Projected ENSO teleconnection on the Southeast Asian climate under global warming

Given the importance of El Niño–Southern oscillation (ENSO) teleconnection on the Southeast Asia (SEA) climate, the ENSO-induced precipitation and near-surface air temperature anomalies over SEA and its twenty sub-regions are compared between historical (1985–2014) and future (2070–2099) simulations using 30 models from Phase 6 of the Coupled Model Intercomparison Project (CMIP6). Future projections suggest that the Philippines, Malay Peninsula, most of the Maritime Continent, and southern Indochina experience reduced (increased) precipitation in the future El Niño (La Niña) summer. Then, during autumn, amplification of ENSO-precipitation teleconnection is projected in the Borneo, Malay Peninsula, and northern Vietnam, raising flood concerns in these sub-regions in future La Niña autumn. During winter, projected ENSO-driven negative anomalies continue intensifying and shifting northeastward, resulting in drier (wetter) conditions for the Philippines and surrounding areas in future El Niño (La Niña). Conversely, a southeastward shift of ENSO-driven precipitation anomalies is projected in the following spring, leading to dampening (an amplification) of teleconnection over the western (eastern) part of SEA. Regarding near-surface air temperature, a ‘land-sea contrast’ pattern is seen, in which intensified ENSO-driven positive (negative) anomalies are projected over land (ocean). At the sub-region scale, robust amplifications in the ENSO teleconnection are mainly observed when only considering the land temperature. The most noticeable future changes are robust amplification of the ENSO-driven positive temperature anomalies in northern Indochina and Myanmar during winter. These sub-regions typically experience a cooler winter, suggesting that wintertime mean temperature there may be much higher under future El Niño conditions. The projected changes in ENSO-driven precipitation and near-surface air temperature anomalies both appear to scale with the radiative forcing, i.e. a higher radiative forcing corresponds to higher teleconnection changes and more sub-regions of SEA experience robust changes. These results suggest that significant ENSO teleconnection changes can be mitigated by minimizing future warming.


Introduction
Southeast Asia (SEA), located in the tropical Asia-Pacific region, comprises islands and the mainland of eleven countries with a combined population exceeding 680 million people (figure 1).Generally characterized by a tropical climate of high temperatures and humidity, SEA experiences abundant precipitation (PR) under the influence of the Asian monsoon (Räsänen et al 2016).However, the region displays significant spatial variability and seasonality in PR and temperature (Chang et al 2005, Juneng et al 2016, Cruz et al 2017).To better understand and address the climate nuances within the region, SEA is divided into 20 sub-regions identified by Juneng et al (2016) under the framework of the SEA regional climate downscaling/coordinated regional climate downscaling experiment-SEA (SEACLID/CORDEX-SEA) project (Tangang et al 2020).Rooted in the tropical Pacific, ENSO exerts a significant influence on the global climate system via teleconnection, leading to changes in weather patterns and extreme events, impacting agriculture, water resources, and socio-ecosystems worldwide (McPhaden et al 2006, Timmermann et al 2018, Yeh et al 2018).Thus, considerable attention has been given to projecting future changes in ENSO and its teleconnections under warming (e.g.Meehl et al 1993, Timmermann 1999, Power et al 2013, Cai et al 2014, Yan et al 2020, Lee et al 2021, Yun et al 2021, McGregor et al 2022).However, there has yet to be a consensus among global climate models (GCMs) regarding a systematic change of ENSO intensity over the 21st century, particularly in the previous phases of coupled model intercomparison project (CMIP) models, i.e.CMIP5 and below.Also, in CMIP5, the challenge of addressing ENSO teleconnection changes at sub-regional scales was noted due to the coarse spatial resolution of GCMs, which hinders the representation of local features such as topography, land-surface feedback, and land use changes.
To overcome the limitations above, the most recent CMIP6 utilizes a new generation of GCMs with improved spatial resolution, model parameterization, and physical processes (Eyring et al 2016).They are driven by newer scenarios based on shared socioeconomic pathways (SSPs) (Eyring et al 2016, O'Neill et al 2017).With the CMIP6, a higher inter-model consensus regarding the ENSO sea surface temperature (SST) amplitudes has been projected in high emission scenarios (Cai et al 2021).Planton et al (2021) indicate that many ENSO-relevant metrics, including the ENSO teleconnection, are noticeably improved in CMIP6 than CMIP5.Recently, Chen et al (2023) utilized CMIP6 to show that, for both land and sea PR, ENSO-driven negative anomalies (drier under El Niño and wetter under La Niña) over the Maritime Continent intensify significantly in future warming.However, their results mainly relate to the Maritime Continent PR under the SSP585 scenario.Thus, here, we aim to investigate further the ENSO teleconnection on the climate variability over the whole SEA and its 20 sub-regions, focusing on how it changes in response to future warming in CMIP6.
This paper is organized as follows.Section 2 briefly describes the study region, CMIP6 data and our methodology.The main results are given in section 3. Finally, section 4 presents conclusions.

