{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,31]],"date-time":"2026-10-31T11:41:00Z","timestamp":1793446860203,"version":"4.5.0"},"reference-count":70,"publisher":"National Academy of Sciences","issue":"21","license":[{"start":{"date-parts":[[2025,5,20]],"date-time":"2025-05-20T00:00:00Z","timestamp":1747699200000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by-nc-nd\/4.0\/"}],"funder":[{"DOI":"10.13039\/100000006","name":"DOD | USN | Office of Naval Research","doi-asserted-by":"publisher","award":["N000142012722"],"award-info":[{"award-number":["N000142012722"]}],"id":[{"id":"10.13039\/100000006","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/100000183","name":"DOD | USA | AFC | CCDC | Army Research Office","doi-asserted-by":"publisher","award":["W911NF-22-2-0124"],"award-info":[{"award-number":["W911NF-22-2-0124"]}],"id":[{"id":"10.13039\/100000183","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/100000159","name":"NSF | GEO | Division of Atmospheric and Geospace Sciences","doi-asserted-by":"publisher","award":["AGS-2046309"],"award-info":[{"award-number":["AGS-2046309"]}],"id":[{"id":"10.13039\/100000159","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":["www.pnas.org"],"crossmark-restriction":true},"short-container-title":["Proc. Natl. Acad. Sci. U.S.A."],"published-print":{"date-parts":[[2025,5,27]]},"abstract":"<jats:p>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 weather model FourCastNet on the 1979\u20132015 ERA5 dataset with all data, or with Category 3\u20135 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\u20132023 Category 5 TCs (gray swans). All versions yield similar accuracy for global weather, but the one trained without Category 3\u20135 TCs cannot accurately forecast Category 5 TCs, indicating that these models cannot extrapolate from weaker storms. The versions trained without Category 3\u20135 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.<\/jats:p>","DOI":"10.1073\/pnas.2420914122","type":"journal-article","created":{"date-parts":[[2025,5,20]],"date-time":"2025-05-20T13:40:02Z","timestamp":1747748402000},"update-policy":"https:\/\/doi.org\/10.1073\/pnas.cm10313","source":"Crossref","is-referenced-by-count":54,"title":["Can AI weather models predict out-of-distribution gray swan tropical cyclones?"],"prefix":"10.1073","volume":"122","author":[{"ORCID":"https:\/\/orcid.org\/0000-0001-5190-4681","authenticated-orcid":false,"given":"Y. Qiang","family":"Sun","sequence":"first","affiliation":[{"name":"Department of the Geophysical Sciences","place":["Chicago"]},{"name":"University of Chicago","place":["Chicago"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Pedram","family":"Hassanzadeh","sequence":"additional","affiliation":[{"name":"Department of the Geophysical Sciences","place":["Chicago"]},{"name":"University of Chicago","place":["Chicago"]},{"name":"Committee on Computational and Applied Mathematics, Division of the Physical Sciences","place":["Chicago"]},{"name":"University of Chicago","place":["Chicago"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Mohsen","family":"Zand","sequence":"additional","affiliation":[{"name":"Research Computing Center, The Office of the Provost","place":["Chicago"]},{"name":"University of Chicago","place":["Chicago"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-2590-1230","authenticated-orcid":false,"given":"Ashesh","family":"Chattopadhyay","sequence":"additional","affiliation":[{"name":"Department of Applied Mathematics","place":["Santa Cruz"]},{"name":"University of California","place":["Santa