El Niño summers determined by the pace of El Niño decay. J. Climate, 32, 3487–3503, https://doi.org/10.1175/JCLI-D-18-0793.1.
Jin, F. F., S. T. Kim, and L. Bejarano, 2006: A coupled-stability index for ENSO. Geophys. Res. Lett., 33, L23708, https://doi.org/10.1029/2006GL027221.
Jin, Y. S., Z. Y. Lu, and Z. Y. Liu, 2020: Controls of spring persistence barrier strength in different ENSO regimes and implications for 21st century changes. Geophys. Res. Lett., 47, e2020GL088010, https://doi.org/10.1029/2020GL088010.
Jin, Y. S., Z. Y. Liu, C. F. He, and Y. C. Zhao, 2021: On the formation mechanism of the seasonal persistence barrier. J. Climate, 34, 479–494, https://doi.org/10.1175/JCLI-D-19-0502.1.
Kawamura, R., 1994: A rotated EOF analysis of global sea surface temperature variability with interannual and interdecadal scales. J. Phys. Oceanogr., 24, 707–715, https://doi.org/10.1175/1520-0485(1994)024<0707:AREAOG>2.0.CO;2.
Keenlyside, N. S., H. Ding, and M. Latif, 2013: Potential of equatorial Atlantic variability to enhance El Niño prediction. Geophys. Res. Lett., 40, 2278–2283, https://doi.org/10.1002/grl.50362.
Klein, S. A., B. J. Soden, and N. C. Lau, 1999: Remote sea surface temperature variations during ENSO: Evidence for a tropical atmospheric bridge. J. Climate, 12, 917–932, https://doi.org/10.1175/1520-0442(1999)012<0917:RSSTVD>2.0.CO;2.
Latif, M., and Coauthors, 1994: A review of ENSO prediction studies. Climate Dyn., 9, 167–179, https://doi.org/10.1007/BF00208250.
Li, T., B. Wang, B. Wu, T. Zhou, C.-P. Chang, and R. Zhang, 2018: Theories on formation of an anomalous anticyclone in western North Pacific during El Niño: A review. J. Meteor. Res., 31, 987–1006, https://doi.org/10.1007/s13351-017-7147-6.
Martin-Rey, M., B. Rodriguez-Fonseca, I. Polo, and F. Kucharski, 2014: On the Atlantic-Pacific Niños connection: A multidecadal modulated mode. Climate Dyn., 43, 3163–3178, https://doi.org/10.1007/s00382-014-2305-3.
Martin-Rey, M., B. Rodriguez-Fonseca, and I. Polo, 2015: Atlantic opportunities for ENSO prediction. Geophys. Res. Lett., 42, 6802–6810, https://doi.org/10.1002/2015GL065062.
Polo, I., M. Martin-Rey, B. Rodriguez-Fonseca, F. Kucharski, and C. R. Mechoso, 2014: Processes in the Pacific La Niña onset triggered by the Atlantic Niño. Climate Dyn., 44, 115–131, https://doi.org/10.1007/s00382-014-2354-7.
Rayner, N. A., and Coauthors, 2003: Global analyses of sea surface temperature, sea ice, and night marine air temperature since the late nineteenth century. J. Geophys. Res., 108, 4407, https://doi.org/10.1029/2002JD002670.
Rodríguez-Fonseca, B., I. Polo, J. García-Serrano, T. Losada, E. Mohino, C. R. Mechoso, and F. Kucharski, 2009: Are Atlantic Niños enhancing Pacific ENSO events in recent decades? Geophys. Res. Lett., 36, L20705, https://doi.org/10.1029/2009GL040048.
Rong, X., R. Zhang, and T. Li, 2010: Impacts of Atlantic sea surface temperature anomalies on Indo-East Asian summer monsoon-ENSO relationship. Chin. Sci. Bull., 55, 2458–2468, https://doi.org/10.1007/s11434-010-3098-3.
Shi, H., and B. Wang, 2018: How does the Asian summer precipitation-ENSO relationship change over the past 544 years? Climate Dyn., 52, 4583–4598, https://doi.org/10.1007/s00382-018-4392-z.
