On the Decadal Changes of the Interannual Relationship Between ENSO and Tropical South Atlantic SST in Boreal Summer

This study aims to investigate the underlying mechanisms driving changes in the contemporaneous relationship between El Niño–Southern Oscillation (ENSO) and tropical South Atlantic (TSA) sea surface temperatures (SST) during boreal summer over the past century. Our findings indicate that the negative TSA–ENSO relationship is primarily attributed to persistent ENSO years rather than ENSO transition or development years. This is because of the distinct anomalies observed in both spring ENSO and the tropical North Atlantic (TNA) during ENSO persistent years, whose impacts on summer TSA are of the same sign. Thus, their combined effects lead to a strong TSA anomaly that is opposite to that of ENSO. On a decadal timescale, the frequency of persistent ENSO occurrence can explain the fluctuation of the ENSO–TSA relationship over the past century. These findings enhance our comprehension of the relationship between summer ENSO and the tropical Atlantic SST.


Introduction
The boreal summer (June-July-August, JJA) is the rainy season for densely populated regions in the Northern Hemisphere.During summer, the El Niño-Southern Oscillation (ENSO) interacts with other tropical sea areas, collectively affecting regional climate across the globe.Recent investigations (Kucharski et al., 2007(Kucharski et al., , 2008;;Sabeerali, Ajayamohan, Bangalath, & Chen 2019;Sabeerali, Ajayamohan, & Rao, 2019) on the weakening of the relationship between Indian summer monsoon rainfall (IMSR) and ENSO since 1980 have suggested that the summer ENSO and the simultaneous tropical South Atlantic (TSA) sea surface temperature (SST) have become more negatively correlated since 1970-1980s, resulting in the opposite impact of cold TSA on IMSR compared to El Niño, leading to the weakening of the ISMR-ENSO relationship.However, the reason for the decadal modifications in the ENSO-TSA relationship remains unknown, thereby stimulating the need for this study.10.1029/2023GL104355 2 of 8 cross-basin interaction of SSTs and the different initial states of prior SSTs, which may lead to differences in subsequent atmosphere-ocean interaction.Furthermore, the boreal summer season presents a range of temporal evolution states for the ENSO phenomenon, including persistence, development, and transition.The complex interplay between ENSO and TSA could be contingent upon their respective pre-existing states prior to summer.Therefore, the ENSO-TSA relationship may be influenced by different types of ENSO evolution.For instance, Tokinaga et al. (2019) observed that multiyear ENSO events may produce a stronger Walker circulation anomaly compared to single-year ENSO events.This stronger anomaly may be more conducive to inducing SSTAs in the tropical Atlantic.
The above analysis suggests that examining different types of ENSO evolution is crucial for understanding the ENSO-TSA relationship and its variations.In this study, we analyze the evolution of the observed ENSO-TSA relationship spanning over a century.We find that the negative ENSO-TSA relationship in summer is mainly observed during the years when ENSO persists, and this can be attributed to the distinct anomalies in ENSO and TNA during the spring season of persistent ESNO years.Furthermore, the changes in the frequency of persistent ENSO years at different time periods can largely explain the interdecadal variations in the ENSO-TSA relationship.These results are presented in Section 3 and Section 4 following a brief introduction of the data and methods in Section 2, and a summary is provided in Section 5.

Data and Methods
We use ERSST v5 (2° × 2°, Huang et al., 2017) data set  for SST and NCEP/NCAR reanalysis data set  for atmospheric variables including precipitation, 10-m wind and 500 hPa vertical velocity fields.Composite analysis, regression, simple correlation, and partial correlation analysis were used to investigate summer ENSO-TSA relationship under different ENSO evolution types.
The target region of this study, the TSA, is defined as 10°S-5°N, 30°W-12°E, as illustrated by the green box in Figure 1a.The selection of this spatial range is primarily based on the significant correlation shown in Figure 1a, which encompass the Atlantic Niño region and the region further south.Therefore, we refer to this region as the TSA.Other indices such as the Niño3.4(5°S-5°N, 170°-120°W, cyan box in Figure 1a), the TNA (0°-20°N, 70°W-10°W, purple boxes in Figure 2) and the tropical Indian (TI) Ocean SST index (20°S-20°N, 30°-110°E) are also used.
We employed the criterion of whether the November-December-January (NDJ) averaged Niño3.4 index is greater than 0.5 times the standard deviation to determine the occurrence of ENSO events at the beginning or end of a given year.Subsequently, we classified the years associated with ENSO events into three categories based on their distinct evolution types: ENSO development years (years with non-ENSO events at the beginning but with ENSO events at the end), ENSO transition years (years with ENSO events at both the beginning and end but of opposite sign), and ENSO persistent years (years with ENSO events at both the beginning and end with the same sign).The specific years for each type of ENSO are listed in Table S1 in Supporting Information S1.

