Stronger and prolonged El Niño-Southern Oscillation in the Early Eocene warmth

Abhik, S.; Dommenget, Dietmar; McGregor, Shayne; Hutchinson, David K.; Steinig, Sebastian; Zhu, Jiang; Capitanio, Fabio A.; Lunt, Daniel J.; Niezgodzki, Igor; Knorr, Gregor; Chan, Wing-Le; Abe-Ouchi, Ayako

doi:10.1038/s41467-025-59263-7

Download PDF

Article
Open access
Published: 30 April 2025

Stronger and prolonged El Niño-Southern Oscillation in the Early Eocene warmth

Nature Communications volume 16, Article number: 4053 (2025) Cite this article

15k Accesses
83 Altmetric
Metrics details

Subjects

Abstract

The El Niño Southern Oscillation (ENSO) during the Early Eocene Climatic Optimum (EECO, 56–48 million years ago) is investigated using a multi-model ensemble of deep-time climate simulations. We reveal that ENSO sea surface temperature variability during the EECO had significantly longer periodicity and stronger amplitude than present-day conditions. These changes are attributed to intensified ocean-atmosphere feedback processes and enhanced in-phase tropical inter-basin interactions within a broader ocean basin compared to the present-day. Sensitivity experiments in coupled ocean-atmosphere models suggest that tectonic changes, particularly the expansion of the tropical ocean basin, play a dominant role in amplifying ENSO variability and extending its periodicity, while stronger inter-basin connections further enhance ENSO amplitude. Elevated atmospheric CO₂ levels, though driving substantial mean-state changes, partially offset the tectonic influence on ENSO variability by modifying feedback processes. These findings underscore the role of tropical ocean basin geometry and atmospheric CO₂ levels in shaping ENSO variability, offering insights into past climate dynamics and implications for future projections under sustained global warming.

Modulation of late Pleistocene ENSO strength by the tropical Pacific thermocline

Article Open access 23 October 2020

Changing El Niño–Southern Oscillation in a warming climate

Article 17 August 2021

Future high-resolution El Niño/Southern Oscillation dynamics

Article 26 August 2021

Introduction

The year-to-year ocean-atmosphere coupled variability in the tropical Pacific^1,2, known as the El Niño Southern Oscillation (ENSO), remains a focal point for climate researchers due to its significant global impact^3,4. ENSO is a natural climate cycle with a 3–8-year periodicity⁵, characterized by the periodic warming (El Niño) and cooling (La Niña) of sea surface temperatures (SST) in the central and eastern tropical Pacific Ocean. Understanding how ENSO may evolve under the influence of enhanced greenhouse gas emissions and global warming is largely elusive as climate models exhibit larger inter-model discrepancy on future SST variability in the eastern Pacific^6,7,8,9. This inconsistency challenges predicting future ENSO behavior and its global teleconnections^10,11. To address this uncertainty, researchers have expanded their focus beyond instrumental records and turned to past warm periods^12,13,14, such as the Early Eocene Climatic Optimum (EECO) epoch¹⁵, for insights into potential changes in ENSO characteristics.

During El Niño, the warm SST anomalies (SSTA) in the eastern Pacific are associated with westerly zonal wind anomalies, resulting in divergence of Sverdrup transport, shoaling of the equatorial thermocline, and discharge of oceanic heat content^16,17. Conversely, during La Niña events, a similar feedback process occurs but with the opposite sign. An altered tropical interbasin interaction may also favor the enhanced ENSO variability through strengthening the Pacific trade winds, increasing stratification, and coupling between SST and thermocline¹⁸. For instance, wind forcing driven by Indian Ocean warming influences surface wind anomalies in the Pacific, aiding in the decay of ENSO events¹⁹. The asymmetric meridional shift of surface wind anomalies is attributed to the early termination of El Niño^20,21. On the other hand, the importance of tropical North Atlantic and equatorial Atlantic on the ENSO variability is highlighted in recent studies^22,23,24. A comprehensive understanding of ENSO variability during past warm periods like the EECO is crucial for interpreting its evolution in future warm climate and impacts on global climate, including rainfall distribution^25,26, storm activity^27,28, teleconnections^11,29, and ecosystems^30,31,32.

The EECO (56–48 million years ago, Ma) represents one of the warmest epochs in the Cenozoic era (66 Ma–present)^33,34, with atmospheric CO₂ concentrations of 1400 ± 470 parts per million by volume (ppmv)^35,36. The EECO was marked by significantly warmer global mean temperatures, with estimates suggesting a 9º–15 °C warming over present-day values³⁷. The extratropical and polar temperatures over many regions were above freezing year-round without polar ice sheets. Unlike the present, where the tropical oceans are relatively isolated by continental landmasses, majority of the Paleogene period (66–23.03 Ma), including EECO featured extensive connections between the major ocean basins and generally lower elevations of the tropical land masses (Fig. S1b). The effective width of the tropical Pacific Ocean during the EECO was about 1.5 times wider than the present ~150° width, with the western boundary extending as far west as the east coast of the Indian subcontinent (~80 °E). Additionally, the South American continent was positioned ∼20° further eastward, and the Panama Gateway remained open. With a weaker meridional temperature gradient (Fig. S1c), a closed Bering Strait³⁸, and the absence of the Antarctic Circumpolar Current³⁹, the slow oceanic teleconnections between high latitudes and the tropics^40,41 were likely modified during the EECO. These factors may have influenced ocean circulation patterns, altering heat transport and affecting the background state of ENSO dynamics. Huber and Caballero¹⁵ suggested the existence of stronger ENSO variability in the EECO, yet the amplitude of the Walker circulation was comparable to present-day. While some analysis of fossil records⁴² suggested ENSO periodicity during the EECO was similar to present-day values, other proxies¹² indicated a slightly longer periodicity of ENSO during the EECO (~8-year), highlighting some inconsistency in the proxy-based results regarding ENSO periodicity during this warm period. The lack of observational data from the EECO epoch necessitates using climate models and proxy records to reconstruct past climate states and understand their implications for ENSO variability.

Previous studies on ENSO theory examined how changes in ocean basin size could impact the period and strength of the ENSO^{43,44,45,46,47}. Based on simplified climate models, they suggested that ocean dynamics in a wider basin would likely result in a longer ENSO periodicity. The wider tropical ocean basin may also exhibit intensified Bjerknes feedback—an ocean-atmosphere interaction between trade wind and zonal SST gradient^48,49, leading to stronger instability in the tropical Pacific due to modified east-west SST gradient and Walker circulation^50,51. However, the growth of the SST is mitigated by enhanced convection in the tropical Pacific, resulting in alteration in the net surface heat flux anomalies^52,53. A recent coupled model study by Dommenget and Hutchinson⁵⁴, hereafter DH24, investigated ENSO variability using an idealized setup with different tropical ocean basin widths and interbasin interactions. They indicated that while ENSO strength and period varied nonlinearly with ocean basin width, the simulated ENSO characteristics were more significantly influenced by changes in tropical inter-basin interactions rather than by the variations in ocean basin width alone. However, these idealized results have yet to be examined in the context of real-world scenarios, and the EECO epoch provides an opportunity to explore how these dynamics played out in a past warm climate with different ocean basin width relative to the present.

Recent advancements in deep-time climate modeling, such as the DeepMIP-Eocene ensemble⁵⁵, have enabled us to simulate past warm periods with greater fidelity. This paleoclimate modeling framework uses a standard set of boundary conditions and experimental design in a group of Coupled Model Intercomparison Project (CMIP) Phase 3–6 class models⁵⁶. A set of simulations utilizing non-CO₂ (e.g., paleogeography, continental ice-sheet, orbital configuration) boundary conditions at preindustrial atmospheric CO₂ level (~280 ppm, 1xCO₂) and under enhanced CO₂ forcing (3xCO₂) are available. These simulations allow us to understand the impact of tectonic changes and elevated CO₂ concentrations on ENSO, respectively. Previous studies explored different modes of tropical climate variability, such as West African⁵⁷ and Asian monsoons⁵⁸ in EECO using DeepMIP-Eocene simulations. They suggested that tectonically constrained continental movement plays a major role in altering the tropical climate variability, while enhanced atmospheric CO₂ level has a lesser influence.

In this study, we focus on understanding ENSO behavior during the EECO warmth using six DeepMIP-Eocene models (see Table 1 for the models and their specifications), selected based on the availability of pre-industrial control (piControl), EECO simulations with 3xCO₂, and a sensitivity experiment (1xCO₂). To mitigate individual model biases and enhance the reliability of the simulations, a multi-model ensemble mean (MMM) approach is employed, which effectively improves the robustness of simulations^59,60. Additionally, to understand the role of tectonic forcing on ENSO, we conduct a sensitivity experiment in the ocean-atmosphere coupled Geophysical Fluid Dynamics Laboratory (GFDL) paleoclimate model by altering the width of the tropical ocean basin in the EECO. This comprehensive analysis aims to develop theories and conceptual models that accurately represent tropical ocean-atmospheric interactions and provide a framework for understanding ENSO’s evolution in warmer climate. By advancing our understanding of ENSO dynamics in warmer EECO, the study aims to refine projections of ENSO’s evolution under future climate scenarios.

