Multi-year La Niña frequency tied to southward tropical Pacific wind shift

ENSO is the dominant source of year-to-year climate variability, oscillating between El Niño (warm anomalies in the central-eastern equatorial Pacific) and La Niña (cold anomalies). Classic recharge–discharge theory explains ENSO’s oscillatory behavior: westerly wind anomalies during El Niño deepen the eastern Pacific thermocline, shoal it in the west, and drive poleward Sverdrup transport that discharges equatorial heat content, favoring a transition to La Niña; conversely, easterlies during La Niña drive equatorward transport that recharges heat content, preconditioning El Niño. ENSO is asymmetric in amplitude (positive SST skewness in the east, negative in the central Pacific) and in duration, with multi-year La Niña events more common than multi-year El Niño. Since 2000, three multi-year La Niña sequences have occurred, leading to prolonged global impacts (e.g., eastern Australia floods, U.S. drought). Prior observational evidence suggests La Niña-associated recharge is weaker than El Niño-associated discharge, enabling longer La Niña duration. Other proposed contributors include subsurface thermal anomalies, off-equatorial processes, inter-basin interactions, and the meridional width of tropical Pacific SST patterns. However, a unifying explanation for multi-year La Niña occurrences, including observed decadal variations and inter-model diversity, has been lacking. This study proposes and tests the hypothesis that multi-year La Niña frequency is strongly linked to the tropical Pacific upper-ocean heat content propensity, measured by warm water volume (WWV) skewness, and that this propensity is governed by a southward shift of tropical Pacific zonal winds during austral summer.

The study situates its contribution within several strands of ENSO research. Observations indicate asymmetries in ENSO amplitude and duration, with La Niña more likely to persist across years, consistent with weaker recharge during La Niña than discharge during El Niño. Multiple mechanisms have been proposed to explain multi-year La Niña: subsurface thermal state influences, off-equatorial processes and meridional modes, inter-basin (e.g., Indian/Atlantic) interactions, and the meridional width or regime (Central vs Eastern Pacific) of SST anomalies. Some studies emphasize the role of strong preceding El Niño in preconditioning subsequent multi-year La Niña via enhanced discharge, though others argue this role is sometimes overemphasized. The seasonal southward shift of wind anomalies (linked to ENSO phase locking and termination) has been highlighted as influential for thermocline structure in the eastern equatorial Pacific and ENSO asymmetry. Despite these insights, a consistent framework linking multi-year La Niña frequency to the tropical Pacific heat content skewness and the southward wind shift across models and observations has remained elusive, motivating the present analysis.

Definitions and indices: ENSO events are identified using the Niño3.4 SST index (170°W–120°W, 5°S–5°N) averaged over DJF. An ENSO event occurs when DJF Niño3.4 exceeds ±0.75 standard deviations; sensitivity tests use alternative thresholds and Niño3/Niño4-based EP/CP regimes. Multi-year La Niña is defined as two or three La Niña events in consecutive years. WWV (warm water volume) is the integrated ocean volume above the 20°C isotherm over 120°E–80°W, 5°S–5°N and serves as a proxy for tropical Pacific upper-ocean heat content; T300 (0–300 m averaged temperature) is also used in sensitivity tests. WWV skewness is computed in sliding windows (primarily 11-year; also 21-, 31-, and 51-year tested) to quantify discharge–recharge propensity. Observations: SST from HadISST; surface zonal wind stress from NCEP/NCAR Reanalysis 1; ocean temperature from IAP/CAS; period from 1948 onward. Anomalies are referenced to full-period climatologies and quadratically detrended. Models: Historical simulations from 33 CMIP6 models (1900–1999) with available SST, surface zonal wind stress, ocean temperatures, and ocean meridional velocity. Model fields are regridded to 1°×1°, with ocean vertical interpolation to 0–300 m at 10 m intervals. Analysis steps: (1) Count multi-year La Niña occurrences over 1900–1999 for each model given rareness. (2) Compute model WWV skewness over the 20th century. (3) Diagnose the southward tropical Pacific wind shift: regress DJF zonal wind anomalies onto Niño3.4 for each model, then apply EOF analysis across models to the ENSO-related DJF zonal wind fields over 10°S–5°N, 130°E–80°W; the leading EOF represents inter-model differences in southward wind shift (westerlies displaced south of the equator at El Niño peak). (4) Compute net meridional heat transport and recharge rate: T_m(depth, lon) = ρ Cp T(depth, lon) vo(depth, lon) at 10°N + ρ Cp T(depth, lon) vo(depth, lon) at 10°S; integrate over 0–300 m and across the basin (160°E–100°W for recharge-rate summary) to obtain recharge/discharge rate; positive indicates heat into, negative out of the tropical Pacific. (5) Relate southward wind shift to recharge-rate skewness and thermocline gradient skewness (west-minus-east thermocline depth averaged over 5°S–5°N in specified longitude bands). (6) ENSO nonlinearity: apply EOF to tropical Pacific SST (15°S–15°N, 140°E–80°W); fit PC2(t) = α[PC1(t)]² + β PC1(t) + γ; more negative α indicates stronger nonlinear Bjerknes feedback and more realistic EP/CP regime separation. (7) Observational counterpart: sliding-window EOF analyses of DJF zonal wind responses (11-, 31-, 51-yr windows) to derive a principal component representing southward wind shift; correlate with WWV skewness. Statistical significance: inter-model correlations assessed; bootstrap resampling (10,000 realizations) used to estimate uncertainty for correlation between multi-year La Niña frequency and WWV skewness at varied thresholds.