Cruz"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-8745-1821","authenticated-orcid":false,"given":"Jonathan","family":"Weare","sequence":"additional","affiliation":[{"name":"Courant Institute of Mathematical Sciences","place":["New York"]},{"name":"New York University","place":["New York"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Dorian S.","family":"Abbot","sequence":"additional","affiliation":[{"name":"Department of the Geophysical Sciences","place":["Chicago"]},{"name":"University of Chicago","place":["Chicago"]}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"341","published-online":{"date-parts":[[2025,5,20]]},"reference":[{"key":"e_1_3_4_1_2","unstructured":"J. Pathak FourCastNet: A global data-driven high-resolution weather model using adaptive Fourier neural operators. arXiv [Preprint] (2022). https:\/\/arxiv.org\/abs\/2202.11214 (Accessed 1 November 2022)."},{"key":"e_1_3_4_2_2","doi-asserted-by":"publisher","DOI":"10.1038\/s41586-023-06185-3"},{"key":"e_1_3_4_3_2","doi-asserted-by":"publisher","DOI":"10.1126\/science.adi2336"},{"key":"e_1_3_4_4_2","unstructured":"S. Lang AIFS-ECMWF\u2019s data-driven forecasting system. arXiv [Preprint] (2024). https:\/\/arxiv.org\/abs\/2406.01465 (Accessed 12 August 2024)."},{"key":"e_1_3_4_5_2","doi-asserted-by":"publisher","DOI":"10.1038\/s41612-023-00512-1"},{"key":"e_1_3_4_6_2","unstructured":"T. Nguyen \u201cScaling transformers for skillful and reliable medium-range weather forecasting\u201d in ICLR 2024 Workshop on AI4 Differential Equations in Science (2024)."},{"key":"e_1_3_4_7_2","doi-asserted-by":"publisher","DOI":"10.1029\/2023MS004019"},{"key":"e_1_3_4_8_2","doi-asserted-by":"publisher","DOI":"10.1175\/BAMS-D-23-0162.1"},{"key":"e_1_3_4_9_2","doi-asserted-by":"crossref","unstructured":"T. Kurth \u201cFourCastNet: Accelerating global high-resolution weather forecasting using adaptive Fourier neural operators\u201d in Proceedings of the Platform for Advanced Scientific Computing Conference (2023) pp. 1\u201311.","DOI":"10.1145\/3592979.3593412"},{"key":"e_1_3_4_10_2","doi-asserted-by":"crossref","unstructured":"I. Price Probabilistic weather forecasting with machine learning. Nature 637 84\u201390 (2025). https:\/\/doi.org\/10.1038\/s41586-024-08252-9.","DOI":"10.1038\/s41586-024-08252-9"},{"key":"e_1_3_4_11_2","doi-asserted-by":"publisher","DOI":"10.1126\/sciadv.adk4489"},{"key":"e_1_3_4_12_2","unstructured":"A. Mahesh Huge ensembles. Part I: Design of ensemble weather forecasts using spherical Fourier neural operators. arXiv [Preprint] (2024). https:\/\/arxiv.org\/abs\/2408.03100 (Accessed 15 August 2024)."},{"key":"e_1_3_4_13_2","unstructured":"A. Mahesh Huge ensembles. Part II: Properties of a huge ensemble of hindcasts generated with spherical Fourier neural operators. arXiv [Preprint] (2024). https:\/\/arxiv.org\/abs\/2408.01581 (Accessed 15 August 2024)."},{"key":"e_1_3_4_14_2","unstructured":"O. Watt-Meyer ACE: A fast skillful learned global atmospheric model for climate prediction. arXiv [Preprint] (2023). https:\/\/arxiv.org\/abs\/2310.02074 (Accessed 12 December 2023)."},{"key":"e_1_3_4_15_2","unstructured":"S. R. Cachay B. Henn O. Watt-Meyer C. S. Bretherton R. Yu Probabilistic emulation of a global climate model with spherical dyffusion. arXiv [Preprint] (2024). https:\/\/arxiv.org\/abs\/2406.14798 (Accessed 1 September 2024)."