1. Introduction
El Niño–Southern Oscillation (ENSO) is a globally impactful expression of ocean–atmosphere coupling in the tropical Pacific (Bjerknes 1966, 1969; Philander 1983; Banholzer and Donner 2014; Timmermann et al. 2018). ENSO involves abnormal values of sea surface temperature (SST) in the equatorial central and eastern Pacific, with SSTs in this region being abnormally large during the warm phase (El Niño) and abnormally small during the cold phase (La Niña). Human activities, especially fossil fuel combustion and deforestation, have raised atmospheric greenhouse gas concentrations by an unprecedented rate since the middle of the nineteenth century. ENSO activity has exhibited both qualitative and quantitative changes as the background climate has warmed in response to rising greenhouse gas concentrations (Ashok et al. 2007; Kao and Yu 2009; Weng et al. 2009; Yeh et al. 2009; Yu and Kim 2013; Capotondi et al. 2015; Marathe et al. 2015; Xie and Jin 2018; Cai et al. 2020b), with greater ENSO variability over recent decades attributed with increasing confidence to the influence of greenhouse gases (Cai et al. 2023). Given the tremendous global-scale impacts of ENSO on weather, climate, and environment, it is necessary to understand the characteristics and mechanisms of how ENSO responds to greenhouse gas forcing.
Much work has been devoted to understanding and predicting ENSO and its variability, including extensive study of the underlying dynamical and thermodynamic mechanisms (Neelin et al. 1998; Wang and Picaut 2004). Among these, the positive Bjerknes feedback is central to explaining the mechanisms that trigger and maintain El Niño (Bjerknes 1966). Shifts between positive and negative ENSO phases can be explained by negative feedback loops such as the recharge–discharge oscillator (Jin 1997a,b), in which ENSO variability is linked to the build-up and occasional discharge of warm water in the equatorial Pacific, or the delayed oscillator (Schopf and Suarez 1988; Suarez and Schopf 1988; Battisti and Hirst 1989), in which ENSO warm anomalies are enhanced by downwelling waves and then suppressed about 6 months later by upwelling equatorial waves reflected at the western boundary. The top-of-atmosphere energy imbalance associated with increases in atmospheric greenhouse gas concentrations also plays a role in modulating ENSO variability. This imbalance is linked to ocean heat uptake through changes in the ocean surface heat flux, rising evaporation and the release of latent heat during precipitation formation, and associated changes in near-surface winds. These processes induce changes in SST throughout the tropics, often manifesting as enhanced equatorial warming (Liu et al. 2005), which in turn affects ENSO variability in the central and eastern tropical Pacific (Cai et al. 2020b; Freund et al. 2020). The characteristics of this response are sensitive to the simulated thermal structure of the upper ocean (Yeh et al. 2010).
ENSO events involve changes in the equatorial Pacific thermal structure and the atmospheric circulation above it, which in turn impact global weather patterns through teleconnections (Trenberth and Hurrell 1994; Alexander et al. 2002; McPhaden et al. 2006; Huang et al. 2023). When forced by increasing greenhouse gas concentrations, most climate models indicate a weakening of the Walker circulation in the atmosphere above the equatorial Pacific (Tokinaga et al. 2012a,b; Cai et al. 2020b, 2021), which can be explained by energetic constraints on the atmospheric water cycle (Held and Soden 2006). As a consequence of these constraints, atmospheric vertical motion is expected to weaken over tropical convective regions, such as the western Pacific warm pool, leading to a slowdown in the tropical overturning circulation (Zhang and Song 2006). Other factors, such as enhanced equatorial warming (e.g., Liu et al. 2005) and increases in rainfall driven by warmer sea surface temperatures (Vecchi et al. 2006), further contribute to slowing the Walker circulation. Associated changes in wind stress (i.e., a weakening of the easterly trade winds) act to flatten the east–west tilted thermocline (Yeh et al. 2009), further amplifying equatorial warming, especially in the eastern Pacific (Xie et al. 2010; Cai et al. 2020a). These changes create a more El Niño–like mean state similar to the positive interdecadal Pacific oscillation (IPO), which has historically been associated with enhanced ENSO variability (Salinger et al. 2001).
Confidence in projections of future changes in ENSO activity is limited by persistent biases in climate model simulations of the tropical Pacific and its variability, including biases in the double intertropical convergence zone (ITCZ), warm SSTs in the southeastern Pacific, and the intensity and shape of the eastern Pacific cold tongue (Hwang and Frierson 2013; Li and Xie 2014; Planton et al. 2021). These biases affect the simulated amplitude, period, asymmetry, and spatial distribution of ENSO events. Among these, biases in the SST distribution in the eastern tropical Pacific are especially impactful, as SST anomalies within this region are a primary characteristic of ENSO variability (Covey et al. 2003; Huang et al. 2007; Richter and Xie 2008; Zheng et al. 2011; Toniazzo and Woolnough 2014; Wang et al. 2014; Xu et al. 2014; Richter 2015).
Projections of future changes in ENSO variability are further complicated by discrepancies in not only the magnitude but even the sign of changes between transient and equilibrium simulations of climate change. Studies of the ENSO response to global warming have been conducted under a range of scenarios but have yet to reach a consensus on how ENSO dynamics and teleconnections respond to these changes (Michel et al. 2020; Taschetto et al. 2020). Moreover, the extent to which discrepancies in the projected ENSO response between transient (Cai et al. 2021, 2022) and abrupt doubling simulations (Callahan et al. 2021) reflect model biases or expressions of internal variability (Stevenson 2012) versus actual mechanistic differences remains unclear.
Some simulations of twenty-first-century climate change suggest that increasing CO2 strengthens ENSO because the oceanic dynamical response (i.e., currents and thermocline) becomes more sensitive to wind forcing (e.g., Kim and Jin 2010; Chen et al. 2017; Wang et al. 2019). This increased sensitivity may be explained by enhanced upper-ocean stratification associated with a surface maximum in warming and a thinner mixed layer due to weaker winds (Yeh et al. 2010), which both suggest a shoaling of the thermocline toward the surface. Increases in the mean stratification may also intensify ocean–atmosphere coupling and associated positive feedbacks, especially in the eastern Pacific (Carréric et al. 2020), and thereby further enhance ENSO activity (Cai et al. 2018). However, negative feedbacks associated with thermodynamic damping (i.e., SST increases reduce net heat flux into ocean and vice versa) are also projected to strengthen across models owing to increased evaporation fluxes and increased cloudiness (Knutson and Manabe 1995). Callahan et al. (2021) linked decreases in ENSO amplitude under equilibrated greenhouse forcing in several climate models to weaker zonal temperature gradients across the tropical Pacific and speculated that ENSO variability was suppressed in these simulations because the negative impacts of stronger thermodynamic damping outweighed the positive impacts of increased dynamical sensitivity to wind stress. Their results suggest that increased CO2 forcing may reduce ENSO variability in the long term.
Here, we adopt the Bjerknes–Jin (BJ) index linear framework (Jin et al. 2006, 2020) for ENSO stability analysis in a model with reduced SST bias in the eastern tropical Pacific (Ma et al. 2019). Prescribed CO2 concentrations in the model are increased incrementally, and the analysis is conducted for periods after the model reaches quasi-equilibrium with the revised forcing. This approach supports a detailed and systematic diagnosis of both the changes in ENSO variability and the mechanisms behind those changes, allowing us to better identify bifurcations and nonmonotonic changes in the mechanisms that amplify or suppress ENSO variability. Specifically, we address this question: To what extent do changes in ENSO variability under increasing CO2 forcing reflect changes in the dynamical barrier (how many events are triggered) versus changes in damping (events are triggered but are more difficult to grow and maintain)?
2. Model setup and experimental approach
a. Model experiments
Coupled model simulations have been conducted using a version of the Community Earth System Model (CESM1) with revised topography (CESM1Topo; Ma et al. 2019). CESM1Topo is a coupled global climate model run at a horizontal resolution of 0.9° × 1.25° for the atmospheric and land components and a nominal resolution of 100 km for the ocean and sea ice components. A refined land surface topography is specified at 0.47° × 0.63° over South America and Africa (Fig. 1). The atmospheric component is the Community Atmosphere Model version 4 (CAM4; Neale et al. 2013) with a finite-volume dynamical core and 26 vertical levels. The ocean component is the Parallel Ocean Program version 2 (POP2; Danabasoglu et al. 2012) with 60 vertical levels, 20 of which are concentrated within 200 m of the surface. The top layer of the ocean model extends from the surface to 10-m depth. The land component is represented by the Community Land Model version 4 (CLM4; Oleson et al. 2008). All runs use the same preindustrial (year 1850) radiative and boundary conditions but with different levels of atmospheric CO2 (Table 1). Atmospheric CO2 concentrations are increased abruptly at the beginning of each simulation, with a series of eight steps and a range from year 1850 (284.7 ppmv) to 8 times that amount (2277.6 ppmv). Each experiment is initialized from the previous quasi-equilibrium result as described by Ma et al. (2019).
Changes in model topography over South America between CESM1.2.1 and CESM1-Topo. Changes in mean sea surface temperature between control simulations in both configurations are also shown, along with the bounds of the Niño-3 region.
Citation: Journal of Climate 37, 3; 10.1175/JCLI-D-23-0079.1
List of the experiments, specified volume mixing ratios of CO2, and conditions in the tropical Pacific under quasi-equilibrium. Metrics include the standard deviation of deseasonalized SST anomalies in the cold tongue (Niño-3) region (σSSTA,CT), the difference in time-mean SST between the warm pool (Niño-4) and cold tongue regions (ΔSSTWP−CT), the mean wind stress in the central Pacific (Niño-3.4) region (τx,CP), and the change in top-of-atmosphere energy imbalance between years 151–200 and 251–300 of the simulation.