Table 1 Summary of the DeepMIP-Eocene models used in this study

Full size table

Results

Strength and periodicity of ENSO

The maximum SST variability associated with ENSO, as illustrated by the standard deviation of monthly-mean SSTA, is typically observed in the equatorial eastern Pacific region (Fig. 1a). During positive phase of ENSO (El Niño), warmer SSTA in the eastern equatorial Pacific shift atmospheric convection toward the central Pacific, weakening the easterly trade winds^61,62, which in turn modulate equatorial upwelling patterns and further influence SSTA^63,64. The negative ENSO phase (La Niña) exhibits a similar but opposite response. Therefore, the amplitude of SSTA standard deviation is indicative of the strength of ENSO variability. The observed SSTA standard deviation in the eastern equatorial Pacific is broadly captured in the piControl simulations. However, some models tend to overestimate (e.g., CESM, COSMOS) or underestimate (e.g., MIROC) the observed variability across the simulations (Fig. S2). The MMM generally exhibits more realistic SSTA standard deviation (Fig. 1b) compared to individual piControl simulations, albeit with slightly amplified (reduced) variability in the western Pacific region (along the coasts of Ecuador and Peru in the eastern Pacific).

**Fig. 1: El Niño–Southern Oscillation variability and periodicity.**

The SSTA standard deviation in the equatorial eastern Pacific notably intensified in both EECO simulations (Fig. 1c, d). This increase in SSTA standard deviation is generally consistent across most models, albeit the change is relatively minor in the MIROC and opposite in the COSMOS. However, the MMM SSTA standard deviation in both EECO simulations significantly differs at the 5% level across most tropical ocean basins based on the F-test with the null hypothesis of identical variance in piControl and EECO simulations. Between two EECO simulations, we find the maximum SSTA standard deviation in the 1xCO₂ experiment, which produces only a modest increase in tropical mean SST (27.71 ± 0.93 °C) relative to piControl (25.79 ± 0.83 °C). The tectonic changes in the 1xCO₂ experiment also lead to notably warmer high-latitude surface temperatures (Fig. S1c). In contrast, the tropical mean SST is significantly higher (33.29 ± 0.88 °C) in the 3xCO₂ simulation (Fig. S3), accompanied by substantially warmer high latitudes. The positive and negative ocean-atmospheric feedback components, which are crucial for ENSO dynamics, are closely intertwined through the mean SST in the western central equatorial Pacific⁶⁵. The tropical mean state difference is also associated with an increase in the zonal gradient of mean SST between the western and eastern Pacific (see Fig. S3 for these two regions). The east–west SST gradient is 0.63 ± 0.81 °C higher in the 1xCO₂ experiment relative to piControl, while the 3xCO₂ simulation produces a zonal gradient of mean SST similar to the piControl despite the expanded warm pool across the warmer Pacific. Typically, models with a stronger east-west SST gradient in the equatorial Pacific store more heat in the subsurface ocean⁶⁶. As a result, these models tend to become dynamically unstable and produce a stronger ENSO amplitude⁶⁷. The zonal SST gradient controls the Walker circulation, which subsequently regulates the atmospheric response to SSTA^65,68.

To estimate the change in ENSO variability, we compare the SSTA standard deviation over the Niño3.4 region (5°S–5°N, 170°E–120°W) between piControl and both EECO simulations (Fig. 1e). It indicates that SSTA standard deviation in the EECO increases by ~1.5 times in the 1xCO₂ experiment relative to the piControl, suggesting that the tectonic changes in the geometry of the tropical ocean basins and continental orientation have a substantial impact on ENSO strength. The amplification of ENSO variability in EECO is largely consistent across most models. In contrast, the 3xCO₂ simulation shows only a ~1.25 times increase in SST standard deviation relative to the piControl, indicating that the warming effect acts to reduce the strength of ENSO variability.

In addition to assessing ENSO variability, the periodicity of simulated ENSO is compared in Fig. 1f. The power spectra of observed and MMM piControl Niño3.4 SSTA time series have a spectral peak corresponding to a periodicity of ~36-month (Fig. S4). While models exhibit slightly different peak frequencies in the piControl, the peak frequency is shifted towards a lower frequency (>5 years) in the EECO simulations across most models. Fig. 1f indicates that the change in spectral characteristics in EECO is outside the 2σ limits, thus, the frequency shift is significant at a 5% level. The low-frequency shifting in ENSO periodicity in EECO simulations is also noted in Niño3.4 averaged SSTA autocorrelation (Fig. 1g). The presence of stronger ENSO with a longer periodicity in the EECO geography, characterized by more open connections between the three tropical ocean basins, aligns with the findings of DH24, which indicated that tropical basin interactions can significantly increase ENSO variability and periodicity.

To investigate the primary driver behind the enhanced ENSO variability during the EECO, we apply the Recharge Oscillator model^16,69 using Niño3 (150°W-90°W, 5°S-5°N) SSTA and equatorial Pacific (130°E-80°W, 5°S-5°N) thermocline depth anomalies (h_a, estimated based on the maximum gradient in the ocean temperature profile) from piControl, 1xCO₂, and 3xCO₂ simulations. The resulting model parameters for the three different Recharge Oscillator model fits are shown in Table.ST1 and changes in the model parameters are examined by comparing the 1xCO₂ simulations to the piControl, and the 3xCO₂ simulations to the corresponding 1xCO₂ experiments. This assessment provides insights into the key processes driving SSTA stadard deviation, denoted as std(SSTA), under different climate conditions. In this framework, a₁₁ and a₂₂ represent the growth (or damping) rates of SSTA and h_a, while a₁₂ and a₂₁ describe the coupling between these variables. The residuals from the linear regression are interpreted as stochastic noise forcing on SSTA and h_a. Since these parameters are directly linked to std(SSTA), a positive change indicates enhanced ENSO growth, while a negative change corresponds to a damped regime. Fig. 1h shows that the SSTA growth rate (a₁₁) term is the dominant contributor to the std(SSTA) changes in the 1xCO₂ experiment, accounting for ~60% of the std(SSTA) change, which increases to ~90% when combined with noise forcing. A similar contribution pattern is noted in the transition from the 1xCO₂ to 3xCO₂ simulations, with a₁₁ contributing ~70% of the SSTA standard deviation changes. The a₁₁ term incorporates both the positive Bjerknes feedback and the negative atmospheric heat flux damping associated with ENSO variability. Based on these findings, we first examine the positive and negative ocean-atmosphere feedback mechanisms influencing ENSO, then explore tropical inter-basin interactions, and finally assess the impact of tropical Pacific Ocean basin width.

Ocean-atmosphere feedback

The positive feedback between zonal SST gradient across the equatorial Pacific, surface winds, and upper ocean warm water volume leads to the development of ENSO¹. The enhanced ENSO variability in the EECO likely results in an intensification of SSTA through the Bjerknes feedback. This positive ocean-atmospheric feedback amplifies ENSO instability through three key elements: (1) the influence of eastern Pacific SSTA on surface winds in the western Pacific, (2) the impact of western Pacific surface winds on thermocline anomalies, and (3) the adjustment of local SSTA by thermocline anomalies⁷⁰. On the other hand, SST warming increases evaporation and cloudiness over the tropical Pacific, leading to thermal damping that stabilizes ENSO by modulating the surface heat flux^68,71,72. Figure 2 illustrates three components of the positive Bjerknes feedback and negative thermal damping feedback using observations, piControl simulations, and EECO simulations. The first element of the positive feedback is analyzed by regressing surface wind stress onto Niño3 (5°S-5°N, 150°E-90°W) SST indices. In observations, a 1 °C change in SSTA in the eastern Pacific is associated with a maximum westerly wind stress of 1.5 × 10⁻² Pa in the central Pacific (Fig. 2a). The relationship between SST and zonal wind stress is weaker in the piControl simulation, with simulated westerly wind stress extending into the western Pacific (Fig. 2b, see Fig. S5 for the individual models). The wind–SST feedback is also underestimated in most current-generation climate models^73,74,75. In contrast, this relationship is found to be stronger in both EECO simulations, with enhanced westerly wind stress response in the central Pacific (Fig. 2c, d). All models, including those that poorly reproduced the SST-zonal wind stress feedback in the piControl simulation, show enhanced feedback in the EECO simulations. The regression of Niño4 zonal wind stress onto Niño3 SSTA exhibits ~20% amplitude increase in the 1xCO₂ experiment relative to piControl, and another ~20% increase in the 3xCO₂ simulation relative to the 1xCO₂ experiment. The stronger equatorial zonal wind stress intensifies ENSO variability by amplifying the east-west thermocline slope⁶⁷. Additionally, the zonal wind stress response is found to be more symmetric around the equator, and positive zonal wind stress anomalies shift eastward to the central Pacific in the EECO. The equatorially symmetric zonal wind stress likely contributes to the development and maintenance of low-frequency ENSO variability^21,76, while its eastward shift amplifies thermocline and SSTA in the eastern Pacific through the positive wind–Kelvin wave–SST feedback⁷⁷, ultimately increasing ENSO amplitude and periodicity.

