Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2605.30184 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866918529837563904 |
|---|---|
| author | Lehmann, Fanny Ozdemir, Firat Cheng, Yun Hoefler, Torsten Schemm, Sebastian Soja, Benedikt Mishra, Siddhartha |
| author_facet | Lehmann, Fanny Ozdemir, Firat Cheng, Yun Hoefler, Torsten Schemm, Sebastian Soja, Benedikt Mishra, Siddhartha |
| contents | While AI weather models excel at short-to-medium range forecasts (up to 15 days), they frequently suffer from ill-defined "instabilities" when rolled out over longer horizons. This work addresses the lack of a formal taxonomy by categorizing these failures into three distinct regimes: blow-up, drift, and loss of seasonality, through year-long rollouts of nine state-of-the-art AI weather models. Our analysis reveals that stability hinges on the treatment of small spatio-temporal scales: unstable models amplify high-frequency energy, while stable models act as denoisers when noise is added to their inputs. Far from reducing these models to mere stochastic parrots, our findings highlight that stable models generate unique weather trajectories, conditioned on the initial state. We verify our findings through ablation studies on architectural design choices, conducted using state-of-the-art Vision Transformer (ViT) AI weather model architectures. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_30184 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Can AI Weather Models Predict Beyond Two Weeks? A Quantitative Benchmark and Analysis of Long Rollouts Lehmann, Fanny Ozdemir, Firat Cheng, Yun Hoefler, Torsten Schemm, Sebastian Soja, Benedikt Mishra, Siddhartha Machine Learning Atmospheric and Oceanic Physics While AI weather models excel at short-to-medium range forecasts (up to 15 days), they frequently suffer from ill-defined "instabilities" when rolled out over longer horizons. This work addresses the lack of a formal taxonomy by categorizing these failures into three distinct regimes: blow-up, drift, and loss of seasonality, through year-long rollouts of nine state-of-the-art AI weather models. Our analysis reveals that stability hinges on the treatment of small spatio-temporal scales: unstable models amplify high-frequency energy, while stable models act as denoisers when noise is added to their inputs. Far from reducing these models to mere stochastic parrots, our findings highlight that stable models generate unique weather trajectories, conditioned on the initial state. We verify our findings through ablation studies on architectural design choices, conducted using state-of-the-art Vision Transformer (ViT) AI weather model architectures. |
| title | Can AI Weather Models Predict Beyond Two Weeks? A Quantitative Benchmark and Analysis of Long Rollouts |
| topic | Machine Learning Atmospheric and Oceanic Physics |
| url | https://arxiv.org/abs/2605.30184 |