Each experiment is integrated for a minimum of 300 years. The last 250 years are used for analysis of the control simulation to further constrain the influence of internal variability. The last 150 years are used for analysis of the perturbed CO2 experiments under the assumption that the climate state of the coupled tropical Pacific has achieved quasi-equilibrium by this point. Two steps are taken to evaluate this quasi-equilibrium assumption and its potential impacts on the Bjerknes–Jin analysis results. First, the 3×CO2 experiment is extended to 600 years and the analysis steps are applied to the full simulation period. Second, a separate 3×CO2 experiment is initialized from the 4×CO2 quasi-equilibrium state and run for 300 years. The results of these tests are described in section 3b.
With respect to ENSO simulation, the salient difference between the original CESM1 and CESM1Topo is the topography along the western coast of South America, which has been modified to steepen the slopes and better represent the actual topography (Fig. 1). These topographic improvements reduce warm biases in the upwelling region along the eastern boundary of the southern tropical Pacific by improving the nearshore wind field. Improvements are particularly pronounced in the southeastern tropical Pacific along the west coast of South America (Fig. 1), where warm biases in simulated sea surface temperature result primarily from insufficient ocean dynamical upwelling (Ma et al. 2019).
CESM1Topo produces improvements in the amplitude and frequency of ENSO variability relative to the original CESM1 without modified topography (CESM1Orig), as shown by the power spectrum of SST variability in the Niño-3 region (Fig. 2). Both the magnitude and distribution of power in the 2–8-yr ENSO band are in better agreement with observations in CESM1Topo relative to CESM1Orig, although CESM1Topo does not reproduce the bimodal structure of the observed spectrum.
(a) Power spectra based on the ERSST (1854–2020) and TropFlux (1979–2018) observational SST analyses and the control simulations using CESM1Orig and CESM1Topo. (b) Changes in average power within periods of 2–8 years in the CESM1Topo control simulation and the last 150 years of the perturbed CO2 experiments (Table 1). ENSO variability for these power spectra is based on Niño-3 (Fig. 1).
Citation: Journal of Climate 37, 3; 10.1175/JCLI-D-23-0079.1
Further tests conducted based on the CLIVAR (Climate and Ocean: Variability, Predictability and Change) ENSO Metrics Package (Planton et al. 2021) indicate that CESM1Topo compares well with both CESM1Orig and the average CMIP6 model in most aspects of the tropical Pacific mean state and basic ENSO characteristics (Fig. 3). Uncertainties in these metrics are primarily associated with internal variability. For CESM1Topo, these uncertainties are estimated as the minimum and maximum values obtained from 1000 randomly selected 50-yr periods with replacement from the 250-yr simulation. Owing to time series length, comparable uncertainty estimates cannot be calculated for the observational benchmark (40 years) or CESM1Orig (50 years). We therefore assess significant differences between CESM1Orig and CESM1Topo as those for which the CESM1Orig result falls outside the CESM1Topo bootstrap interval.
Results of selected evaluations of the tropical Pacific mean state and basic ENSO characteristics as represented by CESM1Orig and CESM1Topo following the CLIVAR ENSO Metrics Package (Planton et al. 2021): time-mean meridional means (5°S–5°N) of (a) equatorial SST, (b) zonal wind stress, and (c) precipitation; (d) zonal distribution of ENSO-related SST anomalies; (e) seasonal cycle of SST variability in the Niño-3.4 region; (f) amplitude and (g) skewness of SST variability in the Niño-3.4 region; and (h) evolution of SST anomalies in Niño-3.4 regressed onto anomalies during December of ENSO events. ENSO variability for these metrics is defined with respect to Niño-3.4 for comparability with CMIP6 results published by CLIVAR. Uncertainties for CESM1Topo in (a)–(c) and (e)–(g) are based on bootstrapping 50-yr periods from the 250-yr simulation. Similar estimates cannot be calculated for the observational benchmark (40 years) or CESM1Orig (50 years). Uncertainties in (d) and (h) are twice the standard error of the regression slope.
Citation: Journal of Climate 37, 3; 10.1175/JCLI-D-23-0079.1
The zonal distribution of sea surface temperature is in good agreement with observations (Fig. 3a); positive biases in zonal wind stress in the western Pacific (Fig. 3b), excess precipitation in the eastern Pacific (Fig. 3c), and the eastern Pacific double ITCZ (not shown) are all reduced in CESM1Topo relative to CESM1Orig. With respect to basic ENSO processes, CESM1Topo better represents the springtime minimum in ENSO variability relative to CESM1Orig (Fig. 3e) and exhibits smaller biases than the CMIP6 average in ENSO amplitude, asymmetry, and life cycle metrics (Figs. 3f–h). Although CESM1Topo underestimates the sensitivity of zonal wind stress (by 28%) and net surface heat flux (by 23%) to SST anomalies and overestimates the relationship between SSH and SST anomalies (by 31%) in the Niño-3.4 region relative to observational estimates, these biases are smaller than those produced by the average CMIP6 model (Planton et al. 2021). Like CESM1Orig, CESM1Topo exhibits a relatively large bias in the ENSO pattern of sea surface temperature anomalies (Fig. 3d), with a response that is weaker than observed in the eastern Pacific and stronger than observed in the western Pacific. CESM1Topo also underestimates the wintertime maximum in ENSO variability (Fig. 3e) and overestimates the relationship between zonal wind stress and sea surface height in the Niño-3.4 region by 32%, slightly more than the average CMIP6 bias in this metric.
b. Bjerknes–Jin stability analysis
Jin (1996, 1997a,b) developed a simple paradigm of the recharge–discharge oscillator (RO) model for ENSO that allows the ENSO growth rate to be explicitly quantified and decomposed (Jin et al. 2006). This linear framework is referred to as Bjerknes–Wyrtki–Jin (BWJ) stability analysis. In this work, we neglect changes in ENSO periodicity (Lu et al. 2018; Jin et al. 2020) and focus on changes in the linear growth rate as represented by the Bjerknes–Jin index (BJI), which encompasses all major processes that influence the linear growth rate of ENSO variability.
The coefficients in Eq. (1) are calculated through a series of linear regressions between anomalies in key variables describing the state of the tropical Pacific (Jin et al. 2020). All anomalies are calculated relative to the mean annual cycle over years 150–300 of the corresponding simulation. For example, the TH term contains parameters representing the sensitivity of zonal-mean wind stress to Niño-3 temperature anomalies (μa), the sensitivity of the zonal thermocline shape to variations in zonal-mean wind stress (βh), the sensitivity of subsurface (75 m) temperature anomalies to variations in the eastern Pacific thermocline depth (ah), and the mean strength of upwelling in the Niño-3 region. Formal definitions and descriptions of all variables follow those provided by Jin et al. (2020).
Meridional mean (5°S–5°N) temperature changes regressed onto a Niño-3 temperature anomaly of 0.75°C for (a) the CESM1Topo control simulation and (b) the ORAS5 ocean reanalysis. Differences from the control simulation are shown for the (c) 1.5×CO2, (d) 2×CO2, (e) 3×CO2, (f) 4×CO2, (g) 6×CO2, and (h) 8×CO2 experiments. The Niño-3 region is marked by gray dashed lines in each panel.
Citation: Journal of Climate 37, 3; 10.1175/JCLI-D-23-0079.1
Figure 5a shows results for the thermocline feedback TH, Niño-3 temperature anomaly growth rate R, slow adjustment scale −ε, and BJI based on the CTL simulation (Table 1) assuming preindustrial CO2 concentrations. The analysis is conducted on rolling 50-yr windows shifted by 1 month. Time variations in the Niño-3 SST anomaly power spectrum based on wavelet analysis (Torrence and Compo 1998) are also shown (Figs. 5b,c) to evaluate the extent to which internal variations in the linear growth rates relate to variations in ENSO activity. Larger values of R and BJI correspond well to increases in the 2–8-yr variance of SST anomalies in the Niño-3 region, indicating that larger values of R and BJI are associated with increases in ENSO activity. Simulated internal variations in the BJI and its components are on the decadal to multidecadal scale. This internal variability in the CTL simulation is dominated by variations in the TH term (Fig. 5a) with minor supporting contributions from the advective terms and a weak compensation from TD (not shown).
Time variations in (a) 50-yr rolling TH, R, −ε, and BJI, (b) the continuous wavelet power spectrum of the Niño-3 index, and (c) Niño-3 SST variance in the 2–8-yr band based on results from the CESM1Topo control simulation. Periods with enhanced variance in the 2–8-yr band are shaded pink in (a) and (c).
Citation: Journal of Climate 37, 3; 10.1175/JCLI-D-23-0079.1
3. Results
a. Changes in ENSO variability
Changes in ENSO variability under increasing levels of CO2 forcing are evaluated with respect to SST anomalies in the Niño-3 and Niño-3.4 regions for years 151–300 of each simulation. The standard deviation of deseasonalized Niño-3 SST anomalies is 0.89°C in the CTL simulation, peaks at 0.94°C for the 1.25×CO2 and 1.5×CO2 simulations, and then decreases with increasing CO2 (Table 1). The reduction in ENSO variability becomes increasingly pronounced at larger CO2 forcings, with a loss of approximately 25% between the 1.5×CO2 and 8×CO2 scenarios. These changes are consistent with changes in the Niño-3 power spectra (Fig. 2b), which confirm that both the initial increase and subsequent decrease in variance with increasing CO2 reflect changes in the 2–8-yr ENSO band. The total number of warm and cold ENSO events over years 151–300 of each simulation based on 3-month running mean SST anomalies in the Niño-3.4 region (Fig. 6) remains steady at 72–75 (slightly less than one event per 2 years) up to 2×CO2 before declining sharply, with only 39 events during years 151–300 of the 8×CO2 simulation. The frequency of extreme ENSO events peaks at 9 in the 1.5×CO2 simulation and becomes very rare as the CO2 forcing increases. The average amplitude of warm (El Niño) events peaks at more than 1.3°C for the 1.25×CO2 and 1.5×CO2 simulations before declining steadily to less than 1°C in the 8×CO2 simulation. However, no systematic change is noted in the average magnitude of La Niña events (1.1°–1.2°C in all simulations).