- Observations: Using 11-yr sliding windows since 1948, decadal La Niña occurrences covary strongly with WWV skewness (correlation ≈ -0.71), indicating more La Niña events, especially consecutive sequences, when the tropical Pacific is predisposed to discharge (negative WWV skewness). Across the record, 21 La Niña vs 14 El Niño events occurred (DJF Niño3.4 threshold ±0.75 s.d.), with about 70% of La Niña (15 of 21) belonging to seven multi-year sequences; no multi-year El Niño was observed under this definition. Results are robust to alternative thresholds and to using T300 instead of WWV. - Models (33 CMIP6): A strong inter-model relationship exists between multi-year La Niña frequency and WWV skewness over 1900–1999 (correlation ≈ -0.75, slope ≈ -4.7 multi-year events per unit WWV skewness, p<0.001): models with more negative WWV skewness simulate more multi-year La Niña events. - Southward wind shift mechanism: The leading inter-model EOF of ENSO-related DJF zonal wind anomalies captures a southward displacement of westerly anomalies during El Niño. The principal component correlates strongly with WWV skewness: greater westerlies south of the equator are associated with more negative WWV skewness. Inter-model regression indicates that stronger south-of-equator westerlies correspond to greater ocean heat discharge (more negative recharge rate) out of the tropical Pacific; correlation between southward wind shift and integrated recharge-rate skewness is about -0.7. - Thermocline and nonlinearity: Stronger southward wind shift is associated with a thermocline skewed shallow in the western and deep in the eastern tropical Pacific (steeper zonal thermocline tilt), favoring discharge and contributing to ENSO nonlinearity. Models with greater southward westerlies tend to produce more realistic ENSO nonlinearity (more negative α) with clearer CP/EP regime separation. - Observational wind–WWV link: Sliding-window analyses show consistent negative correlations between the southward wind shift index and WWV skewness in observations: ≈ -0.72 (11-yr), ≈ -0.93 (31-yr), and ≈ -0.85 (51-yr), all p<0.001, confirming the wind-shift control on discharge–recharge propensity.

The findings support a conceptual framework in which the frequency of multi-year La Niña is governed by the tropical Pacific’s discharge–recharge propensity, quantified by WWV skewness, and dynamically controlled by a southward shift of zonal winds during austral summer. When models or observed conditions feature stronger westerlies south of the equator at El Niño maturity, the thermocline tilts more steeply (deepened in the east, shoaled in the west), enhancing meridional heat export (discharge) from the tropical Pacific. This sets the stage for La Niña persistence or recurrence, thereby increasing multi-year La Niña frequency. The relationship is evident in observations despite limited sample size and is robust across CMIP6 models, linking atmospheric seasonal asymmetry directly to upper-ocean heat-content skewness and ENSO nonlinearity. The mechanism also clarifies why a strong El Niño can sometimes be followed by prolonged La Niña via strong discharge, while acknowledging that multi-year La Niña can occur without a preceding strong El Niño. By tracing inter-model differences in multi-year La Niña frequency to biases in simulating the southward wind shift, the study highlights a critical target for improving ENSO representation, asymmetry, and predictability.

This work establishes a mechanistic and statistically robust link between multi-year La Niña frequency and the tropical Pacific’s discharge propensity (negative WWV skewness), governed by a southward shift of zonal winds during austral summer. Stronger westerlies south of the equator enhance heat discharge via steeper thermocline tilt, fostering prolonged La Niña and contributing to ENSO asymmetries and nonlinearity. The relationship holds in observations and across 33 CMIP6 models. These insights call for constraining and improving the simulation of the southward wind shift and associated ocean–atmosphere feedbacks to enhance near-term climate prediction and long-term projection. Future research should further diagnose model biases in seasonal wind migration, explore triggers (including inter-basin influences) that initiate multi-year sequences, and assess how greenhouse warming may modulate the wind shift, thermocline tilt, and WWV skewness.

- Observational record length is limited, reducing the number of multi-year La Niña sequences and necessitating sliding-window analyses. - Definitions of ENSO events and thresholds can influence counts, though robustness tests were performed. - Large inter-model spread exists in southward wind shift strength, thermocline tilt, recharge rate skewness, ENSO nonlinearity, and resulting multi-year La Niña frequency. - Attribution focuses on internal tropical Pacific processes; external triggers (e.g., inter-basin interactions, volcanic/aerosol forcings) may also modulate events and are not exhaustively explored. - Reanalysis and model uncertainties (e.g., wind stress, subsurface structure, ocean velocity fields) and methodological choices (regridding, detrending, isotherm depth proxy) may affect quantitative estimates.