Research Watch · November 11, 2026

Research Watch: The Ocean's Memory and El Niño's Possible Brake

Two new studies ask how El Niño starts and whether it can keep growing. One finds a possible ocean memory of the MJO. The other, a preprint, proposes a testable limit on amplitude and sets out how this year's event could challenge it.

By Christopher W. Corwin · IAMElNino.com · 8 min read
← All Posts
~8 months
How far ahead of ENSO onset the authors say MJO-driven current variability intensifies. Song et al., npj Climate and Atmospheric Science, 1 September 2026.
+1.8 to +4.3°C
Seasonal Niño 3.4 anomaly where the fitted east-west contrast reaches zero, April to November. A statistical extrapolation, not a measured limit. Colfescu et al. preprint, 27 October 2026.
29% vs 62%
Chance of further Niño 3.4 growth in 57 coupled models, September to December: near-zero east-west contrast versus 1 to 2°C remaining. Model results from the preprint, not observations.

Two new studies ask how El Niño starts and whether it can keep growing. One, in a peer-reviewed journal, finds that repeated Madden-Julian Oscillation (MJO) forcing may leave an ocean memory that helps precondition El Niño. The other is a preprint that has not been peer reviewed. It proposes that El Niño's own growth erodes the east-west temperature contrast that sustains it, and it sets out tests that this year's event could challenge. They examine different stages of an event with different data, and neither tests the other. The preprint addresses 2026 directly. The peer-reviewed study's data end in 2025.

StudyStatusEvidence baseWhat we read
Song et al., npj Climate and Atmospheric Science, 1 Sep 2026Peer reviewed (accepted 22 Aug 2026); accepted-manuscript version, to be replaced by the final oneObservations and model experiments, 1993 to 2025Full main text; not the supplement
Colfescu et al., EGUsphere, 27 Oct 2026Preprint, under review at Weather and Climate Dynamics; not peer reviewedThree SST reconstructions, 1950 to 2025; 57 coupled modelsFull text; not the supplement

The ocean may remember the MJO

The MJO is an eastward-moving pulse of enhanced and suppressed tropical rainfall that circles the globe in roughly 30 to 60 days. Its wind anomalies over the western Pacific can launch oceanic Kelvin waves, eastward-propagating disturbances that deepen the thermocline and help redistribute heat toward the central and eastern Pacific. We covered the mechanics in What Is the MJO, Kelvin Waves, Explained and The MJO and El Niño.

The usual picture treats the western Pacific ocean as a passive responder. Song and colleagues argue it plays an active role in preconditioning. Using observations and model experiments for 1993 to 2025, they report that variability of the zonal (east-west) ocean current in the tropical western Pacific, measured before winter, strengthens ahead of El Niño and weakens ahead of La Niña. They attribute about 80% of that variability's amplitude to wind forcing linked to the MJO over the Indian Ocean. Their proposed mechanism is cumulative: repeated MJO-driven pulses leave a persistent low-frequency ocean response, which strengthens Kelvin-wave-driven eastward heat transport and builds heat in the central Pacific. They say the signal intensifies about eight months before ENSO onset, adds information beyond monthly-mean circulation, and lifts a Niño 3.4 reconstruction to a correlation of 0.73.

Three limits come from the paper's main text. The sample is nine El Niño events since 1993, which the authors say constrains formal significance. The 0.73 is a reconstruction from a linear model fitted to the record. The main text does not describe it as cross-validated, and it is not real-time forecast skill. And the current signal is not independent of the winds: with wind variability removed, it adds no independent linear correlation.

A possible brake: the shrinking east-west contrast

Normally the western Pacific is much warmer than the east. That contrast drives the trade winds, and the trade winds keep cold water upwelling in the east. In the Bjerknes feedback, El Niño warms the east, the contrast shrinks, the trades weaken, upwelling eases and the east warms further. The loop feeds on the contrast, which raises an obvious question: what happens as the contrast runs down?

Colfescu and colleagues test that question with three SST reconstructions for 1950 to 2025. In each calendar month, the contrast (warm pool minus the Niño 3 box) fell by about 1°C for each 1°C of Niño 3.4 warming. Extending that line to zero gives a seasonal threshold of +1.8°C in April to +4.3°C in November. No month since 1950 has exceeded its threshold. In historical months when the contrast was near zero, the heat-budget terms that normally drive growth reversed, but only in the decay season. Across 57 coupled climate models, a near-zero contrast in September to December went with a 29% chance of further Niño 3.4 growth, against 62% when 1 to 2°C of contrast remained.