Takahashi, K., A. Montecinos, K. Goubanova, and B. Dewitte, 2011: ENSO regimes: Reinterpreting the canonical and modoki El Niño. Geophys. Res. Lett., 38, L10704, https://doi.org/10.1029/2011GL047364.
Tang, H., K. Hu, G. Huang, Y. Wang, and W. Tao, 2022: Intensification and northward extension of northwest Pacific anomalous anticyclone in El Niño decaying mid-summer: An energetic perspective. Climate Dyn., 58, 591–606, https://doi.org/10.1007/s00382-021-05923-5.
Timmermann, A., and Coauthors, 2018: El Niño–Southern Oscillation complexity. Nature, 559, 535–545, https://doi.org/10.1038/s41586-018-0252-6.
Wallace, J. M., E. M. Rasmusson, T. P. Mitchell, V. E. Kousky, E. S. Sarachik, and H. Von Storch, 1998: On the structure and evolution of ENSO-related climate variability in the tropical Pacific: Lessons from TOGA. J. Geophys. Res., 103, 14 241–14 259, https://doi.org/10.1029/97JC02905.
Wang, B., and S. I. An, 2001: Why the properties of El Niño changed during the late 1970s. Geophys. Res. Lett., 28, 3709–3712, https://doi.org/10.1029/2001GL012862.
Wang, B., R. Wu, and X. Fu, 2000: Pacific–East Asian teleconnection: How does ENSO affect East Asian climate? J. Climate, 13, 1517–1536, https://doi.org/10.1175/1520-0442(2000)013<1517:PEATHD>2.0.CO;2.
Wang, B., X. Luo, W. Sun, Y. M. Yang, and J. Liu, 2020: El Niño diversity across boreal spring predictability barrier. Geophys. Res. Lett., 47, e2020GL087354, https://doi.org/10.1029/2020GL087354.
Wang, C., 2019: Three-ocean interactions and climate variability: A review and perspective. Climate Dyn., 53, 5119–5136, https://doi.org/10.1007/s00382-019-04930-x.
Wang, J. Z., and C. Z. Wang, 2021: Joint boost to super El Niño from the Indian and Atlantic oceans. J. Climate, 34, 4937–4954, https://doi.org/10.1175/JCLI-D-20-0710.1.
Wang, L., J. Y. Yu, and H. Paek, 2017: Enhanced biennial variability in the Pacific due to Atlantic capacitor effect. Nat. Commun., 8, 14887, https://doi.org/10.1038/ncomms14887.
Wang, Y., K. Hu, G. Huang, and W. Tao, 2021: Asymmetric impacts of El Niño and La Niña on the Pacific–North American teleconnection pattern: The role of subtropical jet stream. Environ. Res. Lett., 16, 114040, https://doi.org/10.1088/1748-9326/ac31ed.
Weare, B. C., and J. S. Nasstrom, 1982: Examples of extended empirical orthogonal function analyses. Mon. Wea. Rev., 110, 481–485, https://doi.org/10.1175/1520-0493(1982)110<0481:EOEEOF>2.0.CO;2.
Webster, P. J., and S. Yang, 1992: Monsoon and ENSO: Selectively interactive systems. Quart. J. Roy. Meteor. Soc., 118, 877–926, https://doi.org/10.1002/qj.49711850705.
Webster, P. J., and T. N. Palmer, 1997: The past and the future of El Niño. Nature, 390, 562–564, https://doi.org/10.1038/37499.
Wei, K., and Coauthors, 2020: Reflections on the catastrophic 2020 Yangtze River basin flooding in southern China. Innovation, 1, 100038, https://doi.org/10.1016/j.xinn.2020.100038.
Wu, B., T. Zhou, and T. Li, 2009: Seasonally evolving dominant interannual variability modes of East Asian climate. J. Climate, 22, 2992–3005, https://doi.org/10.1175/2008JCLI2710.1.
Wu, R., and B. Wang, 2002: A contrast of the East Asian summer monsoon–ENSO relationship between 1962–77 and 1978–93. J. Climate, 15, 3266–3279, https://doi.org/10.1175/1520-0442(2002)015<3266:ACOTEA>2.0.CO;2.