Changes in the frequency of standard (light colors) and extreme (dark colors) warm (El Niño; red) and cold (La Niña; blue) ENSO events during years 151–300 under nine levels of CO2 forcing. ENSO events are defined by the 3-month running mean Niño-3.4 index exceeding ±0.5°C for at least 5 consecutive months. Extreme ENSO events are defined as the peak 3-month mean anomaly exceeding ±2°C during an identified ENSO event.
Citation: Journal of Climate 37, 3; 10.1175/JCLI-D-23-0079.1
These changes in basic ENSO statistics indicate that ENSO events continue to occur at roughly the same rate up to at least 2 × CO2 before declining with further increases in CO2 but become more difficult to grow at forcings larger than about 1.5 × CO2. The latter change mainly affects the growth of warm events, with little change in the average amplitude of cold events. This distinction further motivates detailed analysis of the growth rate and stability with respect to anomalies in the more eastward-located Niño-3 region, as SST anomalies associated with warm ENSO events are shifted eastward on average relative to those associated with cold events. Current atmospheric CO2 levels (∼420 ppmv) are roughly in line with the 1.5 × CO2 concentration (427 ppmv), with the 1.25 × CO2 level having been passed in the early 1990s. The initial increase in ENSO variance is therefore consistent with increases in ENSO activity during the latter half of the twentieth century, which can be attributed at least in part to greenhouse gas forcing (Cai et al. 2023).
b. Bjerknes–Jin stability analysis
As an initial clue to the mechanisms behind changes in the occurrence frequency and intensity of El Niño events, we refer back to the zonal distribution of ocean temperature anomalies (Fig. 4). These anomaly distributions, which are regressed onto a constant Niño-3 anomaly of 0.75°C, show that a similar magnitude of ENSO event requires an increasingly large discharge of warm water from the western Pacific as atmospheric CO2 increases. Moreover, warm ocean temperature anomalies in the Niño-3 region are increasingly confined to the eastern boundary around 50-m depth as CO2 concentrations increase beyond 2 × CO2. These changes in ENSO-related temperature structure in the upper ocean suggest that the mechanisms behind changes in ENSO variability may derive from the thermocline feedback and its interactions with increasing static stability.
The results of Bjerknes–Jin stability analysis (Fig. 7) further support this conclusion. The ENSO anomaly growth rate (R; Fig. 7b) also suggests a trend of first increasing and then decreasing, determined to leading order by changes in the thermocline feedback (TH) and supplemented by changes in thermodynamic damping (TD), especially at larger values of CO2 (Fig. 7a). Changes in the TH term largely follow changes in R, with an initial increase and a subsequent decrease pivoting around 1.5×CO2, while the amplitude of the TD term consistently increases with increasing CO2. However, changes in these terms are mostly within the range of internal variability up to 2×CO2, here illustrated as the interquartile range of 50-yr rolling results in each simulation (error bars in Fig. 7). Combination with the slow adjustment scale −ε creates somewhat larger separation in the stability index BJI, with the amplitude of −ε first increasing (indicating faster basin-scale adjustment) and then decreasing. Although the 1.75×CO2 experiment appears as an outlier in this series of experiments, it is nonetheless within the range of internal variability in the neighboring 1.5×CO2 and 2×CO2 simulations. The results of this simulation therefore highlight the potential confounding influence of internal variability in stability analysis conducted on 150-yr periods of model simulation (e.g., Zheng et al. 2018; Maher et al. 2018), which are not substantially longer than the multidecadal variations seen in Fig. 5. We therefore adopt the conservative approach of treating changes between the CTL and 2×CO2 primarily as expressions of internal variability at near-present-day CO2 concentrations rather than responses to external forcing. Although an increase in the ENSO growth rate at 1.25–1.5 times preindustrial CO2 is consistent with recent increases in ENSO activity (Cai et al. 2023), resolving the quasi-equilibrium response of the growth rate and stability index to these smaller increments in CO2 forcing will require larger ensembles or longer runs.
Results of the Bjerknes–Jin index analysis for all CESM1Topo experiments: (a) component terms for thermocline feedback (TH), advective feedbacks (ZA + MA + VA), dynamical damping (DD), and thermodynamic damping (TD); (b) the temperature anomaly growth rate R, −ε, and the Bjerknes–Jin index [BJI = (R − ε)/2]; and (c) the dynamical damping and its zonal, meridional, and vertical components. The scale of internal variability in each term and each simulation is shown as the interquartile range (middle 50%) of rolling 50-yr results.
Citation: Journal of Climate 37, 3; 10.1175/JCLI-D-23-0079.1
Each simulation has reached approximate equilibrium with the CO2 forcing by year 150 as defined by stability in the global-mean surface air temperature and top-of-atmosphere energy budget, with differences in mean top-of-atmosphere energy fluxes between years 251–300 and years 151–200 less than 0.1 W m−2 for most simulations (Table 1), well within the interannual standard deviation in CTL (0.21 W m−2). Larger differences are found for the largest CO2 forcings, but even these imbalances are less than 0.3 W m−2. To evaluate the potential impacts of any residual disequilibrium after year 150 on the results of the Bjerknes–Jin stability analysis, we have extended the 3×CO2 run by an additional 300 years and conducted an additional companion experiment in which the 3×CO2 experiment was initialized from the 4×CO2 result rather than 2×CO2. The results are shown in Fig. 8. While R is within the range of internal variability essentially from the beginning of the 3×CO2 simulation, ε is subject to an extensive adjustment period (Fig. 8a). This adjustment period is expected given that ε is calculated as a memory effect of SST (Jin et al. 2020) and the initial period of the simulation is characterized by strong temperature trends. As a test, we repeat the Bjerknes–Jin stability analysis by linearly detrending each rolling 50-yr period (light dotted lines). Results for ε based on the raw and detrended outputs converge over the first 100–150 years of the simulation, achieving good agreement by year 150 (which includes outputs from year 125 to year 174). Results during the analysis period (unhatched region) and 300-yr extension (blue stippled region) show internal variability in R, ε, and BJI on a similar scale to that identified in CTL (Fig. 5). The results of the 3×CO2 experiment initialized from 4×CO2 (Fig. 8b) show a shorter adjustment period but similar magnitudes of internal variability (Fig. 8c). The centers of the distributions for all three variables are stable across the core analysis period, the 300-yr extension, and the companion experiment initialized from 4×CO2. This indicates that residual disequilibrium after year 150 does not substantially affect the results of the Bjerknes–Jin analysis in the 3×CO2 experiments, and suggests that the stability of ε and its difference from results based on detrended data may serve as a useful metric of disequilibrium in the other simulations. Applying this criterion suggests the elimination of only 15 years from the 8×CO2 simulation, which is applied to all analyses concerning the Bjerknes–Jin results. This 15-yr period is at the end rather than the beginning of the 8×CO2 simulation, suggesting that it results from the development of an instability in the simulation rather than residual disequilibrium at year 150.
Time variations in 50-yr rolling R, −ε, and BJI for rolling 50-yr intervals of (a) the 3×CO2 experiment initialized from the 2×CO2 experiment and (b) a companion experiment in which the 3×CO2 simulation is initialized from the 4×CO2 experiment. (c) Distributions of R, −ε, and BJI during the spinup [gray cross-hatched region in (a)], analysis, and extension [blue stippled region in (a)] periods of the 3×CO2 experiment. Dotted lines in (a) and (b) show variations based on detrending all variables in each 50-yr interval before conducting the Bjerknes–Jin index analysis. Light shading in (c) shows the 10th–90th percentile range from the time series in (b).
Citation: Journal of Climate 37, 3; 10.1175/JCLI-D-23-0079.1
The magnitude of the dynamical damping (DD) term changes little across the simulations (Fig. 7a). The DD term is defined by changes in the mean state as described by Eq. (1), including mean zonal damping (U damping), mean meridional damping (V damping), and mean upwelling damping (W damping). As shown in Fig. 7c, changes in these terms indicate that reductions in mean upwelling (negative) are compensated by reductions in zonal damping (positive), with little change in meridional damping. This is due to a weakening of westward surface currents and equatorial upwelling under higher CO2 scenarios (Fig. 9), which both correspond to a net decrease in wind stress. Mean wind stress in the central Pacific remains relatively constant up to about 2 × CO2 before decreasing by about 20% between the 2×CO2 and 8×CO2 scenarios (Table 1). Advective terms, including Ekman divergence, are also relatively stable across the nine experiments (Fig. 7a; ZA + MA + VA). As with DD, this stability results from offsetting changes in the zonal and vertical advective terms. The net effect of weaker upwelling acting on a stronger vertical temperature gradient (Fig. 9) produces only slight increases in VA, which are essentially compensated by reductions in the ZA term due to a weaker east–west temperature gradient (Fig. 9; see also Table 1).
(a) Mean temperatures relative to the Niño-3 region (θ − TE) and circulation fields for the CESM1Topo control simulation. Differences in both fields relative to the control simulation are shown for the (b) 1.5×CO2, (c) 3×CO2, and (d) 6×CO2 simulations. Vertical velocities are scaled to emphasize the distribution of dynamical upwelling and its changes with increasing CO2.
