Just after one in the morning UTC on 1 September 2026, the meteorologist Ryan Maue posted a satellite loop of the Pacific to X. "You're looking live at El Niño going 'full-tilt' to start meteorological fall," he wrote. "Once in a lifetime event." Three hurricanes were turning in the eastern Pacific at the same time, with Hawaii, as he put it, "splitting the uprights." Moisture from Hurricane Marie was about to "inundate U.S. desert Southwest with major monsoon."

Five days earlier, Science had published a paper by Julia Cole of the University of Michigan and twelve colleagues. They had dived the reefs at Wolf and Darwin in the northern Galápagos and cored thirteen corals: living colonies eight to ten feet tall, and dead boulders washed up along the shorelines, some of them 4,500 years old. Drilled a millimetre at a time and read for the ratio of strontium to calcium and for oxygen isotopes, the skeletons give a month-by-month temperature history of the eastern equatorial Pacific across the past thousand years. What that history shows is a rise in the year-to-year swing of the ocean, 36.5 percent above its pre-industrial level, which "exceeds simulated natural variability" and "results from stronger El Niño events." Michigan's announcement rendered the finding as El Niño events "nearly 40% stronger" than pre-industrial ones. "We don't see a time in the past where El Niños have been as strong as today," Cole said, "and we show that the strength of El Niño changes in parallel with the warming of global temperature."

Two claims about the same ocean, made in the same week, both broadly right.

And on the morning Maue posted, the file NOAA's Climate Prediction Center now uses to decide whether there is an El Niño at all, and how big it is, put the event two-hundredths of a degree short of moderate.

The Climate Prediction Center publishes its ENSO record as plain text. Anyone can read it. The traditional file, oni.ascii.txt, runs from the winter of 1950 to the present in overlapping three-month seasons, and carries two numbers for each: the absolute sea surface temperature in the Niño-3.4 box of the central Pacific, and the departure of that temperature from a thirty-year average.

The most recent season in the file is May–June–July 2026. The absolute temperature is 29.02 °C. The departure is +1.39.

Twenty-nine degrees is bathwater at the low end. It is also, in that box, extraordinary. Across the 918 seasons in the record it is the fourth-warmest reading of any kind, behind only three seasons from the peak of the 2015–16 event. Compared like with like, May–June–July against May–June–July, which removes the ordinary swing of the annual cycle, it is the warmest such season ever recorded, ahead of MJJ 2015 by 0.20 °C and ahead of MJJ 1997 by 0.35 °C. There is no reading anywhere in the whole of the 1997–98 event, the one that was called the El Niño of the century while it was happening, as warm as the water sitting there this summer.

The second number, +1.39, places the event in the middle of the "moderate" band.

A third number now exists, and since February it is the official one.

What changed

On 13 January 2026, at 2:40 in the afternoon Eastern time, the National Weather Service issued Public Information Statement 26-05, over the name of Marina Timofeyeva, Supervisory Physical Scientist. Its subject line reads: "Public Information Statement Implementing a Relative Oceanic Niño Index effective February 1, 2026."

The Climate Prediction Center's public explanation is short. The traditional Oceanic Niño Index, it says, "relies on a departure from 30-year average that struggles to keep pace with anomalous changes in tropical sea surface temperature (SST), which is particularly problematic in real-time when using a time lagged climatology. RONI solves this problem by comparing the ENSO region to the global tropics, thereby reducing the dependency on the climate base period."

The mechanism is a subtraction. Take the Niño-3.4 anomaly. Subtract from it the average sea surface temperature anomaly of the whole tropical band, 20° North to 20° South. What is left is the Relative Oceanic Niño Index, RONI. The 0.5 °C threshold survives, and so does the five-consecutive-seasons rule for declaring an event.

Air over warm water rises. But it does not rise because the water is warm in some absolute sense; it rises because that water is warmer than the air's other options nearby. Tropical thunderstorms cluster where the ocean is hot relative to the rest of the tropics, because the atmosphere spreads heat sideways very efficiently and the whole tropical belt tends to settle towards a common temperature aloft. Warm the entire tropical ocean by a degree and you have not necessarily moved a single storm. Warm one patch by a degree while the rest stays put and the storms migrate to it, the trade winds slacken, the jet stream over North America bends, and the rains fail somewhere in Africa.