},{"key":"e_1_3_4_16_2","first-page":"e2024JH000136","article-title":"Application of the AI2 climate emulator to E3SMV2\u2019S global atmosphere model, with a focus on precipitation fidelity","volume":"1","author":"Duncan J. P.","year":"2024","unstructured":"J. P. Duncan , Application of the AI2 climate emulator to E3SMV2\u2019S global atmosphere model, with a focus on precipitation fidelity. J. Geophys. Res.: Mach. Learn. Comput. 1, e2024JH000136 (2024).","journal-title":"J. Geophys. Res.: Mach. Learn. Comput."},{"key":"e_1_3_4_17_2","doi-asserted-by":"crossref","unstructured":"H. Guan T. Arcomano A. Chattopadhyay R. Maulik Lucie: A lightweight uncoupled climate emulator with long-term stability and physical consistency for o (1000)-member ensembles. arXiv [Preprint] (2024). https:\/\/arxiv.org\/abs\/2405.16297 (Accessed 1 June 2024).","DOI":"10.5194\/egusphere-egu25-20616"},{"key":"e_1_3_4_18_2","unstructured":"C. Y. Lai Machine learning for climate physics and simulations. arXiv [Preprint] (2024). https:\/\/arxiv.org\/abs\/2404.13227 (Accessed 22 August 2024)."},{"key":"e_1_3_4_19_2","unstructured":"N. Cresswell-Clay Deep learning earth system model for stable and efficient simulation of the current climate. arXiv [Preprint] (2024). https:\/\/arxiv.org\/abs\/2409.16247 (Accessed 1 October 2024)."},{"key":"e_1_3_4_20_2","unstructured":"S. Dheeshjith Transfer learning for emulating ocean climate variability across CO2 forcing. arXiv [Preprint] (2024). https:\/\/arxiv.org\/abs\/2405.18585 (Accessed 5 August 2024)."},{"key":"e_1_3_4_21_2","doi-asserted-by":"publisher","DOI":"10.1038\/s42254-024-00776-3"},{"key":"e_1_3_4_22_2","doi-asserted-by":"publisher","DOI":"10.1038\/s41586-024-07744-y"},{"key":"e_1_3_4_23_2","unstructured":"A. Chattopadhyay P. Hassanzadeh Long-term instabilities of deep learning-based digital twins of the climate system: The cause and a solution. arXiv [Preprint] (2023). https:\/\/arxiv.org\/abs\/2304.07029 (Accessed 7 December 2024)."},{"key":"e_1_3_4_24_2","doi-asserted-by":"publisher","DOI":"10.1017\/CBO9781139177245"},{"key":"e_1_3_4_25_2","doi-asserted-by":"publisher","DOI":"10.1073\/pnas.1712645115"},{"key":"e_1_3_4_26_2","doi-asserted-by":"publisher","DOI":"10.1038\/s41558-024-02095-y"},{"key":"e_1_3_4_27_2","doi-asserted-by":"publisher","DOI":"10.1007\/s13748-016-0094-0"},{"key":"e_1_3_4_28_2","doi-asserted-by":"publisher","DOI":"10.1029\/2019MS001958"},{"key":"e_1_3_4_29_2","doi-asserted-by":"publisher","DOI":"10.1016\/j.earscirev.2020.103324"},{"key":"e_1_3_4_30_2","doi-asserted-by":"publisher","DOI":"10.1175\/BAMS-D-20-0097.1"},{"key":"e_1_3_4_31_2","doi-asserted-by":"publisher","DOI":"10.1103\/PhysRevFluids.8.040501"},{"key":"e_1_3_4_32_2","first-page":"e240033","article-title":"Validating Deep Learning Weather Forecast Models on Recent High-Impact Extreme Events","volume":"4","author":"Pasche O. C.","year":"2025","unstructured":"O. C. Pasche, J. Wider, Z. Zhang, J. Zscheischler, S. Engelke, Validating Deep Learning Weather Forecast Models on Recent High-Impact Extreme Events. Artif. l. earth syst. 4, e240033 (2025).","journal-title":"Artif. l. earth syst."},{"key":"e_1_3_4_33_2","doi-asserted-by":"publisher","DOI":"10.1038\/s41612-024-00638-w"},{"key":"e_1_3_4_34_2","unstructured":"M. DeMaria Evaluation of tropical cyclone track and intensity forecasts from artificial intelligence weather prediction (AIWP) models. arXiv [Preprint] (2024). https:\/\/arxiv.org\/abs\/2409.06735 (Accessed 15 September 2024)."