Citation: Journal of Climate 37, 3; 10.1175/JCLI-D-23-0079.1
To evaluate the extent to which the results of our analysis extend to other models, we analyze results from a selection of six global climate models (Table 2) that contributed outputs to phase 6 of the Coupled Model Intercomparison Project (CMIP6; Eyring et al. 2016). Models are selected based on whether the necessary outputs were provided for all four of the piControl, abrupt-0p5xCO2, abrupt-2xCO2, and abrupt-4xCO2 simulations. The latter three simulations, while conceptually similar to our experiment set (Table 1), were typically run for only 150 years (with the exception of MIROC6, which ran each experiment for 250 years) and were all initialized from the corresponding piControl simulations rather than the previous step. As a result, these simulations cannot be assumed to be in quasi-equilibrium. To partially address this issue, we conduct the analysis on 50-yr rolling intervals from the last 100 years of each simulation (i.e., intervals centered on years 75–124 for a 150-yr simulation), as a compromise between uncertainties related to disequilibrium and uncertainties related to internal variability.
List of the CMIP6 models evaluated in this work, including the nominal resolution and top-level depth of their ocean components.
Figure 10 shows the results of the Bjerknes–Jin stability analysis applied to the CMIP6 models listed in Table 2. Changes in TH (Fig. 10a) are consistent with the idea that the thermocline feedback first intensifies and then weakens with increasing CO2, with peak values for most models associated with either the piControl or the 2×CO2 experiment. The MRI-ESM2.0 is an exception, with the strength of TH increasing monotonically from low 0.5×CO2 to 4×CO2. Changes in R (Fig. 10b) again typically follow those in TH, including for MRI-ESM2.0, although the peak value for R is more likely to be shifted to 2×CO2. Changes in −ε (Fig. 10c) are more diverse and show less agreement with the CESM1Topo results, often presenting as a mirror image of the changes in TH, with minimum amplitudes for piControl and larger values for both smaller and larger values of CO2. As the sum of R and −ε, BJI thus follows a similar pattern (Fig. 10d), with smaller values for 0.5×CO2 and 4×CO2 and larger values for piControl and 2×CO2. This pattern also holds for MRI-ESM2.0, for which increases in the magnitude of ε as CO2 increases from piControl compensate the steady increases in TH and R.
Bjerknes–Jin stability analysis results for the (a) TH term, (b) growth rate R, (c) slow adjustment scale −ε, and (d) BJI. Results of the analysis are shown based on rolling 50-yr intervals from the full piControl simulation and the last 75 years of the abrupt 0.5×CO2, 2×CO2, and 4×CO2 simulations. Details of the evaluated models are listed in Table 2.
Citation: Journal of Climate 37, 3; 10.1175/JCLI-D-23-0079.1
Increases in the (negative) magnitude of the stability index can be interpreted as steepening the barrier to ENSO occurrence and persistence, while decreases in the positive growth rate can be interpreted as suppressing the intensity of events when they occur. These interpretations are especially relevant for warm events as the analysis is conducted for anomalies in the Niño-3 region. Larger negative values of the BJI are thus consistent with the reduced occurrence of both standard and extreme warm events as CO2 increases (Fig. 6), while smaller positive values of R (tending to zero for the 8×CO2 simulation) are consistent with the sharp decrease in average warm event amplitude at forcings larger than 2 × CO2. Accordingly, addressing the primary motivating question for this work requires detailed analysis of the terms that contribute most to changes in R and BJI: TH, TD, and ε.
The TD term describes the extent to which intensification of SST anomalies is limited by the damping effect of changes in surface air–sea heat fluxes (Kim and Jin 2010). Interannual variations in air–sea heat fluxes are generally dominated by shortwave radiation and latent heat flux (Wang and McPhaden 2000; Santoso et al. 2011), a conclusion that also appears to hold for changes in thermodynamic damping of ENSO events under increasing CO2 (Callahan et al. 2021). Due to the exponential dependence of latent heat flux on SST stemming from the Clausius–Clapeyron relation (Ferrett and Collins 2019), TD strengthens as a result of mean SST warming. This dependence implies a stronger negative latent heat feedback with increasing SST as CO2 increases (Lübbecke and McPhaden 2013). Studies have also highlighted the role of cloud feedbacks in the tropical Pacific in amplifying TD as climate warms (Zhao and Fedorov 2020; Callahan et al. 2021). Our results are consistent with these prior conclusions, implicating changes in the surface latent heat flux and cloud radiative effects as the primary reasons why TD increases with increasing CO2. As a result, and noting that both internal variations and the forced response to changes in CO2 are largely determined by variations in TH and ε, the following section focuses on the possible mechanisms behind changes in these terms.
4. Discussion: Possible mechanisms
The results in the previous section indicate that variations in the ENSO growth rate R and the stability index BJI mainly depend on changes in the thermocline feedback term. Although TH dominates both internal variability and the response to external forcing, its relationships with changes in the mean state are distinctly different between these two situations (Fig. 11). Negative anomalies in TH based on internal variability in the CTL simulation are consistent with the negative phase of the interdecadal Pacific oscillation, with negative anomalies in surface and subsurface temperature in the eastern tropical Pacific, an enhanced zonal sea surface temperature gradient, a deeper thermocline in the west with shoaling in the east, and stronger westward zonal wind stress in the central Pacific. By contrast, differences between the 3×CO2 and CTL experiment, associated with the same magnitude of change in the TH term as those regression-based results, are both qualitatively opposite and quantitatively much larger. The difference in the mean state between these two simulations indicates enhanced surface and subsurface warming in the east resulting in a reduced zonal gradient (see also Table 1), a flatter thermocline with shoaling in the west, especially off the equator, and weaker westward zonal wind stress in the central and eastern Pacific. The scale of these changes is approximately 5 times larger than those associated with the same magnitude of TH anomaly arising from internal variability. Changes in TH can therefore not be explained by changes in the mean state associated with internal variations in TH.
Distributions of anomalies associated with a −0.80 yr−1 deviation in the TH term based on (a) regression of internal variations in the CTL simulation and (b) the difference between the 3×CO2 and CTL simulation. Anomaly distributions are shown for SST, subsurface (75 m) temperature, thermocline depth, zonal wind stress, and surface freshwater flux, shown from top to bottom. There is a factor of 5 difference in the color scales between (a) and (b).
Citation: Journal of Climate 37, 3; 10.1175/JCLI-D-23-0079.1
A more direct assessment can be conducted by breaking down the TH term according to its definition [Eq. (1); Fig. 12]. As introduced in section 2b, the TH term consists of four components. The first (μa) indicates the strength of the linear sensitivity of zonal-mean wind stress to sea surface temperature anomalies in the eastern tropical Pacific. The second (βh) indicates the strength of the linear response of changes in thermocline shape (i.e., the difference in the thermocline depth between the western and eastern tropical Pacific) to anomalies in zonal-mean zonal wind stress. The third (ah) indicates the strength of the linear relationship between thermocline depth anomalies and subsurface (75 m) temperature anomalies in the eastern tropical Pacific. The fourth and final term represents the strength of the mean upwelling in the Niño-3 region in the eastern tropical Pacific. Figure 12 breaks down the relative contributions of these four terms both to internal variability (across the quasi-equilibrium portions of all CESM1Topo simulations) and to changes in the TH term relative to its value in the CTL simulation. Among the four terms, βh (thermocline sensitivity to wind stress) makes the largest contribution to internal variations in TH (Fig. 12a), followed by ah (wind stress sensitivity to SST anomalies in Niño-3). The expression of these relationships with respect to internal variability is shown by the first, third, and fourth rows of Fig. 11a.
(a) Changes in the component factors of the TH term for the last 150 years of the CO2 perturbation experiments relative to the CTL simulation: μa, βh, ah, and mean upwelling (see text for details). Relative contributions to internal variations in TH across all simulations are summarized as an inset at lower left. Normalized changes in (b) TH, (c) μa, (d) βh, (e) ah, and (f) mean upwelling in the Niño-3 region are shown for CESM1Topo (dark lines and dots) and six CMIP6 simulations (light lines). All changes are reported relative to the average value for the corresponding control simulation.
Citation: Journal of Climate 37, 3; 10.1175/JCLI-D-23-0079.1
Changes in the four terms across the 1.25×CO2–2×CO2 scenarios are small, with the weak increases in TH for the 1.25×CO2 and 1.5×CO2 experiments relative to CTL resulting mainly from stronger connections between thermocline depth and subsurface temperature in the eastern Pacific (Fig. 12a). Notably, the relatively low value of TH in the 1.75×CO2 scenario relative to CTL is dominated by weaker SST–wind stress and wind stress–thermocline feedbacks. These terms are the two leading contributors to internal variability (Fig. 12a) and show substantially smaller contributions to the neighboring 1.5×CO2 and 2×CO2 simulations, supporting our interpretation of this outlier in the analysis results as an expression of internal variability. This result reaffirms the potential confounding influence of internal variability in interpreting changes in ENSO variability across simulations (Stevenson 2012; Maher et al. 2018; Zheng et al. 2018; Ng et al. 2021). Changes in TH for the 3×CO2 and 4×CO2 experiments relative to CTL result from substantially stronger decreases in the sensitivity of thermocline shape to wind stress, augmented by increasingly large reductions in mean upwelling in the eastern Pacific. All four contributing terms are smaller than in CTL for the 6×CO2 and 8×CO2 experiments, with the reduction in mean upwelling playing a key role in both and a weakened relationship between thermocline depth and subsurface temperature in the eastern tropical Pacific being the most influential contributor in 8×CO2. This latter term is larger than in CTL for all but the two largest CO2 forcing scenarios.