**Fig. 2: The positive and negative feedback mechanisms in El Niño–Southern Oscillation.**

The second element of Bjerknes feedback is examined by regressing h_a onto Niño4 (5°S–5°N, 160°E–150°W) wind stress. The observed central Pacific zonal wind stress anomalies are associated with a deeper thermocline over the eastern Pacific and a shallower thermocline in the western Pacific (Fig. 2e), leading to a dipole pattern in the equatorial Pacific. In the piControl simulation, this relationship is weaker than observed, particularly in the eastern Pacific (Fig. 2f). All the individual models underestimate the east-west thermocline difference in piControl simulation, while CESM produces a strikingly different thermocline response (Fig. S6). As a result, the east-west thermocline difference is underestimated in the MMM. However, it still produces a shallower thermocline in the western Pacific. In contrast, both EECO simulations (Fig. 2g, h) exhibit a stronger thermocline response across the Indo-Pacific region compared to piControl. The east-west thermocline difference increases by ~2.7 times in the 1xCO₂ experiment relative to piControl. This amplification of the thermocline response slightly reduced in the 3xCO₂ simulation. In both EECO simulations, the shallower thermocline in the western Pacific extends zonally into the central Pacific and even reaches the central Indian Ocean. This result is largely consistent across the individual models. The intensified thermocline response in the EECO simulations suggests that the ocean-atmospheric feedbacks during this epoch were more robust, leading to the amplified and low-frequency ENSO events in the EECO.

The third component of Bjerknes feedback is estimated by regressing h_a onto local SSTA (Fig. 2i–l). This positive feedback is observed to be the strongest in the equatorial eastern Pacific but appears to be weak or non-existent in other ocean basins. The thermocline response to the eastern Pacific SSTA is weak in piControl simulation, suggesting weaker than observed oceanic feedback in the simulation (see Fig. S7 for individual models). Both EECO simulations produce intensified h_a-SSTA feedback with the strongest response in the 3xCO₂ simulation. Interestingly, this enhancement is consistent across the models, with the positive SSTA over the eastern Pacific extending towards the central Pacific in the 3xCO₂ simulation. Despite the stronger oceanic feedback in the 3xCO₂ simulation, the ENSO variability is slightly weaker than the 1xCO₂ experiment, suggesting that negative ocean-atmospheric feedback processes might compensate for the enhanced Bjerknes feedback in the 3xCO₂ simulation.

To elucidate the negative thermal damping feedback, we regressed the surface net heat flux anomalies onto the Niño3 SSTA (Fig. 2m–p). During El Niño, the eastward shift of convection weakens the upward net surface heat flux anomalies, which are greatest in the central and eastern equatorial Pacific. The negative feedback is weaker than observed in the piControl simulation, as some models underestimate the weakening of the net heat flux anomalies in the central Pacific (Fig. S8). A similar underestimation of the negative feedback has been noted in climate models^68,78, which emphasized that weaker damping of the heat flux-SST feedback is attributed to weak shortwave feedback in the presence of equatorial cold bias⁷⁹. Notably, models with stronger ENSO variability, such as CESM and COSMOS, also exhibit stronger thermal damping. Similarly, the enhanced ENSO amplitude in EECO simulations also corresponds to an intensification of the negative feedback, particularly in the 3xCO₂ simulation. The enhanced evaporation due to warmer mean SST (not shown) in the 3xCO₂ simulation likely amplifies the negative feedback. This balance between positive and negative feedback processes suggests that intensified ocean-atmosphere interactions played a crucial role in sustaining strong ENSO variability during the EECO.

Tropical inter-basin climate interaction

In the current climate, the tropical Pacific Ocean is interconnected with the Atlantic and Indian Oceans through atmospheric circulation^18,53,80 as well as with the latter through the oceanic pathway of the Indonesian seas⁸¹. Although the Indian Ocean typically exhibits a passive response to ENSO^53,80, its internal variability^53,82,83 may also influence the Pacific through atmospheric teleconnection^84,85,86. Recent studies also indicated that the Atlantic Ocean exerts considerable influence on the Indo-Pacific SST variability on interannual to decadal time scales by accelerating the Pacific trade winds^87,88,89. As the enhanced SST variability in the EECO extends beyond the Pacific Ocean basin, it might have a stronger trans-basin relationship that can provide positive feedback to ENSO through climatic interaction.

To estimate this relationship, we first examine the cross-correlation between the Indian Ocean (50°E-120°E, 20°S-20°N) and Pacific Niño3.4 (170°W-120°W, 5°S-5°N) SSTA (Fig. 3a). Note that the position of the tropical warm pool remained largely similar to present-day conditions (Fig. S3), likely preserving the location of the ascending and descending branches of the Walker circulation across simulations. Therefore, we use the same region to assess tropical inter-basin interactions consistently in all cases. In observation, the Pacific SSTA leads the Indian Ocean SSTA by about 4 months with a strong correlation ~ 0.5, indicating the development of passive response of ENSO on the basin-scale Indian Ocean warming after about a season⁸⁵. ENSO drives this canonical interannual Indian Ocean SST variability, which also contributes to the demise of ENSO by enhancing convective activities over the Indian Ocean^90,91. Although the piControl captures the observed phase lag between SSTA warming of two ocean basins, the correlation is marginally underestimated, particularly when the Indian Ocean leads the Pacific. This suggests that the capacitor effect of the Indian Ocean⁸⁵ is weaker in the piControl. In the EECO simulations, the trans-basin relationship between the Indo-Pacific Oceans appears stronger than that in piControl. The positive cross-correlation between the two basins, when the Pacific leads the Indian Ocean, is even stronger than the observed and shifted to shorter lead times. The stronger and more in-phase correlation suggests that the inter-basin interaction between the Indian and Pacific Oceans in the EECO was stronger than present. DH24 indicated that such a stronger inter-basin coupling likely increases the SSTA variability in the Pacific.

**Fig. 3: Tropical inter-basin climate interaction.**

Similarly, the trans-basin relationship between the Pacific and Atlantic Ocean basins is examined using cross-correlation between SSTA in the Pacific Niño3.4 (170°W-120°W, 5°S-5°N) and Atlantic (40°W-20°W, 5°S-5°N) boxes (Fig. 3b). In observation, the maximum correlation is noted by about a 6-month lead with negative Atlantic SSTA leading to positive Pacific SSTA. It indicates a warm (cold) Atlantic SSTA causes a cooling (warming) in the Pacific Niño3 region and is consistent with the results in Keenlyside and Latif⁹². However, piControl simulations are unable to capture the observed relationship, likely due to a weaker-than-observed Atlantic warming trend and inherent Atlantic SST mean biases in the models, which underestimates the Walker circulation response to the Pacific⁹³. During the EECO, the wider and interconnected tropical ocean basin facilitated stronger and more direct interactions between the Pacific and Atlantic, likely leading to enhanced inter-basin teleconnections when the Pacific SSTA variability leads the Atlantic. The broader ocean basin allowed for more coherent and synchronous variability between the two ocean basins, strengthening the inter-basin interaction. While the exact mechanisms remain to be fully explored, this strengthened pan-tropical climate interaction likely contributed to the enhanced ENSO variability in the EECO.

Role of interconnected tropical ocean basins

The wider tropical ocean basin in the EECO appears to be key to the enhanced amplitude and longer periodicity of ENSO. To examine this hypothesis, we conduct a sensitivity experiment by restoring the present width of the Pacific within the EECO 1xCO₂ configuration in the GFDL model. In this experiment, referred to as the “land block” experiment, the deep ocean connections between the Indian and Pacific Ocean basins, and the Atlantic and Pacific Ocean basins are blocked by placing two flat landmasses with a uniform topographic height of 10 m (see map in Fig. 4c). This essentially produces an isolated Pacific basin, more similar to the piControl simulation, while keeping all other geographic changes in land topography or ocean bathymetry the same as in the 1xCO₂ experiment.

**Fig. 4: Standard deviation of sea surface temperature anomalies and periodicity in GFDL experiments.**

Figure 4a–c illustrates the role of ocean basin width on the monthly SSTA standard deviation. The enhanced SSTA standard deviation in the 1xCO₂ experiment is considerably reduced in the “land block” experiment. Specifically, the Niño3.4-averaged standard deviation ratios between the EECO experiments and the piControl simulation are markedly different, with ratios of 3.07 in the GFDL 1xCO2 experiment, while that reduces to 1.72 in the “land block” experiment, indicating nearly a 50% reduction in SSTA standard deviation under the narrower Pacific Ocean basin conditions. The decline in SSTA standard deviation is also apparent in other ocean basins, underscoring the impact of wider tropical ocean basins in the EECO climate system. Despite the “land block” experiment limits any oceanic interactions among the Pacific, Indian, and Atlantic Oceans, the experiment still produces 72% more SSTA standard deviation than piControl, suggesting the other tectonic factors, such as lower land topography, a closed Bering Strait, and a narrower Southern Ocean, may also contribute to modifying ENSO dynamics through altered pantropical interbasin interactions. Additionally, the weaker meridional temperature gradient in the 1xCO₂ experiment points to mean state changes that further influence ENSO amplitude.

The change in ENSO periodicity in the experiments is measured using autocorrelation of the Niño3.4 SSTA in Fig. 4d. In the 1xCO₂ experiment, Niño3.4 SSTA showed stronger negative autocorrelations that are shifted to longer lead times compared to the piControl, indicating stronger ENSO oscillations on longer periods in the EECO. Consistent with the results in Fig. 1g, the periodicity of the ENSO in the 1xCO₂ experiment is about 7 years relative to ~3 years in piControl simulation. The autocorrelation decays faster in the “land block” experiment, suggesting that the reduced basin width shifts the ENSO periodicity back to the piControl values. This notable decline in autocorrelation underscores the crucial influence of the tropical Pacific Ocean basin’s width in maintaining long-term ENSO dynamics, suggesting that the more extensive oceanic connections among the three basins during the EECO significantly modified ENSO behavior.