Datasets
We analyzed the projected changes in the ENSOdriven PR and near-surface air temperature (TAS) anomalies over the period of 2070-2099, with respect to the historical simulation

Methodology to calculate the ENSO teleconnection
Because the spatial resolution of each CMIP6 GCM is different, all GCM data were interpolated into 1 • × 1 • grid data sets using the bilinear interpolation.Model variables (PR and TAS) in June-July-August(0) [JJA(0)], September-October-November(0) [SON(0)], December(0)-January-February(1) [DJF(0/1)], and March-April-May(1) [MAM(1)] were detrended.Then, the ENSO-driven anomalies were defined as the regression coefficients of these detrended variables onto DJF(0/1) Niño3.4 region SST anomalies for all the historical, SSP126, SSP245 and SSP585 simulations.Finally, the differences in these ENSO-driven anomalies between the future and historical simulations represented the ENSO teleconnection changes under global warming.This process was conducted for each model, and the MME mean was then calculated.A Student's t-test was applied to check the significance of the MME difference between historical and future simulations.An amplification of ENSO teleconnection was identified when positive (negative) ENSO-driven anomalies significantly increase (decrease).In contrast, a dampening of teleconnection was identified when positive (negative) anomalies significantly decrease (increase).

Land-sea difference in the ENSO teleconnection
For 20 sub-regions of SEA, the land-sea difference in the ENSO teleconnection, particularly for the TAS, was investigated by applying land-sea masks to separate the land and sea areas.In the main text, we mostly show the land-only results.The ENSO teleconnections for two other cases, all-surface and seaonly, were provided in the supplements.

Projected changes of ENSO teleconnections on the SEA climate 3.2.1. PR
In the historical simulation, strong ENSO-driven negative PR anomalies initially emerge over the part of the Maritime Continent over the southern hemisphere in the preceding summer JJA(0) (figure 4(a2)), then intensify and extend northward to dominate most SEA areas during SON(0) (figure 4(b2)).The ENSO-PR teleconnection peaks in winter DJF(0/1) (figure 4(c2)) and subsequently weakens or even exhibits sign reversal in the following spring MAM(1) (figure 4(d2)).
Under global warming, distinct disparities of ENSO-PR teleconnection between future projections and historical simulations are observed across all projection scenarios.Notably, the magnitude of the ENSO-PR teleconnection difference scales with the radiative forcing of the SSP scenarios.There is also The intensification and eastward shift induce pronounced ENSO-PR teleconnection changes over SEA, noticeably over the eastern part, i.e. the Philippines and eastern Maritime Continent.However, variations however are evident in space and across different seasons over the region.Specifically, in JJA(0), the ENSO-driven negative PR anomalies over the Maritime Continent are projected to not only intensify and extend further eastward, particularly compared to other seasons, as reported in previous studies (e.g.Chen et al 2023) but also stretches northward, affecting the Philippines, Malay Peninsula, and southern Indochina (figures 4(a2)-( a4)).In SON(0), an increase of negative PR anomalies become apparent across the majority of SEA, but only the Malay Peninsula, western Borneo and some areas in eastern Indochina have robust signals (figures 4(b2)-( b4)).In DJF(0/1), the ENSO-driven negative PR anomalies not only intensify but also expand northeastward relative to historical period, causing the Philippines and the Philippines sea regions exhibit significantly drier (wetter) conditions under future El Niño (La Niña) (figures 4(c2)-(c4)).Meanwhile, in MAM(0), a southeastward shift of ENSO-driven PR anomalies is projected, resulting in a dampening (amplification) of teleconnection over the western (eastern) parts of SEA (figures 4(d2)-(d4)).
It is also suggested that the number of subregions expecting notable changes increases as the SSP scenario radiative forcing increases.In detail, for the SSP126 scenario, no sub-region has a significant difference.However, for the SSP245 scenario, 11 of 20 (55%) sub-regions display substantial changes, primarily in JJA(0) and DJF(0/1).For the SSP-585 scenario, significant ENSO-PR teleconnection changes are found in 17 of 20 regions (85%), spanning all seasons.Furthermore, several sub-regions have both significant amplification and dampening of ENSO-PR teleconnection.The East Philippines (R2), northeastern Borneo (R3), and southeastern Malay Peninsula (R13) are prominently responsive to the future changes in teleconnection.Regions like northwestern and southern Borneo (R4-R5) also show heightened sensitivity.On the other hand, the ENSO-PR teleconnections in Java, Nusa Tenggara, and Timor-Leste (R10-R11) generally remain unchanged in future warming.
Figure 5. Regional ENSO-precipitation (PR) teleconnections (unit: mm/day/ • C of the Niño3.4SST anomalies) on 20 sub-regions of SEA and their projected changes.Significant regional projected differences (at the 95% level) are identified by a colored triangle near the lower x-axis, where the colors again represent the scenario displaying the significant change, while the up (down) triangle symbol indicates an amplification (a damping) of the teleconnection.