Changes in TH and its four contributing terms are summarized for the nine CESM1Topo experiments and the 24 CMIP6 experiments (4 scenarios × 6 models) in Figs. 12b–f. Changes in all terms are normalized relative to the corresponding control simulation value for ease of comparison. Three of the six models show relative changes in TH consistent with those in CESM1Topo (Fig. 12b), while the other three show much sharper increases in TH in the 2×CO2 experiment. Changes in the strength of the SST–wind stress (Fig. 12c) and thermocline–subsurface temperature (Fig. 12e) feedbacks are also broadly consistent across the models, while the CMIP6 models indicate much stronger decreases in eastern Pacific upwelling with increasing CO2 than produced by CESM1Topo. The largest discrepancies between CESM1Topo and the other models are in the sensitivity of thermocline shape to zonal-mean wind stress (Fig. 12d), for which the CESM1Topo result falls outside the envelope of the six CMIP6 models.
Across the six CMIP6 models, values of the BJI are consistently smaller for the 0.5×CO2 and 4×CO2 experiments than for the two intermediate experiments, with the exception of the HadGEM 2×CO2 experiment (Fig. 10d). In many cases, these differences are also reflected in the changes of TH and R with increasing CO2 (Figs. 10a,b). The consistently smaller values of TH, R, and BJI for the 0.5×CO2 simulation are especially interesting, as they lend credence to the idea that ENSO variability peaks at concentrations near or slightly above the present-day value. The reduced TH term in 0.5×CO2 relative to piControl is robustly associated with weaker coupling between the atmosphere and subsurface ocean, which more than compensates for enhanced mean upwelling in the eastern Pacific in all six models. Similarly, the atmosphere–ocean coupling terms generally decrease or plateau between the 2×CO2 and 4×CO2 experiments, while mean upwelling sharply decreases. The CMIP6 results thus provisionally suggest that a peak in ENSO variability at CO2 concentrations near present-day values could result from changes in the coupled atmosphere–ocean contributions to the thermocline feedback, which first strengthen and then weaken with increasing CO2, superimposed on a steadily weaker mean upwelling in the eastern Pacific.
Although Fig. 12 suggests that reductions in the magnitude of the thermocline feedback result in large parts from weaker coupling between the atmospheric wind stress forcing and the subsurface ocean, possible reasons for this weaker coupling are still unclear. Indeed, previous results suggest that stronger stratification, as is observed as the upper ocean warms in response to increasing CO2, typically leads to stronger atmosphere–ocean coupling rather than weaker. Some thermally driven mechanisms may contribute, such as the depletion of the western Pacific warm pool containing water warmer than Niño-3 (Fig. 9) and the requirement for greater changes in subsurface temperatures to support a similar magnitude of ENSO event (Fig. 4). Another contributing factor may emerge from basinwide increases in precipitation (see also Kim et al. 2023), which represent a potentially important distinction between internal variability and external forcing (Fig. 11). Mean precipitation in the Niño-4 region more than doubles between the CTL and 8×CO2 experiments, from 3.8 to 8.1 mm day−1, and the pace of increase in this region (14% °C−1) is double the expected rate of increase in water vapor based on the Clausius–Clapeyron relation (e.g., Held and Soden 2006). These increases in precipitation in the central and western Pacific produce remarkable reductions in surface salinity across the tropical Pacific (Figs. 13c–h). Associated changes in static stability, for which the peak is approximately centered on the thermocline in the CTL simulation (Fig. 13a), are concentrated above the thermocline as CO2 increases (Figs. 13c–h). The changes in salinity, static stability, and thermocline depth suggest the formation of a barrier layer across the tropical Pacific that could restrict interactions between the surface ocean and the thermocline, thereby influencing the intensity of ENSO variability (Guan et al. 2022; Kim et al. 2023). The maximum increases in static stability for the 3×CO2 and 4×CO2 experiments are centered near 150°W, near the east–west inflection point for the thermocline (Fig. 13), and may contribute to the substantial weakening of wind stress–thermocline feedbacks in these simulations. Similarly, strong increases in static stability extend farther into the eastern Pacific in the 6×CO2 and 8×CO2 experiments, potentially explaining the breakdown in the relationship between thermocline depth and subsurface temperature anomalies in those simulations.
Longitude–depth distributions of mean (a) static stability and (b) salinity across the tropical Pacific in the CTL simulation. Differences relative to CTL are shown for both mean salinity (shading) and static stability (pink contours) for years 150–300 of the (c) 1.5×CO2, (d) 2×CO2, (e) 3×CO2, (f) 4×CO2, (g) 6×CO2, and (h) 8×CO2 simulations. Mean thermocline depth for each simulation is shown as a thick black contour in the corresponding panel.
Citation: Journal of Climate 37, 3; 10.1175/JCLI-D-23-0079.1
This freshwater suppression mechanism in CESM1Topo is most pronounced for larger levels of CO2. It is therefore difficult to evaluate whether and to what extent it is present in the evaluated CMIP6 models, although similar signals have been noted in high-emission CMIP6 scenarios of future climate (Kim et al. 2023). All six models indicate that the relative contribution of decreasing surface salinity to the increase in upper-level stability is larger between the 2×CO2 and 4×CO2 scenarios (average contribution: 13%) than between the piControl and 2×CO2 scenarios (6%). However, this contribution is still only about half of that simulated by CESM1Topo, and no robust relationship is identified between changes in TH or R and changes in upper-level stability across these models. Further experiments and evaluations will be necessary to test the effectiveness of this freshwater suppression mechanism and its consistency across models.
5. Conclusions and outlook
Introducing higher-resolution topographic data in the CESM reduces the SST bias in the tropical Pacific and contributes to improvements in several aspects of ENSO simulation (Figs. 1 and 3). We have used this revised model configuration to explore how ENSO changes after climate equilibrates with different levels of atmospheric CO2 ranging from preindustrial to 8 × CO2. The simulations are initialized by applying abrupt incremental increases in CO2. The analysis is then conducted on outputs produced after each simulation has achieved quasi-equilibrium with its corresponding CO2 forcing.
The results show that the occurrence frequency and intensity of ENSO events are roughly steady over the range from preindustrial control to 2 × CO2 (Figs. 2 and 6), with some hints of an increase in ENSO activity and intensity near the middle of this range. Changes in intensity are more pronounced for warm (El Niño) events, for which the average intensity decreases by about 20% between the 1.5×CO2 and 8×CO2 experiments, while no systematic change is simulated in the intensity of cold (La Niña) events. These changes correspond to variations in the ENSO growth rate R and stability index (BJI) calculated using Bjerknes–Jin stability analysis (Jin et al. 2020). Variations in R are determined primarily by variations in the strength of the thermocline feedback term (TH) in both internal variability and the model’s response to external forcing. Variations in the Bjerknes–Jin stability index are in turn determined primarily by R, modulated by variations in the slow adjustment scale ε, which represents the rate of basin-scale adjustment within the tropical Pacific. Changes in TH, R, and BJI across the simulations are within the range of internal variability for CO2 forcings up to twice the preindustrial value. At CO2 concentrations larger than this, the thermocline feedback, the growth rate, and the stability index all decrease substantially, indicating progressively steeper barriers to triggering and growing ENSO events.
Decreases in the magnitude of the TH term between the 2×CO2, 3×CO2, and 4×CO2 simulations are dominated by a sharp weakening in the sensitivity of anomalies in thermocline shape to anomalies in zonal wind stress (Figs. 12a,d). This change in the ocean dynamical response to atmospheric forcing accounts for more than 90% of the decrease in TH across these three simulations and occurs despite increased ocean static stability, which is typically found to intensify atmosphere–ocean interactions in the tropical Pacific (Kim and Jin 2010; Chen et al. 2017; Cai et al. 2018). Further decreases in the magnitude of the TH term in the 6×CO2 and 8×CO2 simulations are characterized by a sharp decline in the strength of the relationship between thermocline depth anomalies and subsurface (75 m) temperature in the eastern tropical Pacific (Fig. 12e). Steady decreases in mean upwelling in the tropical Pacific also contribute to reductions in TH for CO2 concentrations exceeding 2 × CO2 (Fig. 12f). Reductions in the strength of the ocean dynamical response to wind stress forcing in CESM1Topo, including weaker wind stress–thermocline and thermocline–subsurface temperature feedbacks, may be linked to surface freshening, which produces maximum increases in static stability above the thermocline (Fig. 13). The contribution of surface freshening to increasing stability in the Niño-3 region doubles from about 13% for changes in CO2 less than 2 × CO2 to 25% for changes in CO2 greater than that threshold. This surface freshening, which is driven by sharp increases in precipitation in the western Pacific, causes the pycnocline to shoal more swiftly than the thermocline and may explain the weaker connection between the surface and the thermocline (see also Guan et al. 2022; Kim et al. 2023).
To provide context to our model experiments, we have further analyzed simulations from phase 6 of the Coupled Model Intercomparison Project (CMIP6; Eyring et al. 2016), including experiments applying abrupt changes of 0.5 × CO2, 2 × CO2, and 4 × CO2 to preindustrial conditions (Table 2). Although the relatively short duration of these simulations means that they may be farther from quasi-equilibrium than our simulations, changes in TH, R, and BJI are qualitatively consistent with our results in most cases (Fig. 10). In particular, the largest values of the growth rate R are associated with either the preindustrial or 2×CO2 simulation in five out of six simulations, and the most positive values of the stability parameter BJI are associated with either the preindustrial or 2×CO2 simulation in all six models. These results match results from coupled model experiments carried out in long-time equilibrium states (e.g., Callahan et al. 2021), which simulated decreases in ENSO variability with increasing CO2. However, most of these models featured only piControl and 4×CO2 experiments and can only be roughly compared with the smaller steps used in our experiments.
