Deep ocean warming-induced El Niño changes

The changes in tropical Pacific mean state

To simulate the climate after mitigating global warming by reducing CO₂ emissions, the Community Earth System Model (CESM) was run with constant atmospheric CO₂ (restoring) after it was increased for 140 years, then symmetrically decreased (see “Methods” and Supplementary Fig. 1). The mean state change was quantified by calculating the difference between the restoring period, defined as the entire 220 years of the restoring simulation, and the present-day (PD) period, defined as the last 400 years of the PD simulation. Figure 1a, b show the changes in the tropical mean state from the PD period. Even when the CO₂ concentration returns to current levels, the tropical mean state shows considerable differences. SST anomalies (SSTA) show positive signals across the tropics, indicating incomplete recovery and a warmer state compared to the present climate. In particular, El Niño-like warming (i.e., stronger warming in the eastern Pacific) is evident in the spatial pattern of the SST difference, which is similar to the slow response to global warming^33,34.

**Fig. 1: The changes in the tropical mean state.**

The increase in precipitation in the tropics and the decrease in the subtropics indicate a southward shift of the ITCZ in the Pacific (Fig. 1b). The latitudinal position of the ITCZ is linked to the energy transport across the equator, which is regulated by the meridional energy exchange between the hemispheres³⁵. As the climate recovers after CO₂ reaches current levels, the Northern Hemisphere (NH) cools much faster than the Southern Hemisphere (SH) (Supplementary Fig. 2). The larger heat capacity of the SH¹⁷ and the pronounced heat release from the Southern Ocean¹⁸ contribute to a relatively slow cooling rate in the SH. Therefore, the reduction in interhemispheric thermal contrast drives the shift of the ITCZ toward the relatively warmer SH. The southward shift of the Pacific ITCZ is accompanied by an increase in precipitation and a weakening of the trade winds over the equator, which further amplifies El Niño-like warming through the Bjerknes feedback. The mean state change shown in Fig. 1a, b will be eventually determined as a result of this positive feedback.

Previous study¹⁸ suggested that the SST response in the restoring period originates from deep ocean warming and climatological ocean stratification. To investigate the influence of deep ocean warming on changes in the tropical mean state, three initial warming experiments (IW_EXPs) were conducted. Specifically, the IW_EXPs were conducted with a constant atmospheric CO₂ level but with horizontally uniform vertical profiles of ocean temperature and salinity anomalies added to the ocean model initial conditions. These IW_EXPs allow us to explore the impacts of deep ocean warming on the mean state of the tropical ocean and subsequent alterations in ENSO characteristics (see “Methods” for a detailed description of the IW_EXPs).

The mean state change was quantified by calculating the difference between the average for the last 100 years after 50 years from the IW_EXPs and the PD period. Figure 1c, d shows the SST and precipitation responses when the initial ocean warming is added below 700 m (IW_be700). Although spatially uniform warming is only added below 700 m, the initial deep ocean warming leads to a particular SST pattern. Interestingly, the pattern is almost identical to that in the restoring period, with a high pattern correlation of 0.95 over the tropical Pacific. Similarly, the pattern correlation for the precipitation pattern is 0.97. Furthermore, the other IW_EXPs, adding initial ocean warming in the whole ocean depth (IW_whole) and below 100 m (IW_be100), also show similar changes in SST and precipitation, although their magnitudes are different (Supplementary Fig. 3), supporting strong robustness. These results suggest that deep ocean warming triggers specific patterns in the tropical mean state, such as El Niño-like warming and southward shift of the ITCZ. As reported in the previous study, the deep ocean acts as a stove, continuously supplying heat to the ocean surface, and the efficiency of this process depends on ocean stratification¹⁸. As a result, the energy from the deep ocean is more effectively transferred to the surface in the eastern Pacific, where stratification is weaker and climatological upwelling prevails¹⁸. The initial warming over the eastern Pacific (Supplementary Fig. 4) will be intensified by the Bjerknes feedback and the southward shift of the Pacific ITCZ.