TAS
In the historical simulation, weak ENSO-driven positive TAS anomalies (warmer under El Niño and cooler under La Niña) are initially observed over limited areas around the equator in the preceding summer JJA(0) (figure 6(a2)), then extend northward to the Malay Peninsula and the Indochina region during the preceding autumn SON(0) (figure 6(b2)).These positive anomalies continue intensifying and spreading to cover the majority of SEA during the ENSO mature winter DJF(0/1) (figure 6(c2)) and eventually reach their peak in the following spring MAM(1) (figure 6(d2)).
In future projections, similar to what is seen for PR teleconnection, the magnitude of ENSO-TAS teleconnection changes appears to scale with the magnitude of the SSP scenarios radiative forcing.In addition, a pronounced visual resemblance is found in the projected teleconnection changes for each emissions scenario, as indicated by significant spatial correlations among difference maps, ranging from 0.72 to 0.95 (table S2).During the ENSO developing seasons JJA(0) and SON(0), noteworthy amplification of ENSO-driven positive TAS anomalies is only found over some limited regions such as southern Borneo and northern Sumatra, mainly during SON(0) (figures 6(b3)-( b5)).In contrast, robust intensification of ENSO-driven negative TAS anomalies (cooler under El Niño and warmer under La Niña) are observed in JJA(0) to the east of the Philippine Sea and parts of the Maritime Continent over the southern hemisphere (figures 6(a3)-( a5)).In the ENSO mature winter DJF(0/1), amplification of positive anomalies is significant over most Indochina and Myanmar regions and a small part of Borneo (figures 6(c3)-(c5)).Meanwhile, a robust amplification of negative TAS anomalies is projected over the Philippine Sea.Eventually, during the ENSO decaying spring MAM(1), a significant amplification of ENSO-driven positive TAS anomalies is projected over northern Sumatra, eastern Philippines, and eastern Maritime Continent (figures 6(d3)-( d5)).
The sub-regional analysis confirms the above findings (figure 7).Furthermore, noticeable changes are mainly projected in the scenario with high radiative forcing SSP585, occurring in 10 of 20 (50%) sub-regions of SEA.Most of these sub-regions (9 of 10) expect an amplification of the ENSOdriven positive TAS anomalies in SON(0), DJF(0/1), and MAM(1).Remarkably, northern Indochina and Myanmar (R18-R20), known for their cooler winter, exhibit a robust amplification in winter, DJF(0/1) (figure 7(c)).Given the future warmer world, this suggests that these regions will experience an even warmer mean state during the future El Niño winters.Interestingly, we see a 'land-sea contrast' pattern that amplifies ENSO-driven positive (negative) TAS anomalies that are primarily projected over land (ocean) areas.It could be related to the 'land-warmerthan-ocean' pattern of the CMIP6 projections, in which the future warming will be more significant over land than the ocean (Endo et al 2018, Wang et al 2020b) due to the Greenhouse gas-induced horizontally differential warming (Liu et al 2009, Wang et al 2020b).It has been suggested that enhanced land-sea thermal contrast contributes to the enhancement of Asian monsoon circulation and PR (e.g.Endo et al 2018, Wang et al 2020b).Our results also show that robust intensification primarily applies to the ENSOinduced TAS anomalies over land areas since very few sub-regions exhibit robust intensification of sea-only TAS anomalies (figure S20).
Focusing on land areas alone, it is seen that ENSO-TAS teleconnection changes are consistent with ENSO-PR teleconnection, as presented through their significant spatial correlations (table 1), except in DJF (0/1).Physically, fewer PR falls on land might lead to higher TAS via altering soil moisture, reflecting through their high negative correlation coefficients.Note that winter is a dry season in the northern part of SEA, whereas it is a rainy season in the Maritime Continent.Thus, if we only consider the Maritime Continent (12 • S-7.5 • N, 90 • E-145 • E), the ENSO-TAS and ENSO-PR teleconnections also have high negative spatial correlations in DJF(0/1), which are −0.32,−0.37, and −0.40 with SSP126, SSP245 and SSP585 scenarios, respectively.Meanwhile, winter TAS in northern SEA is modulated by the East Asian winter monsoon (EAWM).During El Niño (La Niña) mature winter, the Philippine Sea anomalous anticyclone (cyclone) forms and its related meridional anomalies weaken (enhance) the EAWM, leading to warmer (cooler) winter in SEA (Wang et al 2000).Earlier studies (e.g.Wang et al 2013) have reported an amplification in the ENSO-EAWM relationship at the end of the 21st century due to global warming.This projection thus provides an explanation for the intensified ENSO-TAS teleconnection in northern Indochina and Myanmar during DJF(0/1).