Consensus on the mechanisms behind changes in TH and BJI proves more elusive. The CMIP6 models consistently indicate that smaller growth rates and stability indices in the 0.5×CO2 scenario relative to piControl result from weaker wind stress sensitivity to SST anomalies, weaker thermocline shape sensitivity to wind stress, and a weaker relationship between subsurface temperature and thermocline depth anomalies in the eastern Pacific (Fig. 12). However, the same models produce a range of responses in these components with increasing CO2, and the only consistent contribution to reducing TH in 4×CO2 relative to 2×CO2 is the reduction in mean upwelling in the tropical eastern Pacific. Overall, the number of models is small (six), the durations of the sensitivity simulations conducted by the CMIP6 models are short (∼150 years), and the attendant potential for sensitivity to both climate disequilibrium and internal variability is large (Maher et al. 2018; Zheng et al. 2018). Our evaluation of the influence of disequilibrium in the CESM1Topo results suggests that ε in particular is sensitive to these uncertainties, and changes in this term are influential in the CMIP6-based results.
Although our findings provide insight into how and why ENSO variability changes under elevated levels of CO2, they depend on the choice of model and are sensitive to internal variability. Regarding the choice of model, our results are specific to the model formulation and parameters outlined above, a caveat that can only be addressed by conducting similar tiered quasi-equilibrium experiments using other model formulations. In this context, the stability analysis results may also be useful for assessing how changes in ENSO variability induced by changes in the model formulation, such as atmospheric convection, compare to those induced by changes in CO2. Sensitivity to internal variability could be further evaluated by constructing larger ensembles within the tiered CO2 experiments. Here, owing to limited resources, we have opted for a simpler test: lengthening the 3×CO2 experiment by an additional 300 years. Results across the periods covering years 151–300, 301–450, and 451–600 of this simulation are mutually consistent within the scope of internal variability (Fig. 8c). Our use of smaller-than-usual steps in the tiered-CO2 approach also mitigates sensitivity to internal variability by defining a general pattern and highlighting simulation results that do not fit this pattern, such as the 1.75×CO2 simulation (e.g., Fig. 2b). Moreover, the use of rolling Bjerknes–Jin stability analysis allows us to show where differences between runs exceed the range of internal variability within runs (Fig. 7). A limited assessment of sensitivity to initial conditions has been conducted by comparing two separate versions of the 3×CO2 experiment, one initialized from 2×CO2 and one initialized from 4×CO2 (Fig. 8). Results based on these two simulations are very similar both in average values and in the scale of internal variability, indicating that the 150-yr spinup period is sufficient to mitigate any “shock” in the coupled tropical Pacific caused by abruptly changing CO2.
To close, we return to the motivating question: To what extent do changes in ENSO variability under increasing CO2 forcing reflect changes in the dynamical barrier (how many events are triggered) versus changes in damping (events are triggered but are more difficult to grow and maintain)? Our results based on the quasi-equilibrium CESM1Topo simulations suggest that both factors are important. Changes in damping that reduce the growth rate set in for CO2 concentrations larger than about 2 times preindustrial values and appear to result mainly from weaker coupling between zonal wind stress and the shape of the underlying thermocline. Changes in the dynamical barrier that reduce the number of events triggered set in at slightly higher levels of CO2, about 3–4 times preindustrial values, consistent with the increasing influence of surface freshening on static stability in the tropical Pacific at higher levels of CO2. The quasi-equilibrium response to CO2 forcings closer to the current level, approximately 1.5 × CO2, is indistinguishable from internal variability. These simulations are most relevant to scenarios in which global CO2 emissions are reduced or regulated to stabilize atmospheric CO2 at an approximately constant level for a long time. Our results therefore suggest that ENSO variability is unlikely to deviate much from current levels unless the stabilized value exceeds approximately twice the present-day concentration.
Acknowledgments.
This work was supported by the National Natural Science Foundation of China (Grants 42030602, 42275053, 41975094, U2242206, and 41975094), the International Partnership Program of the Chinese Academy of Sciences (Grant 183311KSYB20200015), and the Ministry of Science and Technology of China (Grant 2017YFA0603902). We thank Jialiang Ma for his suggestions and discussion regarding the manuscript and the National Supercomputing Center in Wuxi for their technical assistance and support, and Dr. Yuko M. Okumura and three anonymous reviewers for insightful suggestions that improved the quality and scope of this work. We acknowledge the World Climate Research Programme’s Working Group on Coupled Modelling, which is responsible for CMIP, and we thank the climate modeling groups (listed in Table 2 of this paper) for producing and making available their model output. We thank the NOAA Physical Sciences Laboratory for providing access to the ERSSTv5 dataset, the Indian National Centre for Ocean Information Services for providing access to the TropFlux dataset, and the Integrated Climate Data Center at the University of Hamburg for providing access to ORAS5 reanalysis products. TropFlux data were produced under a collaboration between the Laboratoire d’Océanographie: Expérimentation et Approches Numériques (LOCEAN) from Institut Pierre Simon Laplace (IPSL, Paris, France) and the National Institute of Oceanography/CSIR (NIO, Goa, India), and supported by the Institut de Recherche pour le Développement (IRD, France). TropFlux relies on data provided by the ECMWF interim reanalysis (ERA-I) and ISCCP projects.
Data availability statement.
CESM1Topo outputs and codes used to reproduce the results are available upon request to the authors. CMIP6 model products were obtained from the Earth System Grid Federation through https://aims2.llnl.gov/metagrid/search/?project=CMIP6. ERSSTv5 data were acquired from the NOAA Physical Sciences Laboratory via https://psl.noaa.gov/data/gridded/data.noaa.ersst.v5.html. TropFlux data were acquired from the Indian National Centre for Ocean Information Services at https://incois.gov.in/tropflux/tf_products.jsp. ORAS5 ocean reanalysis products were obtained from the Integrated Climate Data Center at the University of Hamburg (https://www.cen.uni-hamburg.de/en/icdc/data/ocean/easy-init-ocean/ecmwf-oras5.html). Although the CLIVAR ENSO Metrics Package software was not directly used in this work, descriptions for all metrics were obtained from https://github.com/CLIVAR-PRP/ENSO_metrics/wiki/.
REFERENCES
Alexander, M. A., I. Bladé, M. Newman, J. R. Lanzante, N.-C. Lau, and J. D. Scott, 2002: The atmospheric bridge: The influence of ENSO teleconnections on air-sea interaction over the global oceans. J. Climate, 15, 2205–2231, https://doi.org/10.1175/1520-0442(2002)015<2205:TABTIO>2.0.CO;2.
Ashok, K., S. K. Behera, S. A. Rao, H. Weng, and T. Yamagata, 2007: El Niño Modoki and its possible teleconnection. J. Geophys. Res., 112, C11007, https://doi.org/10.1029/2006JC003798.
Banholzer, S., and S. Donner, 2014: The influence of different El Niño types on global average temperature. Geophys. Res. Lett., 41, 2093–2099, https://doi.org/10.1002/2014GL059520.
Battisti, D. S., and A. C. Hirst, 1989: Interannual variability in a tropical atmosphere-ocean model: Influence of the basic state, ocean geometry and nonlinearity. J. Atmos. Sci., 46, 1687–1712, https://doi.org/10.1175/1520-0469(1989)046<1687:IVIATA>2.0.CO;2.
Bjerknes, J., 1966: A possible response of the atmospheric Hadley circulation to equatorial anomalies of ocean temperature. Tellus, 18A, 820–829, https://doi.org/10.3402/tellusa.v18i4.9712.
Bjerknes, J., 1969: Atmospheric teleconnections from the equatorial Pacific. Mon. Wea. Rev., 97, 163–172, https://doi.org/10.1175/1520-0493(1969)097<0163:ATFTEP>2.3.CO;2.
Cai, W., and Coauthors, 2018: Increased variability of eastern Pacific El Niño under greenhouse warming. Nature, 564, 201–206, https://doi.org/10.1038/s41586-018-0776-9.
Cai, W., B. Ng, T. Geng, L. Wu, A. Santoso, and M. J. McPhaden, 2020a: Butterfly effect and a self-modulating El Niño response to global warming. Nature, 585, 68–73, https://doi.org/10.1038/s41586-020-2641-x.
Cai, W., A. Santoso, G. Wang, L. Wu, M. Collins, M. Lengaigne, S. Power, and A. Timmermann, 2020b: ENSO response to greenhouse forcing. El Niño Southern Oscillation in a Changing Climate, Geophys. Monogr., Vol. 253, Amer. Geophys. Union, 289–307, https://doi.org/10.1002/9781119548164.ch13.
Cai, W., and Coauthors, 2021: Changing El Niño–Southern Oscillation in a warming climate. Nat. Rev. Earth Environ., 2, 628–644, https://doi.org/10.1038/s43017-021-00199-z.
Cai, W., B. Ng, G. Wang, A. Santoso, L. Wu, and K. Yang, 2022: Increased ENSO sea surface temperature variability under four IPCC emission scenarios. Nat. Climate Change, 12, 228–231, https://doi.org/10.1038/s41558-022-01282-z.
Cai, W., and Coauthors, 2023: Anthropogenic impacts on twentieth-century ENSO variability changes. Nat. Rev. Earth Environ., 4, 407–418, https://doi.org/10.1038/s43017-023-00427-8.
Callahan, C. W., C. Chen, M. Rugenstein, J. Bloch-Johnson, S. Yang, and E. J. Moyer, 2021: Robust decrease in El Niño/Southern Oscillation amplitude under long-term warming. Nat. Climate Change, 11, 752–757, https://doi.org/10.1038/s41558-021-01099-2.
Capotondi, A., and Coauthors, 2015: Understanding ENSO diversity. Bull. Amer. Meteor. Soc., 96, 921–938, https://doi.org/10.1175/BAMS-D-13-00117.1.