El Niño's effects on weather thousands of kilometres away are transmitted almost entirely through that mechanism. As the public notice puts it: "It is the change in tropical rainfall and heating that ultimately drives the subseasonal to seasonal variations we see in the midlatitudes and over the United States."

Measured against that job, an index built on absolute departure from a fixed thirty-year baseline has been failing for years. The whole tropical ocean has been warming, and an anomaly reckoned against thirty years of a cooler past reads high no matter what the atmosphere is doing.

RONI is the correct instrument for the job it names.

It was not adopted for convenience. It was argued into existence over years, in the open literature, most influentially by Geert Jan van Oldenborgh of the Royal Netherlands Meteorological Institute and colleagues in Environmental Research Letters in 2021, in a paper called "Defining El Niño indices in a warming climate." Tropical-cyclone researchers reached the same conclusion independently and earlier, according to Michael Tippett of Columbia University, whose own recent work found that "modern ENSO indices (especially RONI) had modest but statistically detectable advantages over the traditional index," with the benefit clearest in forecasting named-storm counts.

The failures the old index produced were real and they cost lives. The 2021 paper points to the 2016–17 La Niña, weak by the traditional measure, whose effects were far more intense than expected, contributing to devastating autumn-rain shortfalls in the Horn of Africa and excessive rain in Indonesia. Run in the other direction, the 2023–24 El Niño was officially strong and its midlatitude impacts underdelivered; Clara Deser and colleagues, in modelling work published this January, find that the atmospheric effects of global ocean warming may have largely cancelled the impacts that event was expected to produce.

Deser's own assessment of the change is unambiguous: "The Relative Oceanic Niño Index will provide a more accurate assessment of El Niño and La Niña events as the global oceans continue to warm. This background warming, undoubtedly a reflection of human-caused climate change, obscures the true magnitude of these events, hindering operational prediction and associated climate impact risk."

L'Heureux, who leads NOAA's ENSO team, puts the diagnosis simply: "Our traditional indicator, which is based on sea surface temperatures, is increasingly misaligned with changes in the atmosphere that are intrinsic with the coupled ENSO phenomenon. The RONI seems to better target this coupling." She notes a further virtue: "One of the nice attributes of RONI is that it is relatively insensitive to the choice of climatology."

Nobody has hidden anything. Australia and New Zealand have already made the switch, India is considering it, and the legacy files remain live, updated, and public, "for users that require that continuity," exactly as the notice promises.

Re-deriving the divergence

The two files sit next to each other on the same server, updated in parallel, 918 seasons each. Subtracting one from the other, decade by decade, produces a clean and monotonic drift.

In the 1950s, RONI ran on average 0.17 °C above ONI. In the 1970s, +0.17. In the 1990s, +0.11. In the 2000s the two indices agree almost exactly, −0.01. In the 2010s, −0.22. In the 2020s so far, −0.44.

That is a swing of six-tenths of a degree across the record, and it is not noise; it is the tropical warming trend being removed, which is precisely what the index was designed to do. L'Heureux and colleagues say so themselves in the 2024 Journal of Climate paper that established the method: "the RONI damps the positive SST trend found in the ONI, making seasonal ENSO variability a more distinctive feature of the time series."

The correction runs in both directions. The easy criticism, that NOAA has simply made El Niño look smaller, is not what the data shows. Since 2000, warm seasons come down by an average of 0.23 °C under RONI, and cold seasons come down by 0.16 °C. Down, in a cold season, means more La Niña, not less.

The clearest case is the winter of 2024–25. Under the traditional index, December–February 2025 reads −0.46: below zero, but short of the −0.5 threshold, and therefore not a La Niña at all. Under RONI the same season reads −1.10, a moderate La Niña, comfortably past the threshold and most of the way to the next band. The gap is 0.64 °C. Bob Henson, writing in Yale Climate Connections in February, noted the consequence directly: the marginal cool conditions since 2024 that were not making the grade under the old index become, under RONI, two moderate La Niña events: 2024–25 and 2025–26, the latter still under way when he wrote.