Summaries and conclusions
Given the importance of ENSO teleconnection on the SEA climate, particularly PR and TAS, it is crucial to understand how they change under the global warming.The most recent CMIP6 offers an opportunity to explore this change across 20 sub-regions of SEA based on prescribed radiative forcing scenarios and the latest generation of GCM.
Regarding PR, an amplification and eastward shift of tropical ENSO-driven anomalies in response to global warming has been robustly projected.However, previous studies mainly looked at the boreal winter ENSO mature season.Here, we confirm these changes can be evident across all four seasons: JJA(0), SON(0), DJF(0/1), and MAM(1).More complex variations in the ENSO-PR teleconnection could be seen over sub-regions of SEA.First, it is shown that the Philippines, Malay Peninsula, most of the Maritime Continent, and southern Indochina will experience drier (wetter) conditions in the future El Niño (La Niña) summer.Then, during autumn, amplification of ENSO-PR teleconnection is projected in the Borneo, Malay Peninsula, and northern Vietnam, raising flood concerns in these subregions in future La Niña autumn.In winter, projected ENSO-driven negative PR anomalies continue intensifying and shifting northeastward, resulting in drier (wetter) conditions for the Philippines and surrounding areas under future El Niño (La Niña).In contrast, a southeastward shift of ENSO-driven PR anomalies is projected during the following spring, leading to a dampening (an amplification) of teleconnection over the western (eastern) part of SEA.
In terms of TAS, a 'land-sea contrast' pattern that intensified ENSO-driven positive (negative) TAS anomalies are mostly projected over land (ocean).Also, at the sub-regional scales, robust intensifications in the ENSO teleconnection are primarily projected considering the land-only TAS.The most prominent change in the future ENSO-TAS teleconnection includes robust amplifications of positive TAS anomalies over northern Indochina and Myanmar in the winter DJF(0/1).This implies that these sub-regions where winter is characterized by cooler conditions will experience much higher wintertime mean temperature under future El Niño conditions.
Note that even with the multi-model mean, there are still notable discrepancies, especially in the case of PR (figure 2), between the simulated and observed ENSO-driven anomalies at the subregional scale.Uncertainties thus persist in projecting ENSO teleconnection on the SEA climate.More efforts should be made to improve the credibility of these future projections.Nevertheless, it is seen that the changes in ENSO-PR and ENSO-TAS teleconnections both appear to scale with the radiative forcing of the SSP scenarios, i.e. the higher radiative forcing leads to the higher teleconnection changes and more subregions display significant changes.Also, no robust change is projected for the low-emission SSP126 scenario.This implies that minimizing future warming could mitigate considerable changes in the ENSO teleconnection.

Figure 1 .
Figure 1.The Southeast Asia (SEA) domain and its 20 sub-regions used for the assessment (delineated by black boxes).Topography (shaded) is obtained from the Global 30 arc-second elevation (GTOPO30) data set.
The observed ENSO-PR teleconnection exhibits a zonal dipole pattern, with positive anomalies over the equatorial central-eastern Pacific and negative anomalies over the tropical western Pacific (figures S1-S4).The MME can generally reproduce the dipole-shaped patterns across all four seasons.The simulated spatiotemporal patterns agree well with documented ENSO-PR relationship in previous studies (e.g.Timmermann et al 2018, Brown et al 2020, Wang et al 2020a, Chen et al 2023).However, the simulated ENSO-driven positive anomalies exhibit a slight westward shift relative to observations, a known feature of coupled GCMs related to mean SST biases (e.g.Bellenger et al 2014, Brown et al 2020).

Figure 2 .
Figure 2. Regional ENSO-precipitation (PR) teleconnections (unit: mm/day/ • C of the Niño3.4SST anomalies) on 20 sub-regions of SEA during the historical period (1985-2014).Observations (obs; black curve) and the multi-model ensemble mean of 30 CMIP6 models for the historical simulation (MME; red curve) are shown.The shading indicates the 95% model range.

Table 1 .
Spatial correlations over the SEA region between the projected changes in the ENSO-precipitation (PR) teleconnection and projected changes in the ENSO-near-surface air temperature (TAS) teleconnection under different SSP scenarios.The star ( * ) denotes the correlation coefficient less than or equal to −0.4.