Carréric, A., B. Dewitte, W. Cai, A. Capotondi, K. Takahashi, S.-W. Yeh, G. Wang, and V. Guémas, 2020: Change in strong eastern Pacific El Niño events dynamics in the warming climate. Climate Dyn., 54, 901–918, https://doi.org/10.1007/s00382-019-05036-0.
Chen, C., M. A. Cane, A. T. Wittenberg, and D. Chen, 2017: ENSO in the CMIP5 Simulations: Life cycles, diversity, and responses to climate change. J. Climate, 30, 775–801, https://doi.org/10.1175/JCLI-D-15-0901.1.
Covey, C., K. M. AchutaRao, U. Cubasch, P. Jones, S. J. Lambert, M. E. Mann, T. J. Phillips, and K. E. Taylor, 2003: An overview of results from the Coupled Model Intercomparison Project. Global Planet. Change, 37, 103–133, https://doi.org/10.1016/S0921-8181(02)00193-5.
Danabasoglu, G., S. C. Bates, B. P. Briegleb, S. R. Jayne, M. Jochum, W. G. Large, S. Peacock, and S. G. Yeager, 2012: The CCSM4 ocean component. J. Climate, 25, 1361–1389, https://doi.org/10.1175/JCLI-D-11-00091.1.
Eyring, V., S. Bony, G. A. Meehl, C. A. Senior, B. Stevens, R. J. Stouffer, and K. E. Taylor, 2016: Overview of the Coupled Model Intercomparison Project Phase 6 (CMIP6) experimental design and organization. Geosci. Model Dev., 9, 1937–1958, https://doi.org/10.5194/gmd-9-1937-2016.
Ferrett, S., and M. Collins, 2019: ENSO feedbacks and their relationships with the mean state in a flux adjusted ensemble. Climate Dyn., 52, 7189–7208, https://doi.org/10.1007/s00382-016-3270-9.
Freund, M. B., J. R. Brown, B. J. Henley, D. J. Karoly, and J. N. Brown, 2020: Warming patterns affect El Niño diversity in CMIP5 and CMIP6 models. J. Climate, 33, 8237–8260, https://doi.org/10.1175/JCLI-D-19-0890.1.
Guan, C., F. Tian, M. J. McPhaden, F. Wang, S. Hu, and R.-H. Zhang, 2022: Zonal structure of tropical Pacific surface salinity anomalies affects ENSO intensity and asymmetry. Geophys. Res. Lett., 49, e2021GL096197, https://doi.org/10.1029/2021GL096197.
Held, I. M., and B. J. Soden, 2006: Robust responses of the hydrological cycle to global warming. J. Climate, 19, 5686–5699, https://doi.org/10.1175/JCLI3990.1.
Huang, B., Z.-Z. Hu, and B. Jha, 2007: Evolution of model systematic errors in the tropical Atlantic Basin from coupled climate hindcasts. Climate Dyn., 28, 661–682, https://doi.org/10.1007/s00382-006-0223-8.
Huang, Y., H.-L. Ren, J.-S. Kug, R. Wang, and J. Li, 2023: Projected change of East-Asian winter precipitation related to strong El Niño under the future emission scenarios. Climatic Change, 176, 81, https://doi.org/10.1007/s10584-023-03551-y.
Hwang, Y.-T., and D. M. W. Frierson, 2013: Link between the double-intertropical convergence zone problem and cloud biases over the Southern Ocean. Proc. Natl. Acad. Sci. USA, 110, 4935–4940, https://doi.org/10.1073/pnas.1213302110.
Jin, F.-F., 1996: Tropical ocean-atmosphere interaction, the Pacific cold tongue, and the El Niño-Southern Oscillation. Science, 274, 76–78, https://doi.org/10.1126/science.274.5284.76.
Jin, F.-F., 1997a: An equatorial ocean recharge paradigm for ENSO. Part I: Conceptual model. J. Atmos. Sci., 54, 811–829, https://doi.org/10.1175/1520-0469(1997)054<0811:AEORPF>2.0.CO;2.
Jin, F.-F., 1997b: An equatorial ocean recharge paradigm for ENSO. Part II: A stripped-down coupled model. J. Atmos. Sci., 54, 830–847, https://doi.org/10.1175/1520-0469(1997)054<0830:AEORPF>2.0.CO;2.
Jin, F.-F., S. T. Kim, and L. Bejarano, 2006: A coupled-stability index for ENSO. Geophys. Res. Lett., 33, L23708, https://doi.org/10.1029/2006GL027221.
Jin, F.-F., H.-C. Chen, S. Zhao, M. Hayashi, C. Karamperidou, M. F. Stuecker, R. Xie, and L. Geng, 2020: Simple ENSO models. El Niño Southern Oscillation in a Changing Climate, Geophys. Monogr., Vol. 253, Amer. Geophys. Union, 119–151, https://doi.org/10.1002/9781119548164.ch6.
Kao, H.-Y., and J.-Y. Yu, 2009: Contrasting eastern-Pacific and central-Pacific types of ENSO. J. Climate, 22, 615–632, https://doi.org/10.1175/2008JCLI2309.1.
Kelley, M., and Coauthors, 2020: GISS-E2.1: Configurations and climatology. J. Adv. Model. Earth Syst., 12, e2019MS002025, https://doi.org/10.1029/2019MS002025.
Kim, H., A. Timmermann, S.-S. Lee, and F. Schloesser, 2023: Rainfall and salinity effects on future Pacific climate change. Earth’s Future, 11, e2022EF003457, https://doi.org/10.1029/2022EF003457.
Kim, S. T., and F.-F. Jin, 2010: An ENSO stability analysis. Part I: Results from a hybrid coupled model. Climate Dyn., 36, 1593–1607, https://doi.org/10.1007/s00382-010-0796-0.
Knutson, T. R., and S. Manabe, 1995: Time-mean response over the tropical Pacific to increased CO2 in a coupled ocean-atmosphere model. J. Climate, 8, 2181–2199, https://doi.org/10.1175/1520-0442(1995)008<2181:TMROTT>2.0.CO;2.
Li, G., and S.-P. Xie, 2014: Tropical biases in CMIP5 multimodel ensemble: The excessive equatorial Pacific cold tongue and double ITCZ problems. J. Climate, 27, 1765–1780, https://doi.org/10.1175/JCLI-D-13-00337.1.
Liu, Z., S. Vavrus, F. He, N. Wen, and Y. Zhong, 2005: Rethinking tropical ocean response to global warming: The enhanced equatorial warming. J. Climate, 18, 4684–4700, https://doi.org/10.1175/JCLI3579.1.
Lu, B., F.-F. Jin, and H.-L. Ren, 2018: A coupled dynamic index for ENSO periodicity. J. Climate, 31, 2361–2376, https://doi.org/10.1175/JCLI-D-17-0466.1.
Lübbecke, J. F., and M. J. McPhaden, 2013: A comparative stability analysis of Atlantic and Pacific Niño modes. J. Climate, 26, 5965–5980, https://doi.org/10.1175/JCLI-D-12-00758.1.
Ma, J., S. Xu, and B. Wang, 2019: Warm bias of sea surface temperature in eastern boundary current regions—A study of effects of horizontal resolution in CESM. Ocean Dyn., 69, 939–954, https://doi.org/10.1007/s10236-019-01280-4.
Maher, N., D. Matei, S. Milinski, and J. Marotzke, 2018: ENSO change in climate projections: Forced response or internal variability? Geophys. Res. Lett., 45, 11 390–11 398, https://doi.org/10.1029/2018GL079764.
Marathe, S., K. Ashok, P. Swapna, and T. P. Sabin, 2015: Revisiting El Niño Modokis. Climate Dyn., 45, 3527–3545, https://doi.org/10.1007/s00382-015-2555-8.
McPhaden, M. J., S. E. Zebiak, and M. H. Glantz, 2006: ENSO as an integrating concept in earth science. Science, 314, 1740–1745, https://doi.org/10.1126/science.1132588.
Michel, C., C. Li, I. R. Simpson, I. Bethke, M. P. King, and S. Sobolowski, 2020: The change in the ENSO teleconnection under a low global warming scenario and the uncertainty due to internal variability. J. Climate, 33, 4871–4889, https://doi.org/10.1175/JCLI-D-19-0730.1.
Neale, R. B., J. Richter, S. Park, P. H. Lauritzen, S. J. Vavrus, P. J. Rasch, and M. Zhang, 2013: The mean climate of the Community Atmosphere Model (CAM4) in forced SST and fully coupled experiments. J. Climate, 26, 5150–5168, https://doi.org/10.1175/JCLI-D-12-00236.1.
Neelin, J. D., D. S. Battisti, A. C. Hirst, F.-F. Jin, Y. Wakata, T. Yamagata, and S. E. Zebiak, 1998: ENSO theory. J. Geophys. Res., 103, 14 261–14 290, https://doi.org/10.1029/97JC03424.
Ng, B., W. Cai, T. Cowan, and D. Bi, 2021: Impacts of low-frequency internal climate variability and greenhouse warming on El Niño–Southern Oscillation. J. Climate, 34, 2205–2218, https://doi.org/10.1175/JCLI-D-20-0232.1.
Oleson, K. W., and Coauthors, 2008: Improvements to the community land model and their impact on the hydrological cycle. J. Geophys. Res., 113, G01021, https://doi.org/10.1029/2007JG000563.
Philander, S. G. H., 1983: El Niño Southern Oscillation phenomena. Nature, 302, 295–301, https://doi.org/10.1038/302295a0.
Planton, Y. Y., and Coauthors, 2021: Evaluating climate models with the CLIVAR 2020 ENSO metrics package. Bull. Amer. Meteor. Soc., 102, E193–E217, https://doi.org/10.1175/BAMS-D-19-0337.1.