Decadal Changes of the ENSO-TSA Relationship During Boreal Summer
First, we examine the evolution of the relationship between summer ENSO and TSA from 1900 to 2022 (Figure 1).It is found that, over the entire period, summer ENSO and TSA shows a general weak negative correlation.However, this correlation shows strong epochal changes.It is not significant before 1970, except for a short period of relatively strong positive correlation in the 1940s-1950s.However, after the 1970s, their correlation is significantly negative, as noted in previous studies (Kucharski et al., 2007;Rodríguez-Fonseca et al., 2009).10.1029/2023GL104355 3 of 8

Summer TSA Anomaly in Different Types of ENSO Evolution Years
As noted in the introduction, an investigation into the cross-basin relationship between ENSO and TSA necessitates a thorough examination of SST evolution for various temporal evolutions of ENSO.Thus, we conducted a composite analysis of the SSTAs from spring to summer (Figure 2) for the three types of ENSO evolution, as introduced in Section 2. Considering that the temporal scope of wind data  is limited in contrast to that of SST, it is appropriate to confine the composite analysis in Figure 2 to their mutual period.This does not affect our conclusion, as the composite map of SST for the entire period 1900-2022 (Figure S1 in Supporting Information S1) has no essential difference from Figure 2.  2d).This feature applies to both El Niño and La Niña (Figure omitted).In contrast, there are no significant SSTAs in TSA for the summer of transitional and developing ENSO years (green boxes in Figures 2e and 2f).This suggests that the negative ENSO-TSA relationship in summer is mainly contributed by persistent ENSO years.
Then, a naturally raised question is why only the ENSO persistent years manifest pronounced TSA anomalies in summer.Tokinaga et al. (2019) proposed that multi-year ENSO can trigger stronger Walker circulation anomalies.However, we seek to provide a more fundamental explanation by examining the differences in the SST field among these three types of years.Considering that the summer ENSO has an equivalent intensity in the three types of events and the interaction of SST between ocean basins always exhibits a seasonal lag in time, the spring SSTA pattern should be our focus of examination.In this regard, we note that during the spring of the persistent ENSO years, there are three regions exhibiting significant differences compared to the spring of the other two types of years (Figures 2a-2c), namely, ENSO (equatorial central-eastern Pacific), tropical Indian (TI) Ocean, and the tropical North Atlantic (TNA).The spring ENSO differences among the three types of years are easy to understand.As for the TI in the spring and summer seasons of the ENSO persistent years, it exhibits a basin-wide SSTA with the same sign as ENSO due to the atmospheric descent anomaly over the TI induced by ENSO (known as the atmospheric bridge mechanism; Klein et al., 1999).Regarding the TNA, Figure 2 demonstrates a significant positive anomaly during the spring of persistent ENSO years, whereas the TNA exhibits a significant negative anomaly during the spring and summer of ENSO transition years and an insignificant anomaly during the ENSO developmental years.The reason is easy to understand.In persistent El Niño years, the warm TNA in spring is forced by the previous winter El Niño (Enfield & Mayer, 1997;Hastenrath et al., 1987;Wang, 2002).
Similarly, the cold spring TNA in the transitional years of El Niño is triggered by the previous winter La Niña.
Regarding the ENSO development years, although cold (warm) TNA anomalies stimulated by atmospheric variability such as the NAO (North Atlantic Oscillation) (Li et al., 2019) can trigger the El Niño (La Niña) development (Ding et al., 2017;Ham, Kug, Park, et al., 2013;Ham, Kug, & Park, 2013;Jiang & Li, 2021), it is worth noting that this is just one of the triggering factors for ENSO, rather than the sole cause.Therefore, the weak negative values in the TNA region are not statistically significant (Figures 2c and 2f).
What are the respective impacts of the three mentioned regions on TSA?In Section 3.3 below, we only present the effects of ENSO and TNA because our study found that the summer TSA anomaly in persistent ENSO years can be attributed to the combined effect of spring ENSO and TNA, while the TI SST can partially counteract the climatic effects of ENSO.The TI effect can be easily understood as the TI warming is a result of the anomalous subsidence caused by El Niño, and thus, the warmed TI would generate negative feedback on its surrounding areas.The convection responses to TI are opposite to that of ENSO, as shown in Figure S2 in Supporting Information S1.Therefore, the TI is not considered to have a promoting effect on the TSA anomaly in the summer of ENSO persistent years.

