The Met Office says its GloSea system is forecasting Niño 3.4 sea surface temperatures above 3°C, and its head of long-range forecasting says he has never seen a signal this intense. That is real news. It is also about to be set beside NOAA’s 69% chance of a “historic” event at RONI +2.5°C, as though 3 beats 2.5. It does not, because nothing published so far establishes that the two numbers measure the same thing.
On 21 August the Met Office published a blog post opening with a sentence that is unusual for a national meteorological service: “2026 is going to be renowned for the development of the largest El Niño in living memory and what is probably the largest such event since the 19th century.”
The number behind it comes from the caption of the forecast chart. Sea surface temperature forecasts from the GloSea system for the Niño 3.4 region are “now showing record values of more than 3°C in the coming months.”
That is a genuinely significant escalation and it deserves to be reported as one. It is also the exact moment when two numbers from two agencies get stacked into a league table they do not belong in.
Professor Adam Scaife, the Met Office’s Head of Long Range Forecasting, is quoted directly:
Read that conditional carefully, because it is doing real work: if the forecast is accurate. Scaife is describing the intensity of a model signal, not an observation. The blog also notes that other forecast systems show similar extreme values, which is the strongest part of the case, and one this site has written about before in the context of how ENSO models work and which ones earn their keep.
The Met Office adds a set of impact expectations, all framed as shifted odds rather than outcomes: heightened drought risk for southern Africa and northeast Australia; low-rainfall risk for Central America, tropical South America and west Pacific nations; increased chances of wetter, stormier conditions for northwest Europe including the UK through autumn and early winter. It notes the Indian summer monsoon is already running well below normal rainfall, and that El Niño suppresses Atlantic hurricanes through wind shear, with only three storms so far this season.
Separately, Dr Nick Dunstone is quoted saying 2027 is “very likely to replace 2024 as the warmest year on record and be another year to temporarily exceed +1.5 C above the preindustrial period.” Note the mechanism he gives: large El Niño events release more ocean heat to the atmosphere, particularly in the calendar year after the winter peak. El Niño supplies the spike. That the elevated baseline it spikes from is set by long-term human-caused warming is this site’s reading, drawn from established attribution science rather than from the post. What the Met Office’s Chief of Science and Technology, Professor Stephen Belcher, says in the same post is that “this large El Niño is a reminder that natural variations in the climate system are also at play, adding to the stresses of climate change.”
NOAA’s Climate Prediction Center said on 13 August that for October through December there is a 69% chance of a historic event, which it defines as a three-month RONI value at or above +2.5°C, a level that would exceed every event back to 1950.
So one agency has 3.0 and another has 2.5, and the arithmetic writes itself. It is wrong. Here is the same information laid out by what each number actually is:
| Met Office | NOAA CPC | |
|---|---|---|
| The number | More than 3°C, a GloSea Niño 3.4 SST forecast | +2.5°C on RONI |
| What kind of number | A forecast value | A threshold, with a 69% probability attached |
| Dataset | Not stated in the Met Office post | ERSSTv6 |
| Index construction | Not stated in the Met Office post | RONI: Niño 3.4 minus the 20°N-20°S tropical mean, then variance-adjusted to the original Niño 3.4 index |
| Region | 5°S-5°N, 120°W-170°W | 5°S-5°N, 170°W-120°W |
| Averaging window | Not stated in the blog post | Three-month running mean |
| Baseline | Not stated in the blog post | 1991-2020 |
| Source system | GloSea, a single forecast system | Official outlook from a ~10-member team using multi-model guidance |
Only the geographic region is confirmed to match. The ocean box is identical, just written in the opposite order. Other attributes either differ or are not stated, and that second category is the important one: the Met Office post does not publish the index construction, baseline, averaging period or dataset behind its 3°C. Without those, the figure cannot be placed on the same scale as RONI even in principle.
The Oceanic Niño Index reports how much warmer the Niño 3.4 region is than its own 1991-2020 average. RONI is built in two steps, and the second one is routinely left out of explanations, including an earlier draft of this one. NOAA’s own definition is a three-month running mean of ERSSTv6 Niño 3.4 anomalies “with average tropical mean (20°N-20°S) SST anomalies subtracted”, after which “the difference is then adjusted so the variance equals the original Niño 3.4 index.” So RONI starts by subtracting the tropical-mean anomaly, then rescales the resulting series so its variance matches Niño 3.4.
