Central-Pacific El Niño-Southern Oscillation less predictable under greenhouse warming

El Niño-Southern Oscillation (ENSO) is the dominant mode of interannual climate variability in the tropical Pacific with global impacts on extreme weather, ecosystems, and agriculture. ENSO events are commonly classified into Eastern-Pacific (EP) and Central-Pacific (CP) types due to their distinct spatial-temporal characteristics and climate impacts. Numerous studies have explored how future climate change may affect ENSO characteristics under different warming scenarios, with projections indicating an intensification of ENSO-associated sea surface temperature anomaly (SSTA) magnitude and wide-reaching climate impacts. Extratropical climate change (e.g., changes in the Walker circulation) may also modulate tropical climate. Despite this body of work, far less attention has been given to how greenhouse warming will affect ENSO predictability, leaving a key question unresolved: how will future warming influence the predictability of ENSO? Some recent studies suggest a decrease in ENSO predictability, but short observational records and modulation by internal multi-decadal variability make attribution uncertain. In equilibrated warmer climates, slight reductions in 6-month persistence have been noted, yet predictability depends on both SSTA persistence and subsurface ocean heat content relationships in the equatorial Pacific. Extratropical processes (e.g., the North Pacific Meridional Mode), inter-basin interactions (Atlantic and Indian Oceans), and sub-seasonal variability (Madden–Julian Oscillation) also influence ENSO dynamics and predictability. Furthermore, the boreal spring predictability barrier (SPB) remains a major challenge, and its response to global warming has not been studied. Using CMIP6 simulations, the authors report a robust decrease in predictability for CP ENSO, linked to a strengthened SPB, indicating that CP ENSO will be less predictable in a warming climate, whereas EP ENSO predictability shows no significant change.

Prior research has projected intensified ENSO-related SST variability under greenhouse warming across multiple emission scenarios and highlighted ENSO’s broad climatic impacts, including effects on Southern Ocean warming and Antarctic shelf ocean warming. Studies also indicate that extratropical climate changes may modulate tropical dynamics such as the Walker circulation. While some analyses suggest a recent decline in ENSO predictability, the brevity of observational records and potential modulation by internal multi-decadal variability complicate attribution to anthropogenic warming. Earlier work has examined ENSO persistence in warmer climates, finding slight reductions over six months, but predictability hinges on both SSTA persistence and subsurface heat content relationships. Additional factors influencing ENSO predictability include extratropical precursors like the North Pacific Meridional Mode (which may strengthen in a warming climate), inter-basin interactions with the Atlantic and Indian Oceans, and sub-seasonal variability such as the Madden–Julian Oscillation. The spring predictability barrier remains a persistent issue, and prior to this study, its response to greenhouse warming had not been assessed comprehensively.

Data and experiments: The study analyzes CMIP6 multi-model ensemble outputs regridded to 1° × 1° for monthly SST, subsurface temperature, zonal/meridional/vertical currents, and downward net surface heat flux. Historical forcings are applied up to 2014 and SSP585 thereafter to 2100. Changes are compared between present-day (1900–1999) and future (2000–2099) climates, with anomalies computed relative to 1900–1999 climatology and quadratic detrending applied jointly across both periods. ORAS5 reanalysis (1958–2022) provides observational constraints. ENSO indices: CP and EP ENSO are defined via the first two EOFs of equatorial Pacific (5°S–5°N, 140°E–80°W) SSTA, using normalized PCs to form C-index = (PC1 + PC2)/√2 and E-index = (PC1 – PC2)/√2. ENSO nonlinearity (Alpha) is estimated by fitting PC2 to a quadratic function of PC1; a subset of 28 CMIP6 models with sufficient nonlinearity is also analyzed for sensitivity. Persistence barrier and SPB metrics: Autocorrelation r(m,t) is used to quantify persistence; the spring persistence barrier strength is defined from the maximum gradient of r with lag for each initial month and summed over months, with timing derived from the corresponding lag at which correlation decays to a threshold. The SPB strength and timing are analogously computed using forecast ACC instead of persistence. Linear Inverse Model (LIM): LIM is constructed with a state vector comprising leading PCs of tropical Pacific SSTA and SSHA (SSTTP20: 12 PCs; SSHTP20: 4 PCs; 20°S–20°N, 140°E–80°W). The linear operator L is estimated from lagged and contemporaneous covariance matrices with a 1-month lag. Seasonal forecasts are generated as xf(m+t) = exp(Lt)x(m), and forecast skill is quantified by ACC using k-fold cross-validation (9 folds over each 99-year period). A coupled LIM including northern Pacific PCs is used to assess extratropical contributions, with decoupled runs (setting cross-basin couplings to zero) for comparison. Recharge Oscillator Model (ROM): A seasonally modulated two-box recharge oscillator is fit to each model for CP/EP ENSO, with prognostic equations for TE, TC, and thermocline depth h. Growth/damping rates (a11, a21, a31) and couplings (a12, a22, a32) are estimated by multilinear regression for present-day and future climates. Ensemble forecasts (perfect-model and CMIP6 frameworks) assess SPB/predictability sensitivity to changes in a21 and a22 between climates, with other parameters held fixed (averaged between scenarios). Model integrations use a 0.3-day timestep over 2050 years, analyzing the final 2000 years. Feedback diagnostics: ENSO growth/damping rates are decomposed into contributions from positive feedbacks (zonal advective, thermocline, Ekman) and negative feedbacks (thermodynamical damping, dynamical damping). Thermodynamical damping is obtained by regressing net surface heat flux anomalies onto the C-index; dynamical damping is estimated from mean zonal, meridional, and vertical advection terms over the equatorial Pacific. Statistical significance: A bootstrap method with 10,000 realizations tests whether multi-model mean differences between periods exceed the combined standard deviations at the 95% confidence level.