The propagation of oceanic waves plays a fundamental role in ENSO dynamics⁹⁴. The eastward-propagating downwelling (upwelling) Kelvin waves drive SSTA warming (cooling) in the eastern Pacific while deepening (shoaling) the thermocline along the equator. The warm SSTA initiates El Niño development, triggering westerly wind anomalies in the central Pacific. These winds, in turn, generate upwelling off-equatorial Rossby waves that propagate westward, reflect at the western boundary, and return as eastward-propagating upwelling Kelvin waves, which dampens the warm SSTA through negative feedback^95,96. A similar but opposite process occurs during La Niña, enabling interannual oscillations in the present ocean-atmosphere coupled climate system⁴⁴. Although stationary responses obscure some wave propagations in observation and piControl simulation (Fig. 5), the alignment between Rossby and Kelvin waves remains evident, consistent with theoretical frameworks. The simulated thermocline response in piControl exhibits over-damped oscillatory features. In contrast, the wider tropical ocean basin in the 1xCO₂ experiment shows a notably stronger thermocline response than piControl, contributing to longer ENSO periodicity. The wave adjustment time in the 1xCO₂ experiment is ~7 years, more than twice that of the 3-year adjustment time in piControl. However, the “land block” experiment shows a notably weaker thermocline response and the wave adjustment time reverts to values comparable to piControl, suggesting that wave reflections from the western and eastern boundaries remain similar to present-day conditions. This result is consistent with the periodicity changes shown in Fig. 4d.

**Fig. 5: Subsurface wave adjustment in GFDL experiment.**

The weaker ENSO variability in the “land block” sensitivity experiments may also be associated with the change in ocean-atmospheric feedback in the experiments. The analysis of the first elements of the Bjerknes feedback in the “land block” experiment, as in Fig. 2, suggests that Niño3 SSTA induces even stronger westerly wind stress feedback over the western Pacific and easterly wind stress feedback over the eastern Pacific (Fig. 6). However, the westward shift of the westerly wind stress reduces the propagation distance of oceanic Rossby waves before their reflection at the western boundary, accelerating their negative feedback on equatorial thermocline adjustment and leading to faster ENSO termination⁷⁷. The amplified wind stress feedback, exhibiting a tropical Pacific-wide dipole pattern in the “land block” experiment, aligns with DH24⁵⁴, which reported weaker Bjerknes feedback for wider ocean basins in an idealized single Pacific basin but its amplification when interactions with two additional ocean basins were included.

**Fig. 6: The positive and negative ocean-atmosphere feedback in GFDL experiment.**

The thermocline feedback to the wind stress anomalies appears to be weaker in the “land block” experiment than in the 1xCO₂ experiment, consistent with a westward shift in the wind stress. The deeper thermocline anomalies in the western Pacific that extended into the Indian Ocean in the 1xCO₂ experiment are truncated and weakened in the “land block” experiment. The reduced oceanic response in the “land block” experiment is likely to weaken the ENSO variability and contribute to shortening the ENSO periodicity. The local SST-thermocline response in the eastern Pacific in the “land block” experiment remains unaltered as in the 1xCO₂ experiment. However, restoring the EECO Pacific Ocean basin to present-day width further amplifies the thermal damping, suggesting a non-linear ocean-atmospheric interaction playing a crucial role in determining the SSTA variability in the Pacific.

The reduced ENSO variability with partially amplified Bjerknes feedback may stem from altered inter-basin climate interaction in the “land block” experiment. The amplified Indo-Pacific interaction in the 1xCO₂ experiment appears to be weakened at the longer lead times (Fig. 3) in the “land block” experiment. This may act to reduce the passive response of the Indian Ocean to the Pacific by weakening the capacitor effect in the “land block” experiment. However, the cross-correlation remains similar at shorter lead times. In contrast to the Indo-Pacific interaction, a much larger reduction in SSTA cross-correlation is evident for the Pacific-Atlantic interaction in the “land block” experiment. The enhanced response of the Pacific to the Atlantic in the 1xCO₂ experiment, indicating the presence of a single mode of variability across the tropical ocean, is weakened in the “land block” experiment. This weakening of the inter-basin interaction is evident at all lags. As noted by DH24⁵⁴, the weaker Pacific-Atlantic interaction may also contribute to the higher-frequency shift in ENSO periodicity in the “land block” experiment. Although the contribution of individual ocean basins is yet to be estimated, our results suggest that tropical basin interaction plays a fundamental role in ENSO dynamics that further needs to be considered in the conceptual ENSO models.

Discussion

Our investigation suggests that ENSO exhibited significantly longer periodicity and greater amplitude during the warmer EECO compared to present-day conditions, with potential implications for global climate patterns of that epoch. A key driver of these changes appears to be intensified ocean-atmosphere feedback within the wider tropical ocean basin. The expansion of the Pacific basin facilitated a stronger and more symmetric zonal wind stress feedback, reinforcing an enhanced east-west thermocline gradient. This led to a greater thermocline depth change in the eastern Pacific, increasing the sensitivity of SST to subsurface processes and amplifying interannual SST variability⁶⁷.

Additionally, well-connected tropical ocean basins during the EECO promoted robust ocean-atmosphere interactions, strengthening the coupling between SSTA and atmospheric responses. The stronger in-phase tropical inter-basin interactions observed in our EECO simulations align with previous findings⁵⁴. The broader ocean basin likely facilitated more extensive and coherent interactions among different tropical ocean regions, synchronizing SSTA across larger spatial scales and further amplifying ENSO variability.

The EECO simulations also reveal an equatorially symmetric zonal wind stress response over the western Pacific, with a broader meridional extent, particularly in the Northern Hemisphere (Fig. S9). This broader meridional extent of zonal wind stress supports more extensive thermocline adjustments, contributing to longer event durations or less efficient ENSO termination⁹⁷. Increased SST in the western Pacific warm pool likely intensified regional precipitation in the EECO, enhancing the zonal wind response⁹⁸. Meanwhile, the limited seasonal migration of the ITCZ during the EECO^58,99 reinforced this symmetric zonal wind pattern over the tropical western Pacific.

Although the “land block” experiment produces a zonal wind stress response with a meridional extent similar to the EECO simulations, the ENSO periodicity remains the same as in the piControl simulation. This result aligns with expectations, as the rigid western boundary in the “land block” experiment prevents oceanic throughflow¹⁰⁰, comparable to the piControl simulation. The westward shift of the zonal wind stress and absence of wave propagation from the tropical Pacific to the Indian Ocean¹⁰¹ lead to faster ENSO termination due to wave reflections from the rigid western boundary, which is positioned ~50° east of its location in the EECO simulations. On the other hand, the broader meridional extent of zonal wind stress in the “land block” experiment helps counteract the stronger western boundary reflection compared to piControl, effectively maintaining a comparable ENSO periodicity to piControl simulation.

The reduced basin width to the present-day configuration in the “land block” experiment results in weaker wind stress-thermocline feedback and a stronger thermal damping effect. The tropical inter-basin interactions are also weaker compared to the 1xCO₂ experiment, suggesting that the wider ocean basin during the EECO played a crucial role in increasing ENSO amplitude. Moreover, ENSO periodicity in the “land block” experiment is significantly reduced to piControl levels, likely due to comparable wave reflections from the western boundary, which facilitate faster ENSO termination relative to 1xCO₂ experiment. The weakened Pacific-Atlantic interbasin interactions may further contribute to the shorter ENSO periodicity.

The 1xCO₂ experiments suggest that tectonic changes during the EECO were the primary drivers of changes in global mean surface temperature, ENSO amplitude, periodicity, and ocean-atmosphere feedback, while elevated atmospheric CO₂ levels partially offset the enhanced ENSO variability in the 3xCO₂ simulations. The stronger east-west mean SST gradient in the 1xCO₂ experiment relative to piControl likely contributes to greater ENSO amplitude by strengthening the Bjerknes feedback. The equatorially symmetrical zonal wind stress response exhibits eastward shift toward the central Pacific, especially in the 1xCO₂ experiment. This eastward shift in zonal wind stress contributes to the displacement of the warm pool⁹⁵. Despite overall warmer mean SSTs and enhanced SST-wind stress feedback in the 3xCO₂ simulation, a relatively weaker east-west SST gradient, wind stress-thermocline feedback, and stronger thermal damping reduce ENSO amplitude compared to the 1xCO₂ experiment. It suggests that elevated CO₂ modifies ENSO through mean state changes and amplifies SST-wind stress feedback, but its impact is offset by thermal damping and weakened thermocline-wind stress feedback. The crucial role of thermocline-wind stress feedback and thermal damping in determining ENSO amplitude aligns with previous findings¹⁴ that identified thermocline feedback as one of the key contributors to ENSO variability.