This is not a theoretical gap. For the most recent completed season, May-June-July 2026, both indices are published and both can be read off NOAA’s own files:
| Season | ONI | RONI | Difference |
|---|---|---|---|
| MJJ 2026 | +1.39°C | +0.98°C | 0.41°C |
During MJJ 2026 the wider tropical oceans were also anomalously warm, so the relative construction produced a lower value than ONI after NOAA’s variance adjustment. Note what that 0.41°C is and is not: it reflects the complete RONI construction, both the subtraction and the rescaling, not simply the temperature of the wider tropics. The sign and size of that difference vary by season, and RONI is not guaranteed to sit below ONI in all conditions. It is lower now.
The practical consequence is straightforward. Set a Niño 3.4 forecast of 3°C beside a RONI threshold of +2.5°C and the apparent half-degree of extra intensity is not a measured margin. It is the residue of two quantities built differently, one of which has not published how it was built.
Do not convert. The obvious next step is to subtract 0.41 from the Met Office number and call it a RONI equivalent. Do not. That offset is not a constant. It depends on how warm the wider tropics run in a given season, which changes, and it has been growing over recent decades. It is also derived from observations, not from the GloSea forecast, whose own baseline and averaging window are not stated in the blog post. There is no honest arithmetic that turns one of these numbers into the other. The correct move is to keep them in separate lanes and say what each one is.
Quite a lot, actually, without any cross-index arithmetic:
What cannot be said is that the Met Office forecast “exceeds” NOAA’s historic threshold, or that the two agencies disagree about severity. Nothing in either statement supports that. This site has already worked through the trap on the NOAA side of it, where a widely quoted +2.6°C figure turns out to be the wrong index for the threshold it is being compared against. The Met Office number is the same error waiting to happen from the other direction.
The testable version. The claim here is narrow: the Met Office forecast is a major escalation, and it cannot be ranked against NOAA’s +2.5°C RONI threshold.
What would support the broader escalation: NOAA raises its RONI strength probabilities at the next ENSO Diagnostic Discussion, due 10 September 2026, or other major forecast systems retain similarly extreme values. This is independent corroboration that a very large event is coming. It is not validation of GloSea’s number, because a NOAA RONI probability cannot verify a Met Office Niño 3.4 forecast.
What would directly support the Met Office forecast: successive GloSea initialisations retain the peak above 3°C, and observations rise toward it measured on the same Niño 3.4 definition the forecast uses.
What would weaken it: later GloSea runs retreat materially from 3°C, or matching observations stay well below the projected trajectory.
What would remove the comparability objection: the Met Office documenting that its plotted value uses the same RONI construction, baseline and averaging period as NOAA. That would retire the different-metric problem. It still would not make a deterministic forecast value equivalent to a probability of crossing a threshold.
Where it gets settled: NOAA’s RONI and ONI series, the CPC diagnostic discussions, and the Met Office’s own El Niño forecast pages. All are public.
The Met Office has done something useful here: it put a specific number in public, attached a named scientist to it, and stated the conditional. The number is striking on its own terms. It does not need to be turned into a scoreboard against NOAA to matter, and turning it into one would misrepresent both agencies.
Method and data vintage. Met Office figures and quotations are from its blog post of 21 August 2026. NOAA probabilities are from the CPC ENSO Diagnostic Discussion of 13 August 2026 and the Official NOAA CPC ENSO Probabilities page issued August 2026. ONI and RONI values were retrieved from NOAA’s ASCII series on 22 August 2026 using cache-busted requests, because the CPC pages can otherwise serve an earlier cached copy. The Met Office post does not state the index construction, baseline, averaging period or dataset behind its 3°C figure, and this article does not infer them. It is described throughout as a GloSea Niño 3.4 SST forecast and is not classified as conventional or relative. The RONI definition quoted here is NOAA’s own, from its RONI product page. NOAA can revise recent seasonal RONI and ONI values for up to two months after first posting. Where a number here differs from the live dashboard, the dashboard carries the newer figure.