The heat record is now being asked to file its own margin note
the heat record is now being asked to file its own margin note: a physics based model still beats graphcast, pangu weather and fuxi on record breaking extremes across nearly all lead times, while a subseasonal ai system cuts root mean square error by 20 to 60 percent on routine surface variables, and emulators smooth out the very extremes that kill.
the forecast now files its own failure log.
Context
A paper on arXiv (2508.15724), Numerical models outperform AI weather forecasts of record-breaking extremes, shows ECMWF's HRES model consistently outperforms the AI models GraphCast, GraphCast operational, Pangu-Weather, Pangu-Weather operational and Fuxi on record-breaking weather extremes, with AI errors larger for record-breaking heat, cold and wind across nearly all lead times. A separate arXiv paper of 30 July 2026, Weather Emulators at the Frontier of Heat Extremes Predictability, evaluates six emulators against dynamical systems at 10 to 15 day lead times and finds several rival or surpass physics-based forecasts in deterministic temperature skill at the cost of blurring.
The first result matches the note's physics-based model beating three AI models on record extremes. The 20 to 60 percent RMSE reduction from a subseasonal AI system on routine variables was not found, so it is unverified. The emulator paper is a different study, covering 10 to 15 days, and its blurring finding is consistent with emulators smoothing extremes, but it is not the note's source. The forecast now files its own failure log is the author's image.
Watch next
- The subseasonal 20 to 60 percent source.
Sources
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 4 October 2026 at 17:34 IST. Sources are the papers and datasets the note draws on.
View the original post ↗Embed this note
More notes
The air is now being asked to keep its own ledger
the air is now being asked to keep its own ledger: ecmwf’s aifs compo becomes the first ai model to forecast atmospheric composition globally every three hours, cleanair simulates 365 days of pm2.5 over china in ten seconds, and a unified framework maps six pollutants at one kilometer across the whole country. the air now files its own composition report.
read the note →The current is now being asked to draw its own map
the current is now being asked to draw its own map: china’s langya 2.0 predicts six ocean phenomena including internal waves and mesoscale eddies, a deep net called wenhai resolves eddies globally with air sea flux formulas built in, and scripps infers surface currents from the way temperature patterns deform in satellite images. the ocean now files its own circulation report.
read the note →The soil is now being asked to report its own carbon
the soil is now being asked to report its own carbon: a nix color sensor paired with generative data augmentation predicts soil organic carbon without a lab, random forest drives 74 percent of soil health mapping studies, and sentinel 2 tracks five year carbon change across france and italy from 922 samples. the dirt now files its own carbon account.
read the note →