Our analysis of the equilibrium warm period reveals ENSO dynamics that differ from the transient responses, where high atmospheric CO₂ levels are often linked directly to increased ENSO variability^7,102. In the equilibrium state, the interactions between positive dynamical feedback and negative thermodynamic damping appear to balance differently compared to transient scenarios^9,103. As the eastern Pacific Ocean surpasses the convective threshold under global warming equilibrium, the collapse of equatorial upwelling weakens the thermocline feedback. Simultaneously, enhanced convection strengthens the negative thermal damping feedback^104,105. These dynamic and thermodynamic adjustments, driven by mean state changes toward a convective eastern Pacific, collectively act to reduce ENSO amplitude under sustained warming⁹. These insights are crucial for paleoclimate reconstructions and advancing our understanding of equilibrium warm climates like the EECO.

Despite our study focuses on ENSO during the warm EECO, the presence of a similarly expanded tropical ocean basin in the earlier Cenozoic era suggests that ENSO activity in the Paleogene period (66–23.03 Ma) was likely prolonged and more intense than today. These findings align with prior studies utilizing proxy data and modeling techniques^12,14,15. However, precise quantification of factors such as orbital forcing, the role of warm poles on ENSO during the Paleogene, remains elusive. Future research exploring ENSO teleconnections, the tropical Pacific heat budget, and the sensitivity of ENSO variability to individual tectonic changes to refine our understanding of ENSO dynamics under sustained EECO warm conditions will be pursued in the next phase of DeepMIP-Eocene model simulations, incorporating updated paleogeography, boundary conditions, and improved climate forcings. The DeepMIP-Eocene models used in this study demonstrate reasonable skill in simulating tropical mean climate (Fig. S10) and SST variability under present-day conditions. While model biases, particularly in representing Bjerknes feedback, tropical inter-basin interactions, and atmospheric teleconnections, affect ENSO asymmetry⁷⁸ and seasonal phase-locking¹⁰⁶, the climate models still provide the best tools for understanding the physical mechanisms of past and future climate change. Our findings highlight how tectonic changes and elevated atmospheric CO₂ levels influenced ENSO dynamics, offering valuable insights into the mechanisms driving past climate variability. These results have important implications for refining future climate projections under sustained global warming.

Methods

DeepMIP-eocene simulations, sensitivity experiments

Eight modeling groups participated in the DeepMIP-Eocene paleoclimate simulations⁵⁵, conducting experiments with standardized atmospheric CO₂ concentrations and other boundary conditions, including paleogeography, vegetation, and river routing based on Herold et al.¹⁰⁷ reconstruction. These Early Eocene experiments were designed without continental ice sheets and with atmospheric CO₂ concentrations at preindustrial levels (280 ppmv, 1x) and various higher levels of CO₂ concentrations (e.g., 3x, 6x, or 9x preindustrial values) for their climate sensitivity. The solar constant, orbital configuration, and non-CO₂ greenhouse gas concentrations were set to preindustrial values, while soil properties were prescribed as homogeneous global mean values based on the preindustrial simulation. Each simulation was run for several thousand years to achieve equilibrium and model spin-up, with only the last 100 years analyzed. We select six of the nine climate models (See Table 1 for details) based on the availability of both piControl and sensitivity experiments (1xCO₂ and 3xCO₂)¹⁰⁸. INM-CM4.8, IPSL-CM5A2, and NorESM1_F are excluded due to the lack of 1xCO₂ simulations, which allows us to assess the impact of tectonic forcing by directly comparing it with the piControl simulation.

We conduct an additional sensitivity experiment using the GFDL model with the 1xCO₂ configuration. This experiment aims to understand how changes in ocean basin width could alter ENSO characteristics by blocking the wider tropical ocean basin in the EECO. In the “land block” experiment, we completely isolate the tropical Pacific Ocean from the Indian and Atlantic Oceans using a narrow landmass with an elevation of 10 m. The experiment runs for 300 years, with the last 100 years analyzed in this study.

Observed datasets and analysis methods

The piControl simulated SST is evaluated against global monthly Extended Reconstructed SST (ERSST), version 5¹⁰⁹, at 2° longitude × 2° latitude resolution from 1979-2018. The net surface heat flux datasets are derived from the fifth-generation ECMWF reanalysis (ERA5)¹¹⁰ during the 1979–2018 period.

The thermocline, zonal, and meridional wind stress simulations in piControl are verified using the reanalysis dataset from Simple Ocean Data Assimilation version 3 (SODA3)¹¹¹ during 1980–2017. The thermocline depth is derived based on the maximum gradient in the ocean temperature profile following the approach of Dommenget et al.¹¹². They demonstrated the superiority of the maximum gradient of the temperature profile over the widely used depth of the 20 °C isotherm for the analysis of the ENSO dynamics.

The model datasets are interpolated onto a common reanalysis or observation grid. The performance of the piControl simulations over the tropical oceans (30°S-30°N) is reasonable in reproducing the observed mean SST and wind stress (Fig. S10). Additionally, the simulation skills of the equatorial Pacific (10°S-10°N, 130°E-90°W) thermocline pattern are evaluated in the Taylor diagram¹¹³, which incorporates three key metrics: correlation coefficient, normalized standard deviation, and normalized root mean squared error (RMSE). Although some models overestimate the zonal wind stress standard deviation, the overall performance of the models is quite similar, with RMSE mostly below 0.75, indicating a consistent portrayal of tropical ocean-atmospheric interactions across the ensemble of simulations.

We subtract the climatological seasonal cycle and the linear trend from all monthly time series datasets to derive the anomalies. The Niño3 region is defined as 90°W-150°W and 5°S-5°N, the Niño3.4 region as 120°W-170°W and 5°S-5°N, and the Niño4 region as 160°E-150°W and 5°S-5°N. The leading principal component (PC1) is derived from the first empirical orthogonal function (EOF) of detrended SSTA over tropical Pacific Ocean (30°S-30°N, 120°E-80°W).

The Recharge Oscillator model is based on two tendency eqs.⁶⁹:

$$\frac{d}{dt}SSTA(t)={a}_{11}SSTA(t)+{a}_{12}{h}_{a}(t)+{\eta }_{SSTA}$$

(1)

$$\frac{d}{dt}{h}_{a}(t)={a}_{21}SSTA(t)+{a}_{22}{h}_{a}(t)+{\eta }_{{h}_{a}}$$

(2)

where a₁₁ and a₂₂ represent the growth rates of SSTA and thermocline depth anomalies (h_a), respectively. The parameters a₁₂ and a₂₁ describe the coupling between SSTA and h_a, and η_SSTA and ${\eta }_{{h}_{a}}$ represents the stochastic noise forcing terms.

The parameters in Eqs. (1 and 2) are estimated (refer to Table ST1 for details) for the simulations by applying multivariate linear regression to the monthly tendencies of SSTA and h_a against their respective monthly values^69,70. The residuals of the linear regression fit can be interpreted as the stochastic noise forcing of the SSTA and h_a, denoted as η_SSTA and ${\eta }_{{h}_{a}}$, respectively.

By integrating the Recharge Oscillator model with the fitted parameters, we can evaluate the statistics of SSTA and h_a, and their relationship to the model parameters^114,115,116. For the 18 DeepMIP simulations, the Recharge Oscillator model successfully reproduces the std(SSTA) across simulations, achieving a correlation of 0.997. This underscores the Recharge Oscillator model’s skill in diagnosing the underlying causes of changes in std(SSTA). To assess the sensitivity of std(SSTA) to changes in each of the model parameters and the strength of stochastic noise, we present the normalized change in std(SSTA) relative to each parameter. A value of +1.0 indicates that the entire change in std(SSTA) is attributed to the change in the parameter under consideration, −1.0 represents an exact inverse relationship, and 0.0 signifies no influence from the parameter.

Data availability

The DeepMIP-Eocene dataset¹⁰⁸ was accessed from the University of Bristol server: https://www.deepmip.org/data-eocene/.The GFDL “land block” experiment variables used in this study are available at https://doi.org/10.5281/zenodo.14934790. ERSST dataset were downloaded from https://www.ncei.noaa.gov/pub/data/cmb/ersst/v5/netcdf/. SODA dataset were obtained from https://dsrs.atmos.umd.edu/DATA/soda3.12.2/REGRIDED/ocean/. Monthly ERA5 reanalysis dataset were downloaded from https://cds.climate.copernicus.eu/cdsapp#!/dataset/reanalysis-era5-single-levels-monthly-means?tab=form.

Code availability

The GFDL CM2.1 climate model code is publicly accessible as part of the MOM5.1.0 release at https://github.com/mom-ocean/MOM5. The “land block” experiment resources, selected variables, and the Python code for thermocline depth calculation are available at https://doi.org/10.5281/zenodo.14934790.

