WMO Forecasts +3.7°C El Niño Warmth. That Is Not NOAA’s RONI.

The World Meteorological Organization says its multi-model forecast for the Niño 3.4 region is +3.7°C for October to December 2026 (standard deviation 0.4°C), peaking at +3.9°C (0.5°C) in December. That is a forecast of a conventional sea surface temperature anomaly. It is not NOAA’s relative index (RONI), and no amount of subtracting turns one into the other. The 1993 to 2009 baseline comes from the captions on the report’s Figures 7 and 8. Table 2 itself states none, so for the table it is inferred.

Nothing in WMO’s material passes the forecast off as an observation. Its graphic labels the plot correctly, as a forecast from leading global centres, issued September 2026. The confusion comes from everything around the number: the baseline, the dates, month versus season, what the ± means, and the “strongest on record” line in the graphic’s headline. This post takes those one at a time, using the report and WMO’s press release, and checks the record claim against NOAA’s files.

What the WMO report forecasts

Table 2 gives multi-model forecasts for the Niño regions, each with a standard deviation. For Niño 3.4 it is +3.5°C for October, +3.8°C for November and +3.9°C for December, with +3.7°C for the three-month season. The Summary says the warming peaks around December 2026 and that the individual forecast systems stay in close agreement, which WMO reads as high forecast confidence. Here is the full table.

Two things stand out. The warmth is concentrated in the east (Niño 1+2 at +4.1°C and Niño 3 at +4.2°C for the season, against +1.9°C in Niño 4), which is the shape of an east-based event. Our post on why Niño 1+2 matters covers that. And what has actually been observed is lower. WMO’s own Table 1 has Niño 3.4 at +1.6°C in June, +2.0°C in July and +2.5°C in August, a June to August mean of +2.0°C against 1991 to 2020.

We kept the two apart on purpose. The report does not say which dataset its observed line comes from, or how the model forecasts are calibrated to observations, so putting CPC’s values and WMO’s forecast on one continuous axis would imply a match nobody has shown. By eye, WMO’s observed line looks somewhat lower than CPC’s published values for April to August. That is an eyeball estimate from a small plot, and we are not reading anything into it about WMO’s data or calibration.

Three dates, three meanings

Preparation, initialization and public release are three different dates, and the report’s numbers belong to the first two.

So the forecast was built before the full September monthly observation was available. September initial conditions already include observations up to the start date, but the monthly value for the whole month did not exist yet. CPC’s monthly file now shows September 2026 at +2.84°C on its 1991 to 2020 anomalies. Table 2 starts at October, and Figure 7’s dashed lines run from the August observation to the October forecasts, so there is no September forecast to hold up against that number.

December’s peak is not the seasonal mean

The graphic says the event peaks in December. That is one month, at +3.9°C (standard deviation 0.5°C). The +3.7°C figure is the October to December average, so it sits below the peak month by construction. They answer different questions, so each needs its own matching observation: December against a monthly value, October to December against a three-month average. WMO’s record comparison uses November 2015 to January 2016, which is a different three months from October to December.

Why +3.7°C is not NOAA’s RONI

WMO’s report and press release never mention the ONI or the RONI, and nothing suggests the forecast was built on either. Their numbers sit on a different construction and baseline from NOAA’s relative index. This table lines up the quantities side by side. It is a labeling aid, not a conversion chart.

Don’t subtract. For the same Niño 3.4 box in July to September 2026, a simple average of CPC’s three monthly OISST values is +2.46°C, the ONI is +2.16°C and the RONI is +1.69°C. Those differ by dataset, averaging, baseline and, for the RONI, the tropical-mean adjustment. The gaps are not conversion factors. The ONI-RONI gap moves with how warm the wider tropics are, as our Met Office comparison showed. NOAA states its threshold for a historic event in RONI terms (+2.5°C for a three-month value), and the largest seasonal RONI in CPC’s file is +2.40°C for December to February 1983. A conventional-index forecast of +3.7°C can’t be placed against that.

The baseline isn’t the reason for the numerical gap in CPC’s observations either. In CPC’s monthly files the 1993 to 2009 and 1991 to 2020 climatologies for the Niño 3.4 box differ by 0.05°C or less for October, November and December (about 0.2°C in January to March), so rebasing the observed series moves late-2026 values by hundredths. That is only about CPC’s observations. It does not show that WMO’s model forecasts are directly comparable with NOAA observations, because the report does not say which observed dataset it uses or how forecast anomalies are calibrated to it.

Is it the strongest on record? What reproduces and what does not

WMO says this a few different ways. The graphic’s headline says the event is forecast to peak in December as the strongest on record. The press release says sea surface temperatures are forecast to reach record levels, that +3.7°C would beat the previous record of +2.6°C (November 2015 to January 2016), and that the event could be one of the strongest since at least 1950. Those are not the same claim, since “since at least 1950” and “on record” cover different periods. The report itself, which we read in full, uses neither phrase.

What reproduces. The historical side holds up. The record three-month Niño 3.4 value is +2.59°C in NOAA’s ONI file for November 2015 to January 2016, matching WMO’s +2.6°C, and +2.5°C to +2.7°C in CPC’s ERSST and OISST monthly files on either baseline (table above). For October to December specifically, the highest observed value is in 2015: ONI +2.45°C, and about +2.5°C to +2.6°C in the monthly files rebased by this site. CPC’s published September 2026 value of +2.84°C already tops the highest earlier monthly value in the same OISST file, +2.72°C (November 2015), which goes back to 1982. The ERSST file we could read stops at June 2026, so we can’t make that check there yet.

