JMA’s September Niño 3 of +4.0°C beats the +3.6°C of December 1997. It is a monthly observation in one region, and it is not a RONI record. Here is what holds, and what does not carry over.
On 9 October the Japan Meteorological Agency said El Niño conditions have reached their strongest intensity on record. The number behind that sentence is a September NINO.3 sea surface temperature deviation of +4.0°C, against a previous record of +3.6°C in December 1997, in a series that starts in 1949. We checked it against JMA’s own data files, and it holds.
What it measures is narrower than the headline sounds. It is one monthly value for one region, from one agency’s analysis. NOAA’s RONI for July to September is +1.69°C, well short of its own record, and that does not make either one wrong.
JMA’s outlook says the September NINO.3 deviation was +4.0°C, exceeding the +3.6°C of December 1997, the highest since records began in 1949. Its monitoring table has August at +3.4°C and September at +4.0°C, with the latest values marked preliminary. JMA’s coupled model expects NINO.3 to keep rising until boreal winter, with the event likely to peak on record, and calls the continuation of the strongest conditions virtually certain. We read the text, not the model figure behind it. JMA’s next update is on 10 November.
We pulled JMA’s index files to check the comparison. On the sliding 30-year baseline, September 2026 is +4.0°C and the next-highest values are December 1997 at +3.6°C, then November 1997 and January 1998 at +3.5°C. The table below shows how 2026 has tracked 1997. The two years matched through July, and 2026 pulled ahead in August.
| JMA NINO.3 monthly deviation | Jun | Jul | Aug | Sep | Oct | Nov | Dec |
|---|---|---|---|---|---|---|---|
| 2026 | +1.9 | +2.5 | +3.4 | +4.0 | pending | pending | pending |
| 1997 | +1.8 | +2.5 | +3.0 | +3.2 | +3.4 | +3.5 | +3.6 |
JMA defines an El Niño event with the five-month running mean of NINO.3, not a single month. That measure was +2.6°C for July 2026, and it cannot be computed for August or September until October and November are in. In JMA’s files the highest five-month value is +3.4°C, in November 1997. So “strongest on record” is a statement about the monthly value. On JMA’s own event definition, the record has not been broken yet, and may be.
NOAA’s RONI is a different measurement in three ways: it is the Niño 3.4 region, it is a three-month mean, and it is relative to the tropical average. For July to September it is +1.69°C, and the highest value in CPC’s RONI file is +2.4°C in December to February 1982 to 1983. CPC’s 8 October discussion gives a 54% chance that RONI reaches +2.5°C or more in September to November, 83% for October to December and 70% for November to January, which we cover in our WMO forecast piece.
The table sets September against each yardstick we could read. The ordering holds in every series that has a comparison: September 2026 is at or near the top in JMA’s two baselines and in CPC’s monthly OISST file. The size of the number does not hold. Niño 3 for September is +4.0°C in JMA’s analysis, +3.64°C in CPC’s conventional file and +3.0°C in CPC’s discussion, which uses monthly relative values.
| Yardstick (September 2026) | Value | Earlier peak in the same series |
|---|---|---|
| JMA NINO.3 monthly deviation, latest sliding 30-year mean | +4.0°C | +3.6°C, December 1997 (since 1949) |
| JMA NINO.3 monthly anomaly, fixed 1991 to 2020 base | +4.0°C | +3.4°C, December 1997. August 2026 (+3.5°C) already topped it. |
| CPC monthly OISST index file, Niño 3 (conventional) | +3.64°C | +3.07°C, December 1997 (since 1982) |
| CPC discussion of 8 October, Niño-3 (monthly relative) | +3.0°C | Not read |
| CPC monthly OISST index file, Niño 3.4 (conventional) | +2.84°C | +2.72°C, November 2015 (since 1982) |
| NOAA RONI, July to September, Niño 3.4, three-month, relative | +1.69°C | +2.4°C, December to February 1982 to 1983 |
The gap between JMA and the CPC conventional file is 0.36°C. Most of it is absolute temperature: JMA has Niño 3 at 28.9°C and CPC at 28.65°C, with the reference values at about 24.9°C and 25.0°C. Both are rounded, so treat the split as approximate. The baseline does not explain it, because JMA’s sliding and fixed 1991 to 2020 versions both give +4.0°C for September.
JMA’s readme says its SST indices come from the MGDSST analysis after June 2015 and from COBE-SST2 before that. The 1997 record and the 2026 value therefore come from different datasets. We have not tested what that splice does to the comparison, and JMA’s outlook does not discuss it. CPC’s OISST file gives the same ordering, +3.64°C against +3.07°C, from a third dataset and a shorter record starting in 1982, which supports the ranking but not the size of the margin. Conventional indices also carry the long-term warming of the tropics, which is the problem relative indices are designed to address. The record is real in these series and still a statement about a month.
This is also an eastern Pacific record. Niño 3 sits in the region our earlier pieces covered in the Niño region explainer and the east-west gradient piece. CPC’s file has Niño 1+2 at +4.69°C in September, the highest value in that file, and Niño 3.4 at +2.84°C, above the earlier high of +2.72°C in November 2015. Those are conventional values, not the relative ones RONI uses.