Medium-range weather forecasts from a graph neural network beat the operational physics model
GraphCast produced more accurate 10-day forecasts than ECMWF's HRES on roughly 90% of the evaluated variable and lead-time combinations, in under a minute on a single TPU against hours on a supercomputer.
- Model
- GraphCast
- Field
- Climate science
- Date
- 2023-11-14
- Human collaborators
- Remi Lam, Peter Battaglia
What was found
GraphCast takes the two most recent global atmospheric states and predicts the next state six hours ahead on a roughly 0.25° grid, applied repeatedly to reach ten days. It is trained on ERA5 reanalysis rather than on the equations of motion, and it beat HRES, the operational gold standard, on the large majority of evaluated variables, rising to almost all of them within the troposphere. It also flagged severe weather events, including cyclone tracks, earlier than the physics model despite never being trained to look for them. Published in Science, with code and weights released.
Novelty check
Machine-learning weather models existed before (FourCastNet, Pangu-Weather), and Pangu had already reported beating HRES on some measures. GraphCast's contribution is the breadth of the win across variables and lead times, at operational resolution, with public code. That is what moved national weather services to adopt learned models.
Caveats and known objections
GraphCast is trained on ECMWF reanalysis and initialised from ECMWF analyses, so it depends on the physics-based system it is measured against rather than replacing it. Deterministic models trained on a mean-squared-error objective blur fine structure and underestimate extremes such as peak cyclone intensity. It has no conservation guarantees and cannot forecast anything outside its trained variables. Later ensemble models such as GenCast address part of this.
Nobody outside the lab has checked this yet.
Reading the primary source closely enough to say whether it supports the claim counts as a check, and you are credited on the entry.
Or on GitHub: submit a check challenge the grade send a correction or send a pull request
Flag this for triage
Signals order the review queue and nothing else. They are never published, and they never move a grade: that takes a citation.
Entry history (1 event)
- AddedEntered the registry graded Peer reviewed and AI-led.
Entries are never deleted. A grade that does not hold up is downgraded on the record, with the reason beside it.
Graded peer reviewed for verification and ai-led for autonomy. What these mean.
Cite this entry
whataifound.org. (2023). Medium-range weather forecasts from a graph neural network beat the operational physics model. whataifound.org: A Registry of AI Scientific and Mathematical Discoveries. https://whataifound.org/finding/2023-11-14-graphcastBibTeX
@misc{whataifound-googledeepmind-2023-graphcast,
title = {Medium-range weather forecasts from a graph neural network beat the operational physics model},
author = {{whataifound.org}},
year = {2023},
howpublished = {whataifound.org: A Registry of AI Scientific and Mathematical Discoveries},
note = {Result by Google DeepMind. Verification: Peer reviewed. Autonomy: AI-led.},
url = {https://whataifound.org/finding/2023-11-14-graphcast}
}