Respective Impacts of ENSO and TNA on TSA
The impact of the early-stage ENSO in spring on the summer TSA is illustrated in Figures 3a-3d, which show that a positive ENSO event can force negative anomalies in TSA throughout the summer.This can be attributed to the abnormal southeast winds over TSA associated with El Niño-induced Walker Circulation anomaly, which reinforce the background southeast winds.Consequently, this amplifies the heat flux loss within TSA and causes negative TSA anomaly (Chang et al., 1997;Huang, 2004).In light of the substantial impact of spring ENSO on summer TSA and the fact that only the spring season of ENSO persistent years exhibits strong ENSO intensity (Figure 2), we can now comprehend the reason behind the significant TSA anomalies observed exclusively during the summer season of ENSO persistent years (Figures 2d-2f).
The above analysis demonstrates that the intensity of spring ENSO can account for the existence of summer TSA anomalies during years of ENSO persistence, as presented in Figure 2.However, can the spring TNA play a role in this besides spring ENSO?We found that although the correlation coefficient between spring TNA and summer TSA is too weak to pass the significance test (figure omitted), the intensity of spring TNA during ENSO persistent years can largely determine the strength of summer TSA anomaly.This can be seen in the partial correlation table (Table S2 in Supporting Information S1), calculated based on data for ENSO persistent years only.
Compared with the correlation between other factors and TSA, the simple correlation coefficient of spring TNA with Jun TSA is the highest (at −0.60), and the spring TNA is the only factor that has a significant partial correlation with TSA (correlation of −0.43 with Jun TSA).This indicates that for years of ENSO persistence, although spring ENSO provides a background favorable for TSA anomalies, the strength of summer TSA is largely determined by the intensity of spring TNA.The spatial patterns for the impacts of spring TNA during ENSO persistent years are shown in Figures 3e-3h.The TNA and TSA with opposite signs form the AMM pattern (Carton et al., 1996;Chang et al., 2000;Nobre & Shukla, 1996) (Figures 3e and 3f).Regarding the mechanism of the TNA impact on the TSA, previous studies (Chang et al., 1997;Chiang & Vimont, 2004) have attributed it to the WES feedback: the meridional SST gradient decreases the strength of trade winds over the warm TNA, while it enhances the trade winds over the TSA which cool the TSA and further amplify the meridional SST gradient.
It is noteworthy that Figures 3e-3h show the spring TNA mainly affects the TSA in early summer, while its impact on the late summer TSA is weak.What is the reason for this?Previous study (Enfield et al., 1999) raised doubts on whether TNA can affect TSA.Here we found that the formation of the AMM exhibits a noteworthy seasonal dependence.The first mode of EOF for the tropical Atlantic SST in consecutive months from April to August (Figure S3 in Supporting Information S1) demonstrates that the spring warmed TNA produces the strongest TSA anomaly in May and June, while the TSA anomaly decays in late summer.The weak TSA anomaly in early spring after an El Niño event may be due to the two counteracting effects (Chang et al., 2006): the warming effect of El Niño via atmospheric bridge and the cooling effect caused by the warm TNA.On the other 10.1029/2023GL104355 5 of 8 hand, in spring and early summer, the mean Inter-Tropical Convergence Zone (ITCZ) is located on the equator and is prone to generate opposite precipitation and convective anomalies on both sides of the equator (Figure 3i), thereby facilitating the TSA cooling due to the WES mechanism.This aligns with the results of previous studies that the WES mechanism leading to AMM is confined to a narrow region near the equator (Chang et al., 2000;Chiang et al., 2002).However, in the late summer, the ITCZ migrates toward the Northern Hemisphere, departing from the equator (Figures 3k and 3l), consequently leading to the absence of antisymmetric convection anomalies on both sides of the equator.This unfavorable condition inhibits the WES mechanism from sustaining the TSA anomalies.In addition, Foltz and McPhaden (2010) explained the weak relationship between the spring AMM and the summer AZM from the perspective of ocean wave propagation.The combined effects of the aforementioned mechanisms, taking into account their respective strengths, has led to the emergence of late spring and early summer as a critical time window for the influence of TNA on TSA.
To sum up, this section highlights that the negative correlation between ENSO and TSA primarily occurs during the summer of ENSO persistent years, rather than during ENSO transition years or development years.This is attributed to the fact that during the ENSO persistent years, the ENSO has stronger intensity in spring and is