What doesn’t. The forecast side. The report does not say whether model anomalies are measured against each system’s own hindcast climatology, whether any bias correction was applied, or which observed dataset the forecast is meant to verify against. So a +3.7°C forecast can’t just be set above an observed +2.6°C as if they were the same quantity, and matching the 1993 to 2009 baseline would not fix that. There is also no probability of beating the record, and the forecast has not verified yet. “Record” is WMO’s attribution, and we have only checked the observed record it is compared with. We did not check years before 1950, which matters because the press release also says “since observations began.”

A standard deviation is not a confidence interval

Table 2’s caption says the forecasts come “with standard deviation” and does not define it. The most natural reading is the spread across forecast systems, and Figure 7 fits: the individual December forecasts run from roughly 2.9°C to 4.8°C by eye, with the multi-model line near 3.9°C. If that is right, the ±0.5°C measures how much the models disagree. It does not tell you the chance the real value lands in any range. The models share physics and errors, so a narrow spread means they agree, not that they are right.

It also can’t be turned into a probability of a record. Setting +3.7°C against the +2.6°C record and counting standard deviations would imply a calibrated predictive distribution, which the report does not claim to provide. Its own guidance says seasonal forecast skill is substantially lower than weather forecast skill and varies considerably by region and season.

Forecast warmth is not uniform regional impact

The ocean forecast is the strongest part of the update. The regional maps are softer. They show probabilities of tercile categories against the 1993 to 2009 record, and the report is explicit that colors show probability and not certainty, that deeper shades mean a higher probability of the shaded category and not a larger anomaly, and that the update is guidance and not an official forecast for any region.

The signals are uneven. Above-normal sea surface temperature probabilities exceed 80% across the equatorial Pacific east of the Date Line, and rainfall probabilities show a strong wet signal there flanked by dry areas. The Indian Subcontinent has an enhanced below-normal rainfall probability with moderate-to-high model consistency. Europe shows only a weak enhancement for above-normal rainfall, with low-to-moderate consistency. The Greater Horn of Africa leans wet, though consistency across Africa is low to moderate. For temperature, above-normal land probabilities are nearly everywhere but weaker over parts of northern Asia and northwestern and central North America, and below-normal probabilities appear over the subpolar North Atlantic and a zone of the South Pacific.

Part of that warm tilt over land is background warming since 1993 to 2009 and not El Niño itself. That is our reading, not WMO’s. A bigger Niño 3.4 anomaly also does not mean a proportionally bigger effect anywhere. Our post on the EUMETSAT study of Europe shows how a composite of past events splits regionally, and why one summer’s weather can’t be assigned to El Niño from a historical average.

What this post is not

In August we compared the Met Office GloSea forecast above +3°C with NOAA’s +2.5°C RONI threshold. This post is about WMO’s new multi-model release. The Exeter system is one of the contributors in Figure 7’s legend, but WMO’s number is an ensemble statement with its own baseline and dates. The ONI versus RONI construction argument stays with that earlier post.

What NOAA said on 8 October

CPC’s ENSO Diagnostic Discussion issued on 8 October 2026 keeps the ENSO Alert System at El Niño Advisory. The synopsis says El Niño continues to strengthen, with a strong-to-very-strong event likely through January to March 2027, a chance it describes as remaining greater than 83%. CPC reports sea surface temperature anomalies above +4.0°C in parts of the eastern equatorial Pacific, and Niño index values of +2.1°C in Niño-3.4, +3.0°C in Niño-3 and +3.9°C in Niño-1+2.

On the outlook, CPC puts the chance of a historic event, one that would exceed previous El Niño events back to 1950 and is defined as +2.5°C or more for a three-month RONI value, at 54% for September to November, 83% for October to December and 70% for November to January. It adds that impacts consistent with El Niño are more likely with an event this big, though not guaranteed. The next discussion is scheduled for 12 November 2026.

Compared with September, the October to December historic-event chance rose from 75% to 83%. CPC’s strength table now puts the chance of a very strong event, a RONI of +2.0°C or more, at 100% for October to December, up from 98% in September. The synopsis wording shifted too. September led with a greater than 90% chance of a very strong event during the fall and winter. October leads with a strong-to-very-strong event being likely through January to March 2027, and the synopsis no longer carries a separate very-strong percentage. The latest seasonal values have not moved: July to September is still the newest season, with the ONI at +2.16°C and the RONI at +1.69°C.

These numbers answer a different question from WMO’s. NOAA is giving the odds that a three-month RONI reaches +2.5°C. WMO is giving a forecast of a conventional index, with no probability attached. The +2.1°C in CPC’s discussion also isn’t the +2.84°C September value from the OISST file earlier in this post. The discussion does not label its basis, but its quoted values match the monthly averages of CPC’s weekly relative Niño series in all four regions for August and in the three regions it gives for September, while conventional OISST values run about 0.6°C to 0.85°C higher. So they appear to be monthly relative Niño-region anomalies. They are not the three-month RONI, and they should not be compared directly with conventional OISST values. NOAA’s 83% and WMO’s +3.7°C agree on direction and not on units.

What to Watch

• CPC’s next ENSO Diagnostic Discussion on 12 November 2026, to see whether the 83% October to December chance of a historic event (a three-month RONI of +2.5°C or more) moves.
• October’s monthly Niño 3.4 value in CPC’s sstoi.indices file, due in early November, to set beside WMO’s +3.5°C forecast (standard deviation 0.3°C).
• WMO’s next Global Seasonal Climate Update, to see whether the Niño 3.4 forecast is revised and whether the baseline and the standard deviation get defined.
• The ERSST monthly file, which ends in June 2026 and will be needed to check whether September already sets a monthly record on that dataset.