References

Bjerknes, J. Atmospheric teleconnections from the equatorial Pacific. Month. Weather Rev. 97, 163–172 (1969).
Article ADS Google Scholar
Philander, S. G. El Niño, La Niña, and the Southern oscillation. Int. Geophys. Ser. 46, 289 (1989).
Google Scholar
Diaz, H. F & Markgraf, V. El Niño and the Southern Oscillation: 748 Multiscale Variability and Global and Regional impacts (Cambridge University Press, 2000).
McPhaden, M. J., Zebiak, S. E. & Glantz, M. H. ENSO as an integrating concept in earth science. Science 314, 1740–1745 (2006).
Article ADS CAS PubMed Google Scholar
Barnett, T. The interaction of multiple time scales in the tropical climate system. J. Clim. 4, 269–285 (1991).
Cai, W. Increasing frequency of extreme El Niño events due to greenhouse warming. Nat. Clim. Change 4, 111–116 (2014).
Article ADS Google Scholar
Cai, W. Changing El Niño–Niño-Southern Oscillation in a warming climate. Nat. Rev. Earth Environ. 2, 628–644 (2021).
Article Google Scholar
Capotondi, A. Understanding ENSO diversity. Bull. Am. Meteorol. Soc. 96, 921–938 (2015).
Article ADS Google Scholar
Callahan, C. W. et al. Robust decrease in El Niño/Southern Oscillation amplitude under long-term warming. Nat. Clim. Change 11, 752–757 (2021).
Article ADS Google Scholar
Trenberth, K. E. et al. Progress during TOGA in understanding and modeling global teleconnections associated with tropical sea surface temperatures. J. Geophys. Res. Oceans 103, 14291–14324 (1998).
Article ADS Google Scholar
Taschetto, A. S. et al. ENSO atmospheric teleconnections. El Niño Southern Oscillation in a changing climate, 309–335 https://doi.org/10.1002/9781119548164.ch14 (2020).
Ripepe, M., Roberts, L. T. & Fischer, A. G. ENSO and sunspot cycles in varved Eocene oil shales from image analysis. J. Sediment. Res. 61, 1155–1163 (1991).
Google Scholar
Zhang, Y. G., Pagani, M. & Liu, Z. A 12-million-year temperature history of the tropical Pacific ocean. Science 344, 84–87 (2014).
Article ADS CAS PubMed Google Scholar
Li, X. et al. Persistently active El Niño–Southern Oscillation since the Mesozoic. Proc. Natl. Acad. Sci. USA 121, 2404758121 (2024).
Article Google Scholar
Huber, M. & Caballero, R. Eocene El Nino: evidence for robust tropical dynamics in the “hothouse”. Science 299, 877–881 (2003).
Article ADS CAS PubMed Google Scholar
Jin, F.-F. An equatorial ocean recharge paradigm for ENSO. Part I: conceptual model. J. Atmos. Sci. 54, 811–829 (1997).
Article ADS Google Scholar
Jin, F.-F. An equatorial ocean recharge paradigm for ENSO. Part II: a stripped-down coupled model. J. Atmos. Sci. 54, 830–847 (1997).
Article ADS Google Scholar
Cai, W. et al. Pantropical climate interactions. Science 363, 4236 (2019).
Article Google Scholar
Okumura, Y. M. & Deser, C. Asymmetry in the duration of El Niño and La Niña. J. Clim. 23, 5826–5843 (2010).
Article ADS Google Scholar
Harrison, D. & Vecchi, G. A. On the termination of El Niño. Geophys. Res. Lett. 26, 1593–1596 (1999).
Article ADS Google Scholar
McGregor, S. et al. Meridional movement of wind anomalies during ENSO events and their role in event termination. Geophys. Res. Lett. 40, 749–754 (2013).
Article ADS Google Scholar
Ham, Y.-G., Kug, J.-S. & Park, J.-Y. Two distinct roles of Atlantic SSTs in ENSO variability: north tropical Atlantic SST and Atlantic Niño. Geophys. Res. Lett. 40, 4012–4017 (2013).
Article ADS Google Scholar
Richter, I., Tokinaga, H., Kosaka, Y., Doi, T. & Kataoka, T. Revisiting the tropical Atlantic influence on El Niño–Southern oscillation. J. Clim. 34, 8533–8548 (2021).
Article ADS Google Scholar
Rensch, P., McGregor, S., Dommenget, D., Bi, D. & Liguori, G. The tropical Atlantic’s asymmetric impact on the El Niño–Southern oscillation. Geophys. Res. Lett. 51, 2023–106585 (2024).
Google Scholar
Dai, A. & Wigley, T. Global patterns of ENSO-induced precipitation. Geophys. Res. Lett. 27, 1283–1286 (2000).
Article ADS Google Scholar
Huang, P. & Xie, S.-P. Mechanisms of change in ENSO-induced tropical Pacific rainfall variability in a warming climate. Nat. Geosci. 8, 922–926 (2015).
Article ADS CAS Google Scholar
Wang, B. & Chan, J. C. How strong ENSO events affect tropical storm activity over the western North Pacific. J. Clim. 15, 1643–1658 (2002).
Article ADS Google Scholar
Ashok, K., Tam, C.-Y., Lee, W.-J. ENSO Modoki impact on the Southern Hemisphere storm track activity during extended austral winter. Geophys. Res. Lett. 36, L12705 (2009).
Yeh, S.-W. et al. ENSO atmospheric teleconnections and their response to greenhouse gas forcing. Rev. Geophys. 56, 185–206 (2018).
Article ADS Google Scholar
Valle, C. A., Cruz, F., Cruz, J. B., Merlen, G. & Coulter, M. C. The impact of the 1982–1983 El Niño-Southern oscillation on seabirds in the Galapagos Islands, Ecuador. J. Geophys. Res. Oceans 92, 14437–14444 (1987).
Article ADS Google Scholar
Glynn, P. & De Weerdt, W. Elimination of two reef-building hydrocorals following the 1982-83 El Niño warming event. Science 253, 69–71 (1991).
Article ADS CAS PubMed Google Scholar
Abhik, S. et al. Influence of the 2015–2016 El Niño on the record-breaking mangrove dieback along northern Australia coast. Sci. Rep. 11, 20411 (2021).
Article ADS CAS PubMed PubMed Central Google Scholar
Kennett, J. P. & Stott, L. Abrupt deep-sea warming, palaeoceanographic changes and benthic extinctions at the end of the Palaeocene. Nature 353, 225–229 (1991).
Article ADS Google Scholar
Zachos, J. C. et al. A transient rise in tropical sea surface temperature during the Paleocene-Eocene thermal maximum. Science 302, 1551–1554 (2003).
Article ADS CAS PubMed Google Scholar
Anagnostou, E. et al. Changing atmospheric CO2 concentration was the primary driver of early Cenozoic climate. Nature 533, 380–384 (2016).
Article ADS CAS PubMed Google Scholar
Anagnostou, E. et al. Proxy evidence for state-dependence of climate sensitivity in the Eocene greenhouse. Nat. Commun. 11, 4436 (2020).
Article ADS CAS PubMed PubMed Central Google Scholar
Inglis, G. N. et al. Global mean surface temperature and climate sensitivity of the early Eocene Climatic Optimum (EECO), Paleocene–Eocene Thermal Maximum (PETM), and latest Paleocene. Climate 16, 1953–1968 (2020).
Google Scholar
Marincovich Jr, L. & Gladenkov, A. Y. Evidence for an early opening of the Bering Strait. Nature 397, 149–151 (1999).
Article ADS Google Scholar
Scher, H. D. et al. Onset of Antarctic Circumpolar Current 30 million years ago as Tasmanian Gateway aligned with westerlies. Nature 523, 580–583 (2015).
Article ADS CAS PubMed Google Scholar
Peterson, R. G. & White, W. B. Slow oceanic teleconnections linking the Antarctic Circumpolar Wave with the tropical El Niño-Southern Oscillation. J. Geophys. Res. Oceans 103, 24573–24583 (1998).
Article ADS Google Scholar
White, W. B., Chen, S.-C., Allan, R. J. & Stone, R. C. Positive feedbacks between the Antarctic Circumpolar Wave and the global El Niño-Southern Oscillation wave. J. Geophys. Res. Oceans 107, 29–1 (2002).
Article Google Scholar
Ivany, L. C., Brey, T., Huber, M., Buick, D. P. & Schöne, B. R. El Niño in the Eocene greenhouse recorded by fossil bivalves and wood from Antarctica. Geophys. Res. Lett. 38, L16709 (2011).
Suarez, M. J. & Schopf, P. S. A delayed action oscillator for ENSO. J. Atmos. Sci. 45, 3283–3287 (1988).
Article ADS Google Scholar
Battisti, D. S. & Hirst, A. C. Interannual variability in a tropical atmosphere–ocean model: influence of the basic state, ocean geometry and nonlinearity. J. Atmos. Sci. 46, 1687–1712 (1989).
Article ADS Google Scholar
Neelin, J. D. et al. ENSO theory. J. Geophys. Res. Oceans 103, 14261–14290 (1998).
Article ADS Google Scholar
An, S.-I. & Wang, B. Interdecadal change of the structure of the ENSO mode and its impact on the ENSO frequency. J. Clim. 13, 2044–2055 (2000).
Article ADS Google Scholar
Dijkstra, H. The ENSO phenomenon: theory and mechanisms. Adv. Geosci. 6, 3–15 (2006).
Article Google Scholar
Cane, M. A. The evolution of El Niño, past and future. Earth Planet. Sci. Lett. 230, 227–240 (2005).
Article ADS CAS Google Scholar
Philander, S., Pacanowski, R. & Yamagata, T. Unstable air-sea interactions in the tropics. J. Atmos. Sci. 41, 604–613 (1984).
Article ADS Google Scholar
Gill, A. E. Some simple solutions for heat-induced tropical circulation. Q. J. R. Meteorol. Soc. 106, 447–462 (1980).
ADS Google Scholar
Lindzen, R. S. & Nigam, S. On the role of sea surface temperature gradients in forcing low-level winds and convergence in the tropics. J. Atmos. Sci. 44, 2418–2436 (1987).
Article ADS Google Scholar
Jin, F.-F., Kim, S. T., Bejarano, L. A coupled-stability index for ENSO. Geophys. Res. Lett. 33, L23708 (2006).
Klein, S. A., Soden, B. J. & Lau, N.-C. Remote sea surface temperature variations during ENSO: evidence for a tropical atmospheric bridge. J. Clim. 12, 917–932 (1999).
Article ADS Google Scholar
Dommenget, D., Hutchinson, D.: El niño southern oscillation and tropical basin interaction in idealized worlds. https://doi.org/10.21203/rs.3.rs-4002023/v1 (2024).
Lunt, D. J. et al. DeepMIP: model intercomparison of Early Eocene Climatic Optimum (EECO) large-scale climate features and comparison with proxy data. Clim. Discuss. 2020, 1–27 (2020).
Google Scholar
Lunt, D. J. et al. The DeepMIP contribution to PMIP4: experimental design for model simulations of the EECO, PETM, and pre-PETM (version 1.0). Geosci. Model Dev. 10, 889–901 (2017).
Article ADS CAS Google Scholar
Williams, C. J. et al. African hydroclimate during the Early Eocene from the DeepMIP simulations. Paleoceanogr. Paleoclimatol. 37, 2022–004419 (2022).
Article Google Scholar
Abhik, S. et al. Unraveling weak and short South Asian wet season in the Early Eocene warmth. Commun. Earth Environ. 5, 133 (2024).
Article ADS Google Scholar
Palmer, T. N. et al. Development of a European multimodel ensemble system for seasonal-to-interannual prediction (DEMETER). Bull. Am. Meteorol. Soc. 85, 853–872 (2004).
Article ADS Google Scholar
DelSole, T., Nattala, J. & Tippett, M. K. Skill improvement from increased ensemble size and model diversity. Geophys. Res. Lett. 41, 7331–7342 (2014).
Article ADS Google Scholar
Hoerling, M. P., Kumar, A. & Zhong, M. El Niño, La Niña, and the nonlinearity of their teleconnections. J. Clim. 10, 1769–1786 (1997).