Decadal Changes in the Frequency of Persistent ENSO Events and Its Implication to the Variation of ENSO-TSA Relationship
Given the pivotal role of ENSO persistent years in the summer ENSO-TSA relationship, it is reasonable to infer that the fluctuation in their occurrence frequency may account for the fluctuation in the ENSO-TSA relationship.By examining the 21-yr sliding correlation curve (Figure 1b), the relationship between the summer Niño3.4 and TSA index can be categorized into three distinct phases: 1900-1929 (negative), 1930-1965 (positive), and 1966-2022 (negative).The occurrence rate of persistent ENSO years in each period is indicated in Figure 4.It shows that during the first and third periods when there is a stronger negative ENSO-TSA correlation, the occurrence rates of persistent ENSO years are around 27% and 26%, respectively, roughly twice as high as the second period (14%).Higher (lower) occurrence rates of persistent ENSO events lead to a stronger (weaker) negative correlation between summer ENSO and TSA.This suggests that the shifts in the frequency of persistent ENSO events may contribute to the decadal variation of the ENSO-TSA relationship.
The increasing frequency of persistent ENSO after the 1970s has been previously identified (An & Wang, 2000;McGregor et al., 2014;Wittenberg et al., 2014), but the underlying mechanism is unclear.An and Wang (2000) suggested that the eastward shift of ENSO-induced wind anomalies since the 1970s led to more long-period ENSO events, while McGregor et al. (2014) found that Atlantic warming contributed to the strengthening of the Pacific Walker circulation and the cooling of the eastern Pacific SST, which helped to increase the ENSO period.We suggest that this issue warrants further investigation from an asymmetry perspective of the ENSO temporal evolution, as La Niña is more likely persistent than El Niño (Larkin & Harrison, 2002;Okumura & Deser, 2010).Therefore, the frequency of La Niña and El Niño events may play a role in modulating ENSO periods, and it may be regulated by the Pacific Decadal Oscillation (PDO) or the Atlantic Multidecadal Oscillation (AMO) (Geng et al., 2020;Lin et al., 2018;Newman et al., 2016).
The amplitude of ENSO is commonly considered a factor for the decadal variation of ENSO teleconnection.Thus, we also examined the role of ENSO amplitude in Figure S4 in Supporting Information S1 and found that the sliding standard deviation of ENSO did not exhibit evident three-stage variations.Hence, the explanatory power of ENSO amplitude for the TSA-ENSO relationship changes is much lower than the frequency of persistent ENSO events.Nevertheless, the frequency of ENSO persistent years is not necessarily the sole factor because in the third period, the frequency of ENSO after 1990 remained like before 1990, but the TSA-ENSO relationship seems different.The statistical test for the correlation difference between these two time periods is  1900-1929, 1930-1965, and 1966-2022, based on the sign of sliding correlations.The percentage of persistent ENSO years for each of these periods is indicated in the figure .10.1029/2023GL104355 7 of 8 challenged by the small sample size.Further study may be needed to examine the specific mechanism involving decadal modulations in the complex relationships among ENSO, TNA, and TSA.

