Learning skillful medium-range global weather forecasting
Problem
Producing accurate 10-day global weather forecasts, traditionally requiring heavy compute rather than learning from historical weather data.
Why it matters
Medium-range forecasts are critical to decision-making across many social and economic domains.
What the AI did
GraphCast, a machine learning method trained directly on reanalysis data, predicts hundreds of weather variables for 10 days at 0.25° resolution in under 1 minute.
Results
Outperforms the most accurate operational deterministic systems on 90% of 1380 verification targets; supports severe event prediction.
Impact
Match or beat operational forecasting systems on most verification targets at far greater speed.
Record
| Venue | Science |
|---|---|
| Published | 2023 |
| Work type | article |
| Authors | Rémi Lam, Álvaro Sánchez‐González, Matthew Willson, Peter Wirnsberger, Meire Fortunato, Ferran Alet, Suman V. Ravuri, Timo Ewalds et al. |
| Citations | 1,416 (OpenAlex, at import time) |
| References | 54 |
| Model or method | GraphCast |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1126/science.adi2336 |
| Open access | hybrid |
| Keywords | weather forecasting, machine learning, GraphCast, reanalysis data, tropical cyclones, extreme events |
Importance rating
major — millions of lives, an industry or a field
Abstract states global medium-range forecasting is critical to decision-making across many social and economic domains; outperforms operational systems on 90% of 1380 targets.