Article ADS Google Scholar
Kang, I.-S. & Kug, J.-S. El Niño and La Niña sea surface temperature anomalies: asymmetry characteristics associated with their wind stress anomalies. J. Geophys. Res. Atmos. 107, 1 (2002).
Article Google Scholar
McPhaden, M. J. Evolution of the 2002/03 El Niño. Bull. Am. Meteorol. Soc. 85, 677–696 (2004).
Article ADS Google Scholar
Trenberth, K.E. El Niño Southern Oscillation (ENSO) (ENSO, 2019).
Bayr, T., Latif, M., Dommenget, D., Wengel, C., Harlaß, J. & Park, W. Mean-state dependence of ENSO atmospheric feedbacks in climate models. Clim. Dyn. 50, 3171–3194 (2018).
Article Google Scholar
Sun, D.-Z. A possible effect of an increase in the warm-pool SST on the magnitude of El Niño warming. J. Clim. 16, 185–205 (2003).
Article ADS Google Scholar
Meehl, G. et al. Factors that affect the amplitude of El Niño in global coupled climate models. Clim. Dyn. 17, 515–526 (2001).
Article Google Scholar
Bayr, T. et al. Error compensation of ENSO atmospheric feedbacks in climate models and its influence on simulated ENSO dynamics. Clim. Dyn. 53, 155–172 (2019).
Article Google Scholar
Burgers, G., Jin, F.-F., Van Oldenborgh, G. J. The simplest ENSO recharge oscillator. Geophys. Res. Lett. 32, L13706 (2005).
Jansen, M. F., Dommenget, D. & Keenlyside, N. Tropical atmosphere–ocean interactions in a conceptual framework. J. Clim. 22, 550–567 (2009).
Article ADS Google Scholar
Lloyd, J., Guilyardi, E., Weller, H. & Slingo, J. The role of atmosphere feedbacks during ENSO in the CMIP3 models. Atmos. Sci. Lett. 10, 170–176 (2009).
Article ADS Google Scholar
Dommenget, D. & Yu, Y. The seasonally changing cloud feedbacks contribution to the ENSO seasonal phase-locking. Clim. Dyn. 47, 3661–3672 (2016).
Article Google Scholar
Bellenger, H., Guilyardi, E., Leloup, J., Lengaigne, M. & Vialard, J. ENSO representation in climate models: from CMIP3 to CMIP5. Clim. Dyn. 42, 1999–2018 (2014).
Article Google Scholar
Richter, I. & Tokinaga, H. An overview of the performance of CMIP6 models in the tropical Atlantic: mean state, variability, and remote impacts. Clim. Dyn. 55, 2579–2601 (2020).
Article Google Scholar
Rensch, P., McGregor, S. & Dommenget, D. Exploration of atmosphere-only model deficiencies in reproducing the 1992–2011 Pacific trade wind acceleration. Geophys. Res. Lett. 49, 2022–099981 (2022).
Google Scholar
McGregor, S., Timmermann, A., Jin, F.-F. & Kessler, W. S. Charging El Niño with off-equatorial westerly wind events. Clim. Dyn. 47, 1111–1125 (2016).
Article Google Scholar
Wang, B. & An, S. A mechanism for decadal changes of ENSO behavior: roles of background wind changes. Clim. Dyn. 18, 475–486 (2002).
Article Google Scholar
Zhao, Y. & Sun, D.-Z. ENSO asymmetry in CMIP6 models. J. Clim. 35, 5555–5572 (2022).
Article ADS Google Scholar
Kim, S. T., Cai, W., Jin, F.-F. & Yu, J.-Y. ENSO stability in coupled climate models and its association with mean state. Clim. Dyn. 42, 3313–3321 (2014).
Article Google Scholar
Latif, M. & Barnett, T. P. Interactions of the tropical oceans. J. Clim. 8, 952–964 (1995).
Article ADS Google Scholar
Sprintall, J. et al. The Indonesian seas and their role in the coupled ocean–climate system. Nat. Geosci. 7, 487–492 (2014).
Article ADS CAS Google Scholar
Saji, N., Goswami, B. N., Vinayachandran, P. & Yamagata, T. A dipole mode in the tropical Indian Ocean. Nature 401, 360–363 (1999).
Article ADS CAS PubMed Google Scholar
Lau, N.-C. & Nath, M. J. Atmosphere–ocean variations in the Indo-Pacific sector during ENSO episodes. J. Clim. 16, 3–20 (2003).
Article ADS Google Scholar
Xie, S.-P., Annamalai, H., Schott, F. A. & McCreary, J. P. Structure and mechanisms of South Indian Ocean climate variability. J. Clim. 15, 864–878 (2002).
Article ADS Google Scholar
Xie, S.-P. et al. Indian Ocean capacitor effect on Indo–western Pacific climate during the summer following El Niño. J. Clim. 22, 730–747 (2009).
Article ADS Google Scholar
Dommenget, D., Semenov, V. & Latif, M. Impacts of the tropical Indian and Atlantic Oceans on ENSO. Geophys. Res. Lett. 33, L11701 (2006).
McGregor, S. et al. Recent Walker circulation strengthening and Pacific cooling amplified by Atlantic warming. Nat. Clim. Change 4, 888–892 (2014).
Article ADS Google Scholar
Li, X., Xie, S.-P., Gille, S. T. & Yoo, C. Atlantic-induced pan-tropical climate change over the past three decades. Nat. Clim. Change 6, 275–279 (2016).
Article ADS Google Scholar
Wang, L., Yu, J.-Y. & Paek, H. Enhanced biennial variability in the Pacific due to Atlantic capacitor effect. Nat. Commun. 8, 14887 (2017).
Article ADS CAS PubMed PubMed Central Google Scholar
Kug, J.-S. & Kang, I.-S. Interactive feedback between ENSO and the Indian Ocean. J. Clim. 19, 1784–1801 (2006).
Article ADS Google Scholar
Ohba, M. & Ueda, H. An impact of SST anomalies in the Indian Ocean in acceleration of the El Niño to La Niña transition. J. Meteorol. Soc. Jpn. Ser II 85, 335–348 (2007).
Article ADS Google Scholar
Keenlyside, N. S. & Latif, M. Understanding equatorial Atlantic interannual variability. J. Clim. 20, 131–142 (2007).
Article ADS Google Scholar
McGregor, S., Stuecker, M. F., Kajtar, J. B., England, M. H. & Collins, M. Model tropical Atlantic biases underpin diminished Pacific decadal variability. Nat. Clim. Change 8, 493–498 (2018).
Article ADS Google Scholar
Wang, C. On the ENSO mechanisms. Adv. Atmos. Sci. 18, 674–691 (2001).
Article Google Scholar
Picaut, J., Masia, F. & Penhoat, Y. An advective-reflective conceptual model for the oscillatory nature of the ENSO. Science 277, 663–666 (1997).
Article CAS Google Scholar
Wang, C. A review of ENSO theories. Natl Sci. Rev. 5, 813–825 (2018).
Article Google Scholar
McGregor, S., Dommenget, D. & Neske, S. Distinct off-equatorial zonal wind stress and oceanic responses for EP-and CP-type ENSO events. J. Clim. 35, 1423–1440 (2022).
Article ADS Google Scholar
Gong, Y. & Li, T. Mechanism for southward shift of zonal wind anomalies during the mature phase of ENSO. J. Clim. 34, 8897–8911 (2021).
ADS Google Scholar
Kelemen, F. D. et al. Meridional heat transport in the DeepMIP Eocene ensemble: Non-CO2 and CO2 effects. Paleoceanogr. Paleoclimatol. 38, 2022–004607 (2023).
Article Google Scholar
Gordon, A. Oceanography of the Indonesian seas and their throughflow. Oceanography 18, 14–27 (2005).
Article Google Scholar
Clarke, A. J. On the reflection and transmission of low-frequency energy at the irregular western Pacific Ocean boundary. J. Geophys. Res. Oceans 96, 3289–3305 (1991).
Article ADS Google Scholar
Cai, W. et al. Anthropogenic impacts on twentieth-century ENSO variability changes. Nat. Rev. Earth Environ. 4, 407–418 (2023).
Article ADS Google Scholar
Collins, M. et al. The impact of global warming on the tropical Pacific ocean and El Niño. Nat. Geosci. 3, 391–397 (2010).
Article ADS CAS Google Scholar
Peng, Q., Xie, S.-P. & Deser, C. Collapsed upwelling projected to weaken ENSO under sustained warming beyond the twenty-first century. Nat. Clim. Change 14, 815–822 (2024).
Article Google Scholar
Geng, T., Cai, W., Jia, F. & Wu, L. Decreased ENSO post-2100 in response to formation of a permanent El Ni{\̃n}o-like state under greenhouse warming. Nat. Commun. 15, 5810 (2024).
Article CAS PubMed PubMed Central Google Scholar
Chen, H.-C. & Jin, F.-F. Dynamics of ENSO phase–locking and its biases in climate models. Geophys. Res. Lett. 49, 2021–097603 (2022).
Google Scholar
Herold, N. et al. A suite of early Eocene (̃ 55 Ma) climate model boundary conditions. Geosci. Model Dev. 7, 2077–2090 (2014).
Article ADS Google Scholar
Steinig, S. et al. DeepMIP-Eocene-p1: multi-model dataset and interactive web application for Eocene climate research. Sci. Data 11, 970 (2024).
Article PubMed PubMed Central Google Scholar
Huang, B. et al. Extended reconstructed sea surface temperature, version 5 (ERSSTv5): upgrades, validations, and intercomparisons. J. Clim. 30, 8179–8205 (2017).
Article ADS Google Scholar
Hersbach, H. et al. The ERA5 global reanalysis. Q. J. R. Meteorol. Soc. 146, 1999–2049 (2020).
Article ADS Google Scholar
Carton, J. A. & Giese, B. S. A reanalysis of ocean climate using Simple Ocean Data Assimilation (SODA). Monthly Weather Rev. 136, 2999–3017 (2008).
Article ADS Google Scholar
Dommenget, D., Priya, P. & Vijayeta, A. ENSO phase space dynamics with an improved estimate of the thermocline depth. Clim. Dyn. 61, 5767–5783 (2023).
Article Google Scholar
Taylor, K. E. Summarizing multiple aspects of model performance in a single diagram. J. Geophys. Res. Atmos. 106, 7183–7192 (2001).
Article ADS Google Scholar
Vijayeta, A. & Dommenget, D. An evaluation of ENSO dynamics in CMIP simulations in the framework of the recharge oscillator model. Clim. Dyn. 51, 1753–1771 (2018).
Article Google Scholar
Wengel, C., Dommenget, D., Latif, M., Bayr, T. & Vijayeta, A. What controls ENSO-amplitude diversity in climate models? Geophys. Res. Lett. 45, 1989–1996 (2018).
Article ADS Google Scholar
Dommenget, D. & Vijayeta, A. Simulated future changes in ENSO dynamics in the framework of the linear recharge oscillator model. Clim. Dyn. 53, 4233–4248 (2019).
Article Google Scholar
Hurrell, J. W. et al. The community earth system model: a framework for collaborative research. Bull. Am. Meteorol. Soc. 94, 1339–1360 (2013).
Article ADS Google Scholar
Zhu, J., Poulsen, C. J. & Tierney, J. E. Simulation of Eocene extreme warmth and high climate sensitivity through cloud feedbacks. Sci. Adv. 5, 1874 (2019).
Article ADS Google Scholar
Jungclaus, J. H. et al. Ocean circulation and tropical variability in the coupled model ECHAM5/MPI-OM. J. Clim. 19, 3952–3972 (2006).
Article ADS Google Scholar
Delworth, T. L. et al. GFDL’s CM2 global coupled climate models. Part I: formulation and simulation characteristics. J. Clim. 19, 643–674 (2006).
Article ADS Google Scholar
Valdes, P. J. et al. The BRIDGE HadCM3 family of climate models: HadCM3@Bristol v1. 0. Geosci. Model Dev. 10, 3715–3743 (2017).
Article ADS CAS Google Scholar
Chan, W.-L., Abe-Ouchi, A. & Ohgaito, R. Simulating the mid-Pliocene climate with the MIROC general circulation model: experimental design and initial results. Geosci. Model Dev. 4, 1035–1049 (2011).
Article ADS Google Scholar