Wu, R., Z.-Z. Hu, and B. P. Kirtman, 2003: Evolution of ENSO-related rainfall anomalies in East Asia. J. Climate, 16, 3742–3758, https://doi.org/10.1175/1520-0442(2003)016<3742:EOERAI>2.0.CO;2.
Wu, R. G., W. Chen, G. H. Wang, and K. M. Hu, 2014: Relative contribution of ENSO and East Asian winter monsoon to the South China sea SST anomalies during ENSO decaying years. J. Geophys. Res. Atmos., 119, 5046–5064, https://doi.org/10.1002/2013JD021095.
Xie, S.-P., K. Hu, J. Hafner, H. Tokinaga, Y. Du, G. Huang, and T. Sampe, 2009: Indian Ocean capacitor effect on Indo–Western Pacific climate during the summer following El Niño. J. Climate, 22, 730–747, https://doi.org/10.1175/2008JCLI2544.1.
Xie, S.-P., Y. Kosaka, Y. Du, K. Hu, J. S. Chowdary, and G. Huang, 2016: Indo-western Pacific ocean capacitor and coherent climate anomalies in post-ENSO summer: A review. Adv. Atmos. Sci., 33, 411–432, https://doi.org/10.1007/s00376-015-5192-6.
Xu, H., J. J. Xu, C. L. Liu, and N. S. Ou, 2020: Distinguishing characteristics of spring and summer onset El Niño events. J. Climate, 33, 4579–4597, https://doi.org/10.1175/JCLI-D-19-0605.1.
Yan, J. N., L. Mu, L. Z. Wang, R. Ranjan, and A. Y. Zomaya, 2020: Temporal convolutional networks for the advance prediction of ENSO. Sci. Rep., 10, 8055, https://doi.org/10.1038/s41598-020-65070-5.
Yang, J., Q. Liu, S.-P. Xie, Z. Liu, and L. Wu, 2007: Impact of the Indian Ocean SST basin mode on the Asian summer monsoon. Geophys. Res. Lett., 34, L02708, https://doi.org/10.1029/2006GL028571.
Yang, X., and P. Huang, 2021: Restored relationship between ENSO and Indian summer monsoon rainfall around 1999/2000. Innovation, 2, 100102, https://doi.org/10.1016/j.xinn.2021.100102.
Zhang, J., W. Duan, and X. Zhi, 2014: Using CMIP5 model outputs to investigate the initial errors that cause the “spring predictability barrier” for El Niño events. Sci. China Earth Sci., 58, 685–696, https://doi.org/10.1007/s11430-014-4994-1.
Zhang, R., A. Sumi, and M. Kimoto, 1996: Impact of El Niño on the East Asian monsoon: A diagnostic study of the ‘86/87 and ’91/92 events. J. Meteor. Soc. Japan, 74, 49–62, https://doi.org/10.2151/jmsj1965.74.1_49.
Zhang, W., and Coauthors, 2016: Unraveling El Niño’s impact on the East Asian monsoon and Yangtze River summer flooding. Geophys. Res. Lett., 43, 11 375–11 382, https://doi.org/10.1002/2016GL071190.
Zhao, Y., G. Feng, Z. Zheng, D. Zhang, and Z. Jia, 2020: Evolution of tropical interannual sea surface temperature variability and its connection with boreal summer atmospheric circulations. Int. J. Climatol., 40, 2702–2716, https://doi.org/10.1002/joc.6361.
Zhou, Q., M. Mu, and W. Duan, 2019a: The initial condition errors occurring in the Indian Ocean temperature that cause “spring predictability barrier” for El Niño in the Pacific Ocean. J. Geophys. Res. Oceans, 124, 1244–1261, https://doi.org/10.1029/2018JC014403.
Zhou, X., F. Liu, B. Wang, B. Xiang, C. Xing, and H. Wang, 2019b: Different responses of East Asian summer rainfall to El Niño decays. Climate Dyn., 53, 1497–1515, https://doi.org/10.1007/s00382-019-04684-6.