The authors are explicit about what this is. The threshold is a candidate constraint on amplitude, not a prediction of any event's peak, and historical non-exceedance does not establish a physical ceiling. The near-unit slope partly reflects that Niño 3 and Niño 3.4 share much of the same SST signal while the warm pool varies little. All eleven historical near-zero months fell between January and May, from only three events, so the observations cannot separate a contrast-imposed limit from the seasonal ending of strong events. The threshold also extrapolates, since no observed autumn anomaly has approached the zero crossing. The models do not settle it either: their November thresholds range from +1.76 to +10.13°C (median +5.05°C), so it is not a common temperature limit. An exploratory attempt to predict event peaks from the mean state performed weakly, with leave-one-out correlations of 0.22 and 0.06.

It is also not an ONI or RONI threshold. The paper uses monthly ERSSTv5 anomalies against a 1991 to 2020 base, plus a relative version that subtracts the 20°S to 20°N ocean-mean anomaly with no rescaling. NOAA's seasonal ONI averages three months, and RONI adds a variance adjustment (see RONI vs ONI). The paper's weekly Niño values come from CPC's conventional weekly file, while this site tracks the relative one. The numbers cannot be read against NOAA's event categories, nor are they the Niño 1+2 versus Niño 3.4 contrast in our July analysis.

What the preprint says about 2026

Unusually, the preprint turns to the current event. By its account, the observed contrast fell from 5.0°C in January to 0.58°C in September, the smallest for any September in ERSSTv5 (the previous minimum was 0.72°C in 1997). It describes the event as strongly eastern Pacific centred, and reports record equatorial ocean recharge in August (+2.4 standard deviations, against +2.2 before 1997/98; see our warm-water volume analysis). In July to September, the two Bjerknes-related advective terms in its heat budget summed to a record +1.5 K per month.

The authors then fix two tests before the outcome. The growth-arrest test: the hypothesis would be contradicted if the contrast fell below 0.3°C in any month from October 2026 to March 2027 and Niño 3.4 rose the following month. The amplitude test: whether the relative Niño 3.4 anomaly exceeds +3.64°C in October, +3.68°C in November, +3.50°C in December or +2.96°C in January. They state that a plateau sustained by stored ocean heat would not by itself contradict the hypothesis, and that the record recharge could keep warming the eastern Pacific after the surface contrast is gone. Full evaluation needs observations through April 2027.

Our interpretation is that the tests are built so that only some outcomes would count against the idea. The authors' own wording is that 2026 offers a prospective test, not a demonstration of a ceiling. Nothing in the preprint says this event must stop strengthening, and we do not say so either. The authors say their scripts and a pre-registration of the tests are deposited on Zenodo, with the DOI to be added at submission. We could not check that deposit.

How the two fit, and how they do not

Our interpretation is that the studies address different stages. Song et al. concern onset: whether the ocean has been primed by repeated MJO forcing, and how heat gets advected into the central Pacific. Colfescu et al. concern amplitude: whether growth can continue as the east-west contrast erodes. The eastward heat advection in the first study is the kind of process the second study's heat budget tracks, which is why the two sit naturally together. They need not conflict: a primed ocean can start a strong event, and a shrinking contrast could still limit how far it goes.

That link is our framing, not either set of authors'. The studies use different periods (1993 to 2025 and 1950 to 2025), different data and different goals, and neither was written to predict the other's result.

What would strengthen or challenge the findings

What to check next

This piece makes no forecast. It reports two studies, and the second one defines its own tests.

Checkpoints: NOAA CPC's ENSO diagnostic discussion on 12 November 2026; the close of the EGUsphere discussion period on 12 December 2026; and monthly values through March 2027 against the preprint's stated criteria, with evaluation complete after April 2027.

Data vintage and limitations. Song et al. were received 11 May, accepted 22 August and published online 1 September 2026; the version read is the accepted manuscript, which will be replaced by the final version. Colfescu et al. were received 22 October and posted 27 October 2026 on EGUsphere (submitted to Weather and Climate Dynamics); no referee comments were visible when we checked on 11 November. We read the main text of both papers and neither supplement. All figures quoted come from those texts. The preprint's 2026 values are preliminary and use its own conventions (monthly ERSSTv5 anomalies and CPC's conventional weekly Niño file), which differ from NOAA's seasonal ONI and RONI.

Sources

← All Posts Teleconnections Dashboard →