Extremes & hazards★ Highlight
Can AI weather models predict out-of-distribution gray swan tropical cyclones?
Why it matters. Shows FourCastNet cannot extrapolate to unseen Category 5 cyclones when trained without strong storms. This is a key limit on AI forecasting of gray swan extremes.
In one sentence. FourCastNet trained without Category 3-5 cyclones cannot forecast Category 5 storms, showing AI weather models fail to extrapolate to unseen extremes.
Abstract
Predicting gray swan weather extremes, which are possible but so rare that they are absent from the training dataset, is a major concern for AI weather models and long-term climate emulators. An important open question is whether AI models can extrapolate from weaker weather events present in the training set to stronger, unseen weather extremes. To test this, we train independent versions of the AI model FourCastNet on the 1979-2015 ERA5 dataset with all data, or with Category 3-5 tropical cyclones (TCs) removed, either globally or only over the North Atlantic or Western Pacific basin. We then test these versions of FourCastNet on 2018-2023 Category 5 TCs (gray swans). All versions yield similar accuracy for global weather, but the one trained without Category 3-5 TCs cannot accurately forecast Category 5 TCs, indicating that these models cannot extrapolate from weaker storms. The versions trained without Category 3-5 TCs in one basin show some skill forecasting Category 5 TCs in that basin, suggesting that FourCastNet can generalize across tropical basins. This is encouraging and surprising because regional information is implicitly encoded in inputs. Given that current state-of-the-art AI weather and climate models have similar learning strategies, we expect our findings to apply to other models. Other types of weather extremes need to be similarly investigated. Our work demonstrates that novel learning strategies are needed for AI models to reliably provide early warning or estimated statistics for the rarest, most impactful TCs, and, possibly, other weather extremes.
Labels
- Topic
- Extremes & hazards, Explainability, uncertainty & physics
- Type
- Evaluation or intercomparison of existing models
- Method
- Neural operator (FNO, SFNO, ...)
- Time scale
- 3–15 days
- Domain
- Global
- Data
- ERA5 / reanalysis
- Status
- Research
- Identifiers
- arXiv:2410.14932 DOI:10.1073/pnas.2420914122 OpenAlex:W4410535525
- Added
- 2026-10-22
Labels and summary are generated by Claude and reviewed weekly. Wrong? Edit this entry.