The changes in the characteristics of the ENSO

The El Niño-like warming and the southward shift of the ITCZ have the potential to intensify the variability of the ENSO in the eastern Pacific²⁸. Therefore, we hypothesized that deep ocean warming may contribute to changes in ENSO characteristics by the tropical mean state changes. To examine this possibility, we first checked the amplitudes of the Niño indices (Fig. 2a, b, see “Methods”). The standard deviation (STD) of the Niño3 SSTA increases significantly, while the Niño4 SSTA remains unchanged except for IW_whole. Interestingly, the STDs of the Niño3 SSTA in IW_be700, the IW_be100, and IW_whole experiments consistently increase relative to the PD period, similar to the change in the restoring period. The largest increases are in the order of: IW_whole, IW_be100, and IW_be700, which are proportional to the tropical mean state warming (Supplementary Fig. 3). These results suggest that deep ocean warming enhances the SST variability over the eastern Pacific.

**Fig. 2: The changes in El Niño-Southern Oscillation amplitude.**

Figure 2c shows the El Niño composites for December-January-February (DJF) during the PD period. Here, El Niño events were defined based on 1 STD of the Niño3 SSTA (see “Methods”). The SSTA pattern shows positive anomalies in the equatorial Pacific and the Niño3 SSTA during the PD period, with a maximum near 141°W. During the restoring period, the Niño3 SSTA increased with the maximum near 134°W, indicating an increased ENSO amplitude and an eastward shift (Supplementary Fig. 5). The SSTA difference from the PD period shows a zonal dipole pattern (Fig. 2d), showing that the equatorial eastern and western Pacific regions have opposite signs, consistent with the increase in STD of the Niño3 (Fig. 2a). In addition, Coupled Model Intercomparison Project Phase 6 (CMIP6) models further support our argument (see Supplementary Note and Supplementary Fig. 6). Interestingly, IW_be700 simulates quite a similar difference in SSTA from the PD period (Fig. 2e). Although only uniform deep ocean warming is added to the initial condition, the model simulates mostly identical El Niño SSTA pattern to that in the restoring period, suggesting that the warming of the deep ocean is the most critical factor in changing El Niño characteristics in the restoring period.

The eastward shift of the maximum in the ENSO synthesis suggests a systematic change in ENSO diversity. To investigate this, we analyzed changes in the spatial diversity of El Niño using a method that detects the longitude of the SSTA peak during DJF (see “Methods”). The zonal distribution of SSTA peaks shows a bimodal structure with maxima at 155 W and 115 W (Fig. 3a), although the model tends to simulate more CP El Niño compared to the observational distribution²⁶ (see Fig. 1a, b). Based on this result, we can divide El Niño events into two distinct types: CP El Niño, with maximum SSTA between 165°E − 145°W, and EP El Niño, between 125°W − 105°W (Supplementary Fig. 7). In addition, to make a quantitative comparison, the CP ratio was defined as the CP events relative to the sum of CP and EP events. The change in SSTA peak occurrence shows that EP El Niño events are more frequent in the restoring period (Fig. 3b), consistent with the decrease in CP ratios in the restoring period relative to the PD period (Fig. 3d). The decrease in the frequency of CP El Niño events is consistently simulated in the IW_be700 (Fig. 3c), IW_be100, and IW_whole experiments (Supplementary Fig. 8). These results suggest that heat release from the deep ocean prefers to more frequent EP El Niño. In addition, the distribution for peaks of La Niña events also shows quite similar changes to that of El Niño events in the restoring period and all IW_EXPs (Supplementary Fig. 9).

**Fig. 3: The changes in El Niño-Southern Oscillation’s flavor.**

EP El Niño is mostly stronger than CP El Niños^36,37 and tends to be associated with extreme events, which are directly linked to anomalous climate conditions worldwide^31,38. To assess the risk by changes in the extreme case of El Niño, we have adopted the concept of a convective extreme El Niño (see “Methods”), which will be more closely related to anomalous atmospheric circulation^28,31. The changes in the frequency of the convective El Niño from the PD period are shown in Fig. 3e. During the PD period, convective El Niño occurs about 13 times per 100 years, accounting for 10% of all El Niño events. On the other hand, the frequency of convective El Niño increased by 40 to 80% in warmer climates. In addition, the CMIP6 models provide additional support for our argument (see Supplementary Note and Supplementary Fig. 6).