Ren, H.-L., and R. Wang, 2020: Distinct growth rates of the two ENSO types. Geophys. Res. Lett., 47, e2020GL088179, https://doi.org/10.1029/2020GL088179.
Richter, I., 2015: Climate model biases in the eastern tropical oceans: Causes, impacts and ways forward. Wiley Interdiscip. Rev.: Climate Change, 6, 345–358, https://doi.org/10.1002/wcc.338.
Richter, I., and S.-P. Xie, 2008: On the origin of equatorial Atlantic biases in coupled general circulation models. Climate Dyn., 31, 587–598, https://doi.org/10.1007/s00382-008-0364-z.
Salinger, M., J. Renwick, and A. Mullan, 2001: Interdecadal Pacific Oscillation and South Pacific climate. Int. J. Climatol., 21, 1705–1721, https://doi.org/10.1002/joc.691.
Santoso, A., W. Cai, M. H. England, and S. J. Phipps, 2011: The role of the Indonesian Throughflow on ENSO dynamics in a coupled climate model. J. Climate, 24, 585–601, https://doi.org/10.1175/2010JCLI3745.1.
Schopf, P. S., and M. J. Suarez, 1988: Vacillations in a coupled ocean-atmosphere model. J. Atmos. Sci., 45, 549–566, https://doi.org/10.1175/1520-0469(1988)045<0549:VIACOM>2.0.CO;2.
Stevenson, S. L., 2012: Significant changes to ENSO strength and impacts in the twenty-first century: Results from CMIP5. Geophys. Res. Lett., 39, L17703, https://doi.org/10.1029/2012GL052759.
Suarez, M. J., and P. S. Schopf, 1988: A delayed action oscillator for ENSO. J. Atmos. Sci., 45, 3283–3287, https://doi.org/10.1175/1520-0469(1988)045<3283:ADAOFE>2.0.CO;2.
Swart, N. C., and Coauthors, 2019: The Canadian Earth System Model version 5 (CANESM5.0.3). Geosci. Model Dev., 12, 4823–4873, https://doi.org/10.5194/gmd-12-4823-2019.
Taschetto, A. S., C. C. Ummenhofer, M. F. Stuecker, D. Dommenget, K. Ashok, R. R. Rodrigues, and S.-W. Yeh, 2020: ENSO atmospheric teleconnections. El Niño Southern Oscillation in a Changing Climate, Geophys. Monogr., Vol. 253, Amer. Geophys. Union, 309–335, https://doi.org/10.1002/9781119548164.ch14.
Tatebe, H., and Coauthors, 2019: Description and basic evaluation of simulated mean state, internal variability, and climate sensitivity in MIROC6. Geosci. Model Dev., 12, 2727–2765, https://doi.org/10.5194/gmd-12-2727-2019.
Timmermann, A., and Coauthors, 2018: El Niño-Southern Oscillation complexity. Nature, 559, 535–545, https://doi.org/10.1038/s41586-018-0252-6.
Tokinaga, H., S.-P. Xie, C. Deser, Y. Kosaka, and Y. M. Okumura, 2012a: Slowdown of the Walker circulation driven by tropical Indo-Pacific warming. Nature, 491, 439–443, https://doi.org/10.1038/nature11576.
Tokinaga, H., S.-P. Xie, A. Timmermann, S. McGregor, T. Ogata, H. Kubota, and Y. M. Okumura, 2012b: Regional patterns of tropical Indo-Pacific climate change: Evidence of the Walker circulation weakening. J. Climate, 25, 1689–1710, https://doi.org/10.1175/JCLI-D-11-00263.1.
Toniazzo, T., and S. Woolnough, 2014: Development of warm SST errors in the southern tropical Atlantic in CMIP5 decadal hindcasts. Climate Dyn., 43, 2889–2913, https://doi.org/10.1007/s00382-013-1691-2.
Torrence, C., and G. P. Compo, 1998: A practical guide to wavelet analysis. Bull. Amer. Meteor. Soc., 79, 61–78, https://doi.org/10.1175/1520-0477(1998)079<0061:APGTWA>2.0.CO;2.
Trenberth, K. E., and J. W. Hurrell, 1994: Decadal atmosphere-ocean variations in the Pacific. Climate Dyn., 9, 303–319, https://doi.org/10.1007/BF00204745.
Vecchi, G. A., B. J. Soden, A. T. Wittenberg, I. M. Held, A. Leetmaa, and M. J. Harrison, 2006: Weakening of tropical Pacific atmospheric circulation due to anthropogenic forcing. Nature, 441, 73–76, https://doi.org/10.1038/nature04744.
Voldoire, A., and Coauthors, 2019: Evaluation of CMIP6 DECK experiments with CNRM-CM6-1. J. Adv. Model. Earth Syst., 11, 2177–2213, https://doi.org/10.1029/2019MS001683.
Wang, B., R. Wu, and R. Lukas, 2000: Annual adjustment of the thermocline in the tropical Pacific Ocean. J. Climate, 13, 596–616, https://doi.org/10.1175/1520-0442(2000)013<0596:AAOTTI>2.0.CO;2.
Wang, B., X. Luo, Y.-M. Yang, W. Sun, M. Cane, W. Cai, S.-W. Yeh, and J. Lu, 2019: Historical change of El Niño properties sheds light on future changes of extreme El Niño. Proc. Natl. Acad. Sci. USA, 116, 22 512–22 517, https://doi.org/10.1073/pnas.1911130116.
Wang, C., and J. Picaut, 2004: Understanding ENSO physics—A review. Earth’s Climate: The Ocean–Atmosphere Interaction, Geophys. Monogr., Vol. 147, Amer. Geophys. Union, 21–48, https://doi.org/10.1029/147GM02.
Wang, C., L. Zhang, S.-K. Lee, L. Wu, and C. R. Mechoso, 2014: A global perspective on CMIP5 climate model biases. Nat. Climate Change, 4, 201–205, https://doi.org/10.1038/nclimate2118.
Wang, W., and M. J. McPhaden, 2000: The surface-layer heat balance in the equatorial Pacific Ocean. Part II: Interannual variability. J. Phys. Oceanogr., 30, 2989–3008, https://doi.org/10.1175/1520-0485(2001)031<2989:tslhbi>2.0.co;2.
Weng, H., S. K. Behera, and T. Yamagata, 2009: Anomalous winter climate conditions in the Pacific Rim during recent El Niño Modoki and El Niño events. Climate Dyn., 32, 663–674, https://doi.org/10.1007/s00382-008-0394-6.
Williams, K. D., and Coauthors, 2018: The Met Office Global Coupled Model 3.0 and 3.1 (GC3.0 and GC3.1) configurations. J. Adv. Model. Earth Syst., 10, 357–380, https://doi.org/10.1002/2017MS001115.
Xie, R., and F.-F. Jin, 2018: Two leading ENSO modes and El Niño types in the Zebiak–Cane model. J. Climate, 31, 1943–1962, https://doi.org/10.1175/JCLI-D-17-0469.1.
Xie, S.-P., C. Deser, G. A. Vecchi, J. Ma, H. Teng, and A. T. Wittenberg, 2010: Global warming pattern formation: Sea surface temperature and rainfall. J. Climate, 23, 966–986, https://doi.org/10.1175/2009JCLI3329.1.
Xu, Z., P. Chang, I. Richter, W. Kim, and G. Tang, 2014: Diagnosing southeast tropical Atlantic SST and ocean circulation biases in the CMIP5 ensemble. Climate Dyn., 43, 3123–3145, https://doi.org/10.1007/s00382-014-2247-9.
Yeh, S.-W., J.-S. Kug, B. Dewitte, M.-H. Kwon, B. P. Kirtman, and F.-F. Jin, 2009: El Niño in a changing climate. Nature, 461, 511–514, https://doi.org/10.1038/nature08316.
Yeh, S.-W., B. Dewitte, B. Y. Yim, and Y. Noh, 2010: Role of the upper ocean structure in the response of ENSO-like SST variability to global warming. Climate Dyn., 35, 355–369, https://doi.org/10.1007/s00382-010-0849-4.
Yu, J.-Y., and S. T. Kim, 2013: Identifying the types of major El Niño events since 1870. Int. J. Climatol., 33, 2105–2112, https://doi.org/10.1002/joc.3575.
Yukimoto, S., and Coauthors, 2019: The Meteorological Research Institute Earth System Model version 2.0, MRI-ESM2.0: Description and basic evaluation of the physical component. J. Meteor. Soc. Japan, 97, 931–965, https://doi.org/10.2151/jmsj.2019-051.
Zebiak, S. E., and M. A. Cane, 1987: A model El Niño–Southern Oscillation. Mon. Wea. Rev., 115, 2262–2278, https://doi.org/10.1175/1520-0493(1987)115<2262:AMENO>2.0.CO;2.
Zhang, M., and H. Song, 2006: Evidence of deceleration of atmospheric vertical overturning circulation over the tropical Pacific. Geophys. Res. Lett., 33, L12701, https://doi.org/10.1029/2006GL025942.
Zhao, B., and A. Fedorov, 2020: The effects of background zonal and meridional winds on ENSO in a coupled GCM. J. Climate, 33, 2075–2091, https://doi.org/10.1175/JCLI-D-18-0822.1.
Zheng, X. T., C. Hui, and S. W. Yeh, 2018: Response of ENSO amplitude to global warming in CESM large ensemble: Uncertainty due to internal variability. Climate Dyn., 50, 4019–4035, https://doi.org/10.1007/s00382-017-3859-7.
Zheng, Y., T. Shinoda, J.-L. Lin, and G. N. Kiladis, 2011: Sea surface temperature biases under the stratus cloud deck in the southeast Pacific Ocean in 19 IPCC AR4 coupled general circulation models. J. Climate, 24, 4139–4164, https://doi.org/10.1175/2011JCLI4172.1.