For the current event the two indices read +1.39 and +0.98, a gap of 0.41, and the difference between a moderate El Niño and a weak one. For the 2023–24 event the gap widens: the traditional index peaks at +1.99 in November–January 2023, deep in "strong"; RONI peaks at +1.42 a month earlier, in the middle of "moderate." At the single season of maximum disagreement, December–February 2024, the two indices are 0.71 °C apart.

Across all seasons since 2000, 103 of 318, very nearly a third, fall into a different CPC intensity band depending on which file you open.

The historical league table also inverts. Under the traditional index, the three great El Niños rank 2015–16 (+2.59), then 1997–98 (+2.37), then 1982–83 (+2.14). Under RONI the order reverses exactly: 1982–83 (+2.40), then 1997–98 (+2.28), then 2015–16 (+2.25). The 1982–83 event gains a quarter of a degree; 2015–16 loses a third of one. A fourth event, 1991–92, gains more than half a degree and climbs from eighth in that ranking to fourth, ahead of 2023–24. Nothing happened to the ocean of 1983 in the intervening years. The ruler moved.

The scaling factor

Subtracting a smoothly rising trend from a wiggly series removes some of the wiggle. RONI, computed plainly, has less variance than ONI. The peaks are lower, the troughs shallower, and a fixed 0.5 °C threshold would therefore fire less often. The 2024 paper is candid about why that is unacceptable: the loss of variance "is problematic for operational agencies that rely on threshold anomalies to determine the status of ENSO."

So a scaling factor is applied. The public notice states it in one line: "After the subtraction, the relative index is rescaled to match the amplitude of the traditional index."

The index is therefore deliberately tuned so that the familiar threshold keeps declaring events at roughly the familiar rate. The notice adds, reasonably: "The RONI is visually very similar to the traditional ONI, so users can use it in the same way as they would the traditional ONI."

So the new number looks like the old number, is bounded like the old number, crosses the same thresholds at roughly the same cadence, and is explicitly recommended for use in the same way. What it measures has changed.

One word, two jobs

RONI is right about what it measures. The difficulty is that a single phrase is being asked to do two incompatible jobs.

"El Niño" is a technical term for a coupled ocean-atmosphere state. It is also the word the world reaches for when it wants to talk about damage: failed harvests, bleached reefs, collapsed fisheries, floods in Piura. The first job wants the warming trend removed, because the trend is not the signal. But a farmer outside Piura is not experiencing a signal. He is experiencing water, at whatever temperature it happens to have reached, and the trend is part of what arrives at the coast.

Jeremy Klavans of the University of Colorado Boulder, who supports the change, names the difficulty without flinching. Regional modes like ENSO, he observes, "are primarily useful if they capture a regional, not global, signal and if they explain meaningful impacts. As such, most of these modes attempt to exclude the global warming signal. As the world warms, this task has become harder." He calls RONI "a well-reasoned attempt to account for the warming signal in a challenging operational environment," and then asks the question underneath all of it: "how societally relevant will they be going forward? Will ENSO or other indices be as useful for predicting regional climate as they used to be?"

Where the money actually looks

If the index has consequences, they should be visible in the instruments that spend money on it.

In 2010, northern Peru became the site of the world's first regulated forecast index insurance product. The Extreme El Niño Insurance Product was sold by La Positiva Seguros, reinsured by PartnerRe, and bought by the microfinance institution Caja Nuestra Gente in 2011 and 2012 to protect its loan book. It was designed to pay out before the flooding in Piura arrived, in early January, so that the money reached people while there was still time to move.

Its trigger is worth reading carefully. The contract uses the monthly sea surface temperature for the Niño 1+2 region, the coastal box off Ecuador and northern Peru, 0–10° South, 80–90° West, as measured and reported by the NOAA Climate Prediction Center. The basis for payment is the simple average of two months, November and December. Three versions were offered, with payments beginning at 23.4, 24.0 and 24.5 degrees Celsius, and reaching the maximum insured sum at 27 degrees. The payout function is linear.

Those are absolute temperatures. Not anomalies, not departures from a baseline, not an index at all in the usual sense. Degrees of water. The contract points at a specific NOAA text file, sstoi.indices, which is still published and still updated; its July 2026 line reads Niño 1+2 at 25.40 °C and Niño-3.4 at 29.33 °C.