Download references

Acknowledgements

S.A., D.D., and F.A.C. acknowledge the support of the Australian Research Council (ARC) Discovery Project (DP220101468). D.K.H. is supported by the ARC grant DE220100279. The CESM project is primarily funded by the National Science Foundation (NSF). J.Z. acknowledges support from the National Center for Atmospheric Research (NCAR), which is a major facility sponsored by the NSF under Cooperative Agreement No. 1852977. We are grateful to Julie M. Arblaster for her assistance in obtaining additional datasets from the NCAR server. The “land block” sensitivity experiment and all analyses are conducted on the National Computational Infrastructure (NCI) supercomputer, supported by the Australian Government.

Author information

Authors and Affiliations

School of Earth, Atmosphere, and Environment, Monash University, Clayton, VIC, Australia
S. Abhik, Dietmar Dommenget, Shayne McGregor & Fabio A. Capitanio
ARC Centre of Excellence for Climate Extremes, Monash University, Clayton, VIC, Australia
Dietmar Dommenget
ARC Centre of Excellence for 21st Century Weather, Monash University, Clayton, VIC, Australia
Shayne McGregor
Climate Change Research Centre, University of New South Wales, Sydney, NSW, Australia
David K. Hutchinson
School of Geographical Sciences, University of Bristol, Bristol, UK
Sebastian Steinig & Daniel J. Lunt
Climate and Global Dynamics Laboratory, NSF National Center for Atmospheric Research, Boulder, CO, USA
Jiang Zhu
Institute of Geological Sciences, Polish Academy of Sciences, Kraków, Poland
Igor Niezgodzki
Alfred Wegener Institute, Helmholtz Centre for Polar and Marine Research, Bremerhaven, Germany
Igor Niezgodzki & Gregor Knorr
Atmosphere and Ocean Research Institute, University of Tokyo, Kashiwa, Japan
Wing-Le Chan & Ayako Abe-Ouchi

Authors

S. Abhik
View author publications
Search author on:PubMed Google Scholar
Dietmar Dommenget
View author publications
Search author on:PubMed Google Scholar
Shayne McGregor
View author publications
Search author on:PubMed Google Scholar
David K. Hutchinson
View author publications
Search author on:PubMed Google Scholar
Sebastian Steinig
View author publications
Search author on:PubMed Google Scholar
Jiang Zhu
View author publications
Search author on:PubMed Google Scholar
Fabio A. Capitanio
View author publications
Search author on:PubMed Google Scholar
Daniel J. Lunt
View author publications
Search author on:PubMed Google Scholar
Igor Niezgodzki
View author publications
Search author on:PubMed Google Scholar
Gregor Knorr
View author publications
Search author on:PubMed Google Scholar
Wing-Le Chan
View author publications
Search author on:PubMed Google Scholar
Ayako Abe-Ouchi
View author publications
Search author on:PubMed Google Scholar

Contributions

S.A. led the analysis and drafted the manuscript following input from D.D. and S.M. D.D. conducted the Recharge Oscillator model analysis. D.K.H. performed the GFDL sensitivity experiment in collaboration with S.A. and D.D. S.S., J.Z., D.K.H., I.N., G.K., W.-L.C. and A.A.-O. contributed to the DeepMIP-Eocene simulations, with D.J.L. supervising the DeepMIP-Eocene simulation project and coordinating with S.S. to process and archive the datasets. S.A., D.D., S.M., D.K.H., J.Z., F.A.C., D.J.L., I.N., G.K., and W.-L.C. participated in discussions and revised the manuscript.

Corresponding author

Correspondence to S. Abhik.

Ethics declarations

Competing interests

The authors declare no competing interests.

Peer review

Peer review information

Nature Communications thanks Jong-Yeon Park, De-Zheng Sun, and Xiao-Tong Zheng for their contribution to the peer review of this work. A peer review file is available.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary Information

Transparent Peer Review file

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Cite this article

Abhik, S., Dommenget, D., McGregor, S. et al. Stronger and prolonged El Niño-Southern Oscillation in the Early Eocene warmth. Nat Commun 16, 4053 (2025). https://doi.org/10.1038/s41467-025-59263-7

Download citation

Received: 14 November 2024
Accepted: 16 April 2025
Published: 30 April 2025
DOI: https://doi.org/10.1038/s41467-025-59263-7