- A robust decrease in CP ENSO persistence and predictability is projected under greenhouse warming, especially when forecasts must pass through boreal spring. - Persistence: 33/36 CMIP6 models (92%) show decreased year-round SSTA persistence for CP ENSO in 2000–2099 vs. 1900–1999; the multi-model mean decrease is significant at the 95% level. The decrease is dominated by reductions for target months in boreal spring–summer. The spring persistence barrier strength for CP ENSO increases across 32/36 models (89%), with a multi-model mean increase of about 21% (significant at >95%). The timing of the persistence barrier occurs about one month earlier in the future. - Predictability (LIM ACC): Year-round and spring-target ACC for CP ENSO decrease significantly; 26/34 models (76%) show increased SPB strength for CP ENSO, with a multi-model mean SPB strengthening of about 25% (significant at >95%). The SPB timing shifts earlier by about one month. The largest ACC decrease peaks around July. - EP ENSO: No significant or robust change is detected in year-round persistence, spring persistence barrier strength, or LIM prediction ACC/SPB strength for EP ENSO. - Mechanism (ROM and feedback analysis): ROM attribution indicates strengthened CP ENSO SPB arises mainly from a decrease in a21 (stronger damping of the CP ENSO mode). While a22 (coupling of thermocline depth to SST) increases in many models, its effect opposes SPB strengthening and cannot offset the impact of decreased a21. Feedback decomposition shows that although positive feedbacks (zonal advective, thermocline, Ekman) and North Pacific impacts tend to strengthen under warming, they are outweighed by increased negative feedbacks—particularly enhanced thermodynamical damping (and increased dynamical damping)—yielding a net increase in the CP damping rate. Spatial patterns reveal amplified thermodynamical damping across the tropical Pacific in the future period. - Implications: With likely increased occurrence and potential dominance of CP ENSO in a warmer climate, the strengthened SPB and reduced predictability imply more challenging seasonal-to-interannual predictions and broader climate forecast uncertainty.

The study addresses the open question of how greenhouse warming alters ENSO predictability by separating CP and EP ENSO types. Using CMIP6 simulations and complementary LIM and ROM frameworks, the authors show that CP ENSO will experience a significant increase in spring predictability barrier strength and a reduction in persistence and forecast skill, with an earlier onset of the barrier. EP ENSO predictability remains largely unchanged. Mechanistically, increased thermodynamical and dynamical damping under warming augments the net damping of the CP ENSO mode (reduced a21), which dominates over strengthened positive feedbacks and enhanced extratropical precursors. This stronger damping particularly reduces predictability when forecasts traverse boreal spring, explaining the heightened SPB and its earlier timing. These results indicate that greenhouse warming will make CP ENSO prediction more difficult, potentially affecting global climate predictability given CP ENSO’s teleconnections and a possible increase in CP event frequency.

Under greenhouse warming, Central-Pacific ENSO becomes less predictable due to a robust strengthening and earlier onset of the spring predictability barrier, driven primarily by enhanced net damping associated with increased thermodynamical (and dynamical) damping in the tropical Pacific. This reduction in predictability is consistently detected across CMIP6 models and confirmed by LIM-based forecasts and ROM-based attribution experiments. In contrast, Eastern-Pacific ENSO predictability shows no significant change. Given CP ENSO’s potential increasing prominence, these findings imply greater challenges for seasonal-to-interannual climate prediction in a warmer world.