The people with capital at risk, designing an instrument in 2009 that had to pay reliably or destroy a microfinance portfolio, went straight to the thermometer.

Peru's government faced the same problem NOAA did and reached the opposite conclusion. The official diagnosis of El Niño and La Niña on the Peruvian coast belongs not to NOAA but to ENFEN, the multi-sector commission that studies the phenomenon, and its instrument is the Índice Costero El Niño, or ICEN: a three-month running mean of SST anomalies in the Niño 1+2 box. In December 2024, "due to current climate changes," ENFEN revised the ICEN — and kept it absolute, updating the fixed baseline to 1991–2020 rather than subtracting a moving one.

The two agencies' readings are not directly comparable, and it would be sloppy to set one against the other as though they measured the same quantity. Different box, different baseline, different purpose. That caveat is real.

It is also the finding. For the season centred on May 2026, ENFEN's coastal index reads +1.98. CPC's RONI for May–June–July reads +0.98. A reader who wants to know how big this El Niño is can ask two governments and receive answers a full degree apart, both official and both correct.

For the humanitarian financing system the answer runs the same way. The inter-agency standard operating procedures for ENSO response are triggered not by ONI or RONI but by probability: a greater-than-55-percent chance of an event in the official forecast produced by Columbia's International Research Institute with NOAA, escalating to 75 percent, with the WMO's ENSO Update as a parallel source. Country-level thresholds are then negotiated in country, and the IFRC's Early Action Protocols hang off local hazard forecasts — rainfall in Ecuador between January and April, drought in Central America between June and August. The systems that move money before a disaster arrives are not reading the anomaly at all.

The 2021 paper that made the case for the relative index lists among its co-authors Erin Coughlan de Perez, Roop Singh and Maarten van Aalst, all of the Red Cross Red Crescent Climate Centre — the people who design anticipatory-action triggers for a living.

The instrument beneath the instrument

Every number in this piece comes from the same place. Both indices are computed from NOAA's Extended Reconstructed Sea Surface Temperature analysis, now in its sixth version, and ERSST is built entirely out of objects floating in the water: ship intakes, drifting and moored buoys, and Argo floats reporting from the top five metres. It takes no satellite input at all.

The exclusion is deliberate and long-standing. Satellites see about 95 percent of the ocean surface in a given month, against roughly 70 percent for the in-situ network since the 1970s, and NOAA's analysts have judged that coverage not worth its price: radiometers drift as their sensors degrade, aerosols and cloud contaminate the retrieval, and errors amplify at high viewing angles. So the thermometer of record is a scatter of physical instruments, and the space between them is filled by inference.

That step is not incidental. The published method calls the interpolation of observed anomalies "to areas without observations" the core of ERSST. Until recently the filling was done by a statistical decomposition of the historical field. In version six it is done by an artificial neural network — 2,058 of them, one for each month of the record — trained on the Hadley Centre's HadISST, which is a dataset that does blend satellite measurements. Satellite information reaches the tropical mean after all, then, not as an observation of this ocean this month but as a learned expectation about what an unobserved patch of ocean probably looks like.

The old index needed one box, five degrees of latitude by fifty of longitude, thick with instruments. RONI needs that box and an accurate anomaly for the entire tropical belt from 20° North to 20° South.

That belt is where the instruments have been going away. The TAO/TRITON moored array, the line of buoys along the equator that has been the backbone of ENSO monitoring since the 1980s and the only reliable source of subsurface equatorial ocean data, once numbered nearly seventy moorings. In 2012 NOAA decommissioned the ship Ka'imimoana, which had serviced the array since 1996. Without maintenance, data return fell to a record low of 30 percent. At about the same time, JAMSTEC, NOAA's principal partner since 2000, announced a phased withdrawal of the Japanese TRITON moorings from the western Pacific. Return recovered from 2014, but the withdrawal continued to its conclusion: the last TRITON sites were decommissioned in June 2021.