Summary
We have examined the decadal variations in the summer ENSO-TSA relationship since 1900 and explored the underlying causes.Considering the unique characteristics of the summer season in the ENSO evolution process, as well as the seasonal lag in the inter-basin interaction, we analyzed the SST evolution from spring to summer for three temporal evolution types of ENSO years.It is found that the negative ENSO-TSA relationship in summer primarily manifests during the ENSO persistent years rather than during the ENSO transition and development years.This is because only in persistent ENSO years that the spring ENSO and TNA anomaly exhibit strong anomaly with the same sign, thereby exerting strong cooperative effects on the summer TSA.Then, we identified a significant relationship between the frequency of ENSO persistence years and the decadal variation of the ENSO-TSA relationship over the past century.The enhanced negative correlation between summer ENSO and TSA since 1970 is attributed to the increase in the frequency of persistent ENSO events.
These findings contribute to a deeper understanding on the summer ENSO-Atlantic relationship.Future studies could focus on the mechanisms behind the variations in the frequency of different types of ENSO evolution.Additionally, it is important to acknowledge that the change in the frequency of persistent ENSO events is not necessarily the sole factor causing changes in the TSA-ENSO relationship.This question remains open, and future investigations involving the complex relationships among ENSO, TNA, and TSA are needed.

•
The El Niño-Southern Oscillationtropical South Atlantic (ENSO-TSA) relationship during boreal summer since 1900 has experienced three stages: strong negative, weak, and resumption • The negative ENSO-TSA correlation is primarily due to the cooperative impacts of spring ENSO and tropical North Atlantic during ENSO persistent years • The changes in the frequency of persistent ENSO years contribute to the decadal changes in the relationship between ENSO and TSA Supporting Information: Supporting Information may be found in the online version of this article.

Figure 1 .
Figure 1.(a) Contemporaneous correlation map of SST with Niño3.4 index (JJA).The black contour lines indicate a significance level of 0.1 (b) Sliding correlation coefficients between Niño3.4 and TSA (JJA) indices.The red, blue, and black lines represent a sliding window of 11, 15, and 21 years, respectively.The black dashed lines denote the critical correlation at the significance of 0.05 for the sample size of 21.The TSA time series is obtained by spatially averaging the data within the green box (TSA) in panel (a).The data used for calculation covers the period of 1900-2022.

Figure 2
Figure 2 clearly shows that during summer (right column), although all three types of ENSO evolution in the central-eastern Pacific exhibit significant ENSO signals, only the persistent ENSO years show TSA anomalies opposite to ENSO (Figures2a and 2d).This feature applies to both El Niño and La Niña (Figureomitted).In contrast, there are no significant SSTAs in TSA for the summer of transitional and developing ENSO years (green boxes in Figures2e and 2f).This suggests that the negative ENSO-TSA relationship in summer is mainly contributed by persistent ENSO years.

Figure 2 .
Figure 2. Composite maps of SSTAs (shading) and 10 m wind (vectors) during boreal spring (a, b, c) and summer (d, e, f) for different types of ENSO evolution.The first, second, and third rows represent ENSO continuation, transition, and development years, respectively.The purple and green boxes denote the TNA and TSA areas, respectively.The green dots denote a significance level of 0.1.Only the wind vectors that pass the significance test (at a level of 0.1) are plotted.These analyses are based on data from the period of 1950-2022.

Figure 3 .
Figure 3. (a-d) Correlation map of SST (shaded) and regression of 10 m wind (vectors) from May to August with the spring (MAM) Niño3.4 index.The green box highlights the TSA region.(e-h) Partial correlation map of SST (shaded) and regression of 10 m wind (vectors) from May to August with the spring TNA index, while excluding the influence of spring ENSO.The black contours in panels (a)-(h) represents the critical correlation coefficient at the significance level of 0.1.(i-l) Regression of the precipitation (shaded) and 500 hPa vertical velocity (contours with positive purple and negative blue) on the spring TNA index, while also excluding the influence of ENSO.The climatological precipitation is also indicated as black contours.Panels (a)-(d) are conducted using data pertaining to the three types of ENSO evolution, while panels (e)-(l) are calculated for ENSO persistent years only.

Figure 4 .
Figure 4. Standardized Niño3.4 index (NDJ average, black line) and 21-year sliding correlation (blue line) between summer Niño3.4 and TSA.The red dot indicates persistent ENSO year, where the Niño3.4index (NDJ) exhibit the same sign at the beginning and end of the year, with values exceeding 0.5 times the standard deviation.The entire time range is divided into three periods: 1900-1929, 1930-1965, and 1966-2022, based  on the sign of sliding correlations.The percentage of persistent ENSO years for each of these periods is indicated in the figure.