Zeke Hausfather (2026), The Climate Brink (Substack newsletter of Andrew Dessler and Zeke Hausfather) — https://www.theclimatebrink.com/p/the-strongest-el-nino-ever
Key takeaways
- In the post, Hausfather reports that the July 2026 initializations of 14 seasonal models (667 ensemble members) put the multi-model median peak Niño 3.4 anomaly for 2026-27 at about 3.6 °C, against a prior record of about 2.75 °C (2015-16). That gives about a 91% chance of a record by Niño 3.4 and about 77% by relative ONI. He cautions that no model has ever been verified in this range.
Seeded from the abstract and a skim, before the reading in NOTE-031. What follows is what the work says about itself.
Hausfather argues that, with the July model runs in, the developing 2026 El Niño is very likely to be the strongest on record, possibly by a wide margin. He compares the multi-model forecast with every event since 1877, looks at the per-model spread and at the relative index that removes tropical-mean warming, and traces how forecasts rose from March to July. He also notes that observed daily Niño 3.4 anomalies are already at “super El Niño” levels in mid-July. He expects most of the global-temperature effect in 2027, and ends by warning that the models are untested at this intensity.
Standing in the record
Filed from the survey of 2026-09-25 of work the anthology set aside as out of scope (tier C): 345 seconds of active reading over 1 session in the papers-feed tracker. It was filed Deferred, unread. NOTE-031 is the close reading of 2026-09-25, and it placed the work: Proposed — This is a dated forecast. What settles it is the observed 2026-27 peak, i.e. monthly ERSSTv5 Niño 3.4 and RONI through about February 2027. The observations to mid-September are consistent with a record but do not yet reach 3.6 °C.
When seeded, before the reading in NOTE-031, the entry gave a priority for a deeper reading: low — Fully read. A time-bound blog post with no bearing on the anthology.
What the seeding asked a deeper reading to check (NOTE-031 is that reading):
- This is a dated forecast. Whether the event verified is now checkable, and a reader should check it rather than rely on the post.
- It has no ML content. The survey’s “drop: a blog post” is reasonable. Its only possible use is as an example of ensemble forecasting outside the training distribution.
Access when seeded: I fetched and read the full post: body text, figure captions and footnotes. The page metadata gives datePublished 2026-07-13, and the byline is Zeke Hausfather. The post’s subtitle is “With all the July runs in, dynamical models give a ~90% chance of a record-setting event”.