The current position, set out in Frontiers in Climate this July by a group writing on the western Pacific gap, is blunt. There are no near-equatorial moorings deployed west of 165° East, and none planned, despite the TPOS2020 project designating sites at 137°E, 147°E and 156°E as high priority. The Argo float array that covers the rest is at just over 65 percent of its recommended 494-float target in the 10°S–10°N band as of May 2026, and 80 percent of the floats in that band are contributed by a single country, the United States.

NOAA's reconstruction team names the weakness itself, without prompting. The shortcoming the neural network was introduced to mitigate is "low SST spatial-variabilities in the data sparse regions," and the improvement it delivers is, in the team's own assessment, "more pronounced in the tropical Pacific" than anywhere else. The place where the method most needed fixing is the place the index is looking.

The Climate Prediction Center says as much, in a notice at the foot of the RONI page that almost nobody will read. Because of a large-scale filter applied to the ERSSTv6 data, "RONI values may change up to two months after the initial 'real time' value is posted. Therefore, the most recent RONI values should be considered an estimate."

The estimate for May–June–July 2026 is +0.98. The boundary between a weak El Niño and a moderate one falls at +1.0.

The water

The MJJ 2026 season closed at 29.02 °C in the Niño-3.4 box: the warmest May–June–July on record, warmer than any reading taken during the 1997–98 event, and by the operational index, an estimate two-hundredths short of moderate.

Nothing in that sentence is contradictory. It is what happens when you index natural variability against the warming and then keep the old word for the result. RONI will go on being right about rainfall and circulation, and it will report ever-smaller El Niños in an ever-warmer ocean.

Under the new index, the largest El Niño in the record is no longer 2015–16 but 1982–83. Cole's divers found that event in every living coral they cored. They call the signature a death horizon: a dense band where the organisms that colonise an injured reef — algae, gastropods, barnacles — interrupt the lattice of the coral's own skeleton. The coral was not recording an anomaly. It was recording the temperature of the water around it, which in 1983 was high enough to stop it growing and let something else move into the wound. In 1983 the relative measure and the absolute one pointed the same way. The whole difficulty of the present is that they have come apart.

The corals are measuring degrees. So is the reef, the anchoveta, the fisher who owns the boat, the wheat. None of them subtracts the tropical mean.

Source note

Primary data. Both index files were downloaded from the Climate Prediction Center on 1 September 2026 and analysed independently: cpc.ncep.noaa.gov/data/indices/oni.ascii.txt (918 seasons, DJF 1950 – MJJ 2026, absolute SST and anomaly), RONI.ascii.txt (918 seasons), and sstoi.indices (monthly Niño 1+2, 3, 4 and 3.4, through July 2026). Decadal means, category counts, event peaks and rankings were computed from the raw files; the analysis script is retained. The commissioning editor's spot-checks of MJJ 2026 (+1.39 / +0.98), the 2023–24 peaks (+1.99 NDJ / +1.42 OND) and the 29.02 °C absolute reading were each confirmed exactly.

A caution for anyone repeating this work: the seasons in these files overlap, so adjacent rows are one month apart and a figure from one season pairs plausibly with a figure from the next, producing a gap that belongs to neither. October–December 2025 reads ONI −0.61 and RONI −0.98; November–January 2025 reads ONI −0.60 and RONI −1.04. Setting the −0.61 against the −1.04 overstates the divergence. An earlier version of this piece made that pairing and it was corrected before publication; the genuinely large gap in that cold event is the DJF 2025 figure given in the text.

Official notice. NWS Public Information Statement 26-05, issued 13 January 2026, effective 1 February 2026; and the CPC announcement page for RONI, dated February 2026. All quotations from both are byte-for-byte. The provisional-value notice quoted in the text appears on the CPC's RONI episode page, which also gives the current definition against ERSST version 6.

Literature. L'Heureux et al., "A Relative Sea Surface Temperature Index for Classifying ENSO Events in a Changing Climate," Journal of Climate 37(4), JCLI-D-23-0406.1 (2024). van Oldenborgh, Hendon, Stockdale, L'Heureux, Coughlan de Perez, Singh and van Aalst, "Defining El Niño indices in a warming climate," Environmental Research Letters 16: 044003 (2021). Deser and Tippett comments, and the Klavans and L'Heureux quotations, are from Bob Henson, "A new and better way to keep tabs on El Niño and La Niña," Yale Climate Connections, February 2026.