The changes in ENSO feedback

So far, we have shown that the deep ocean warming induces the tropical mean state change, resulting in more EP El Niño and convective El Niño. It is important to understand what physical processes lead to these changes in the ENSO characteristics. During the restoring period, there is a significant increase in the equatorial ocean temperature below the 300- m depth (Supplementary Fig. 10a). This result is consistent with all IW_EXPs (Supplementary Fig. 10b–d), suggesting a release of heat from the deep ocean to the surface. In the eastern Pacific, a region of weak ocean stratification and intense upwelling, heat is much more efficiently transferred to the surface. The surface warming is further amplified by the Bjerknes feedback so that the surface warming is greater than the subsurface warming though the warming originated from the deep ocean warming. Therefore, the difference in the vertical temperature gradient is positive in the upper layer, suggesting enhanced ocean stratification (Supplementary Fig. 10e–h).

Enhanced stratification leads to strong air-sea coupling, indicating that the surface-layer responses to a given wind anomaly are amplified. To show enhanced surface-layer responses, we performed a linear regression analysis of the zonal current response to the zonal wind stress anomaly (120°E-90°W) during DJF (Fig. 4a). During the PD period, the current response to wind stress shows a stronger eastward flow in the surface layer, which gradually weakens with depth. Below the thermocline, there is a westward current response, indicating a vertically baroclinic structure. In the restoring period, the eastward current response in the surface layer is intensified (Fig. 4b), indicating stronger zonal advective feedback. Note that the stronger zonal current response is particularly distinctive in the eastern Pacific. In addition, the equatorial upwelling and thermocline responses are also intensified in the eastern Pacific (Supplementary Fig. 11). This enhanced surface-layer feedback plays a critical role in the enhanced eastern Pacific SST variability. The IW_be700 also simulates enhanced feedback in the surface layer (Fig. 4c). This result suggests that deep ocean warming leads to stronger oceanic responses in the eastern Pacific upper layer. The other two IW_EXPs show similar changes in the zonal current response (Supplementary Fig. 12a, b), supporting strong robustness.

**Fig. 4: The changes in El Niño-Southern Oscillation feedback.**

The ocean stratification is responsible for the intensified surface-layer feedback, but it cannot account for the eastward shift of the responses. Previous studies suggested that the El Niño-like mean conditions not only thermodynamically enhance the response of precipitation to equatorial SST but also provide favorable conditions for the shift of anomalous atmospheric heating due to increased mean trade wind convergence³⁹. For example, when the eastern Pacific is in a more moist state, the response of wind and precipitation to SSTA shifts eastward⁴⁰. The eastward shift of the wind response to ENSO anomalies strengthens the thermocline and zonal advection feedback, ultimately increasing the eastern Pacific SST variability^41,42. As mentioned earlier, the El Niño-like warming induces an eastward shift of the ascending branch of the Walker circulation and an equatorward shift of the ITCZ (Fig. 1), which strengthens the shift in the surface-layer current response.

To analyze the eastward shift of the ENSO feedback, we calculated the regressed precipitation and wind onto the Niño3 SSTA during DJF (Fig. 4d). As expected, the precipitation pattern shows positive anomalies associated with westerlies in the equatorial Pacific. The changes in the precipitation and wind response from the PD period indicate an eastward extension of the response, with positive anomalies observed in the eastern Pacific and negative anomalies in the western Pacific (Fig. 4e). The eastward shift of the wind response to the SSTA contributes to the enhanced Ekman feedback. In addition, the strong equatorial wind stress response enhances the thermocline feedback by intensifying the equatorial Kelvin wave responses. In addition, increases in the meridional slope of the wind stress (Fig. 4e), resulting in a stronger off-equatorial Rossby wave response. The IW_EXPs consistently simulate the precipitation response eastward shifted compared to the PD period (Fig. 4f and Supplementary Fig. 12c, d). The pattern of precipitation change is almost identical to that of the restoring period, with a pattern correlation of 0.95. These results suggest that the warming of the deep ocean produces the ENSO system eastward in the restoring period. Note that the westerly anomaly is stronger at 150-120°W, consistent with where the surface current response increases (Fig. 4b, c). In short, the occurrence of more EP El Niño can be attributed to the eastward shift of the ENSO feedback system caused by El Niño-like warming.