The underlying SST analysis. Huang, Yin, Boyer, Liu, Menne, Rao, Smith, Vose and Zhang, "Extended Reconstructed Sea Surface Temperature Version 6 (ERSSTv6): Part I. An Artificial Neural Network Approach," Journal of Climate 38, 1105–1121, doi:10.1175/JCLI-D-23-0707.1 (2025), and "Part II: Upgrades on Quality Control and Large-Scale Filter," Journal of Climate 38, 1123–1136, doi:10.1175/JCLI-D-24-0185.1 (2025). The input sources (ICOADS ships and buoys, Argo floats above 5 m), the exclusion of satellite measurements and the stated reasons for it, the description of interpolation to unobserved areas as the core of the method, the neural network trained on HadISST for 1950–2012, the "low SST spatial-variabilities in the data sparse regions" shortcoming and the tropical-Pacific improvement are all from Part I. This piece does not claim that the western Pacific mooring gap produces a specific error in the tropical mean; it establishes that the mean is an in-situ product, that the in-situ network in that belt has thinned, that the unobserved gaps are filled by inference, and that the agency itself labels recent values estimates.

Observing system. Frontiers in Climate, 10.3389/fclim.2026.1861050, published 16 July 2026, on the missing western Pacific moorings; and "A Decadal Overview of the Global Tropical Moored Buoy Array" (2023) for the 30 percent data-return figure following the Ka'imimoana decommissioning.

Downstream instruments. GlobalAgRisk documentation on the Extreme El Niño Insurance Product, including the ENSO Business Interruption Index Insurance concept note (Skees, 2009) specifying the Niño 1+2 region, the November–December averaging window, the 23.4 / 24.0 / 24.5 °C start triggers and the 27.0 °C exit trigger; and the 29 January 2013 GlobalAgRisk press release. Peru's ICEN definition and monthly values are from ENFEN Nota Técnica 01-2024 as published by IGP and IMARPE. The inter-agency ENSO standard operating procedures and IFRC Early Action Protocol descriptions are as cited. A third official figure exists and is left out of the text because it is not a three-month ENSO diagnostic and would flatter the argument: NOAA's own monthly file puts the Niño 1+2 anomaly for July 2026 at +3.56 °C against its fixed baseline. I could not establish that any specific catastrophe bond or sovereign parametric policy names ONI or RONI directly in its trigger language, and I am not going to guess.

Opening scene. The text and date of the Ryan Maue post of 1 September 2026 were confirmed through India Today's verbatim quotation of it, published the same morning. The engagement figures could not be independently confirmed, because the X interface available for this piece returned no data, and so no view, like or repost count is asserted here. The characterisation of the storms as a "once in a lifetime event" and as El Niño "going full-tilt" is Maue's and is attributed to him; no causal claim linking any specific 2026 storm to El Niño is made in this piece. For context on the season as a whole, the National Hurricane Center's eastern Pacific summary as of 09 UTC on 31 August 2026 recorded 12 named storms (three above normal), 3 hurricanes (two below normal), 2 major hurricanes (at normal), and accumulated cyclone energy of 64.6, one percent above the 1991–2020 normal. Hurricane Marie is not in that list; it formed after the summary was compiled.

Coral study. Cole, Thompson, Dyez, Tripp, Tudhope, Lofverstrom, Stevenson, Okun, Lawman, Conroy, Overpeck, Jimenez and Edwards, on Galápagos coral geochemistry across the past millennium, Science (2026), doi:10.1126/science.ady2660, published 27 August 2026. The quoted abstract language, the thirteen cores, the strontium-to-calcium and oxygen-isotope methods, the 4,500-year-old fossil boulders and the death horizons from the 1982–83 event are from the paper and from Cole's own first-person account of the fieldwork in The Conversation, 27 August 2026. The 36.5 percent figure is the increase in interannual sea surface temperature variability reported from the paper by New Scientist; the "nearly 40% stronger" framing is the University of Michigan announcement's rendering of the same result, and the two are given here separately because they are not the same statement. The coral cores cover roughly half of the thousand-year window, a limit the authors state; the millennial claim rests on that partial coverage.