},{"key":"e_1_3_4_35_2","doi-asserted-by":"publisher","DOI":"10.1029\/2023GL107377"},{"key":"e_1_3_4_36_2","doi-asserted-by":"publisher","DOI":"10.1029\/2023GL105747"},{"key":"e_1_3_4_37_2","doi-asserted-by":"publisher","DOI":"10.1038\/nclimate2777"},{"key":"e_1_3_4_38_2","doi-asserted-by":"crossref","unstructured":"G. J. Hakim S. Masanam \u201cDynamical tests of a deep-learning weather prediction model\u201d in Artificial Intelligence for the Earth Systems (2024).","DOI":"10.1175\/AIES-D-23-0090.1"},{"key":"e_1_3_4_39_2","unstructured":"T. Rackow Robustness of AI-based weather forecasts in a changing climate. arXiv [Preprint] (2024). https:\/\/arxiv.org\/abs\/2409.18529 (Accessed 1 October 2024)."},{"key":"e_1_3_4_40_2","doi-asserted-by":"publisher","DOI":"10.1126\/sciadv.adj7250"},{"key":"e_1_3_4_41_2","doi-asserted-by":"publisher","DOI":"10.1175\/1520-0469(1990)047<0265:GBITC>2.0.CO;2"},{"key":"e_1_3_4_42_2","unstructured":"E. Guo FourCastNext: Improving FourCastNet training with limited compute. arXiv [Preprint] (2024). https:\/\/arxiv.org\/abs\/2401.05584 (Accessed 1 April 2024)."},{"key":"e_1_3_4_43_2","unstructured":"T. Nguyen J. Brandstetter A. Kapoor J. K. Gupta A. Grover Climax: A foundation model for weather and climate. arXiv [Preprint] (2023). https:\/\/arxiv.org\/abs\/2301.10343 (Accessed 20 December 2023)."},{"key":"e_1_3_4_44_2","unstructured":"J. Schmude Prithvi WxC: Foundation model for weather and climate. arXiv [Preprint] (2024). https:\/\/arxiv.org\/abs\/2409.13598 (Accessed 25 September 2024)."},{"key":"e_1_3_4_45_2","doi-asserted-by":"crossref","unstructured":"X. Wang Orbit: Oak ridge base foundation model for earth system predictability. arXiv [Preprint] (2024). https:\/\/arxiv.org\/abs\/2404.14712 (Accessed 24 August 2024).","DOI":"10.1109\/SC41406.2024.00007"},{"key":"e_1_3_4_46_2","unstructured":"X. Man C. Zhang J. Feng C. Li J. Shao W-MAE: Pre-trained weather model with masked autoencoder for multi-variable weather forecasting. arXiv [Preprint] (2023). https:\/\/arxiv.org\/abs\/2304.08754 (Accessed 20 December 2023)."},{"key":"e_1_3_4_47_2","unstructured":"C. Lessig AtmoRep: A stochastic model of atmosphere dynamics using large scale representation learning. arXiv [Preprint] (2023). https:\/\/arxiv.org\/abs\/2308.13280 (Accessed 12 September 2023)."},{"key":"e_1_3_4_48_2","unstructured":"A. McNally Data driven weather forecasts trained and initialised directly from observations. arXiv [Preprint] (2024). https:\/\/arxiv.org\/abs\/2407.15586 (Accessed 27 August 2024)."},{"key":"e_1_3_4_49_2","doi-asserted-by":"publisher","DOI":"10.1029\/2022MS003585"},{"key":"e_1_3_4_50_2","unstructured":"L. M. Yang E. P. Gerber Overcoming set imbalance in data driven parameterization: A case study of gravity wave momentum transport. arXiv [Preprint] (2024). https:\/\/arxiv.org\/abs\/2402.18030 (Accessed 2 March 2024)."},{"key":"e_1_3_4_51_2","doi-asserted-by":"publisher","DOI":"10.1038\/s41586-024-07566-y"},{"key":"e_1_3_4_52_2","doi-asserted-by":"publisher","DOI":"10.5194\/gmd-17-2493-2024"},{"key":"e_1_3_4_53_2","doi-asserted-by":"publisher","DOI":"10.5194\/gmd-15-2221-2022"},{"key":"e_1_3_4_54_2","unstructured":"R. Wang R. Walters R. Yu Data augmentation vs. equivariant networks: A theory of generalization on dynamics forecasting. arXiv [Preprint] (2022). https:\/\/arxiv.org\/abs\/2206.09450 (Accessed 1 February 2023)."},{"key":"e_1_3_4_55_2","doi-asserted-by":"publisher","DOI":"10.1063\/1.5081461"},{"key":"e_1_3_4_56_2","doi-asserted-by":"publisher","DOI":"10.1029\/2020GL091197"},{"key":"e_1_3_4_57_2","doi-asserted-by":"publisher","DOI":"10.3847\/1538-4357\/ac2fa8"},{"key":"e_1_3_4_58_2","doi-asserted-by":"publisher","DOI":"10.1029\/2023AV000881"},{"key":"e_1_3_4_59_2","doi-asserted-by":"publisher","DOI":"10.1029\/2024MS004264"},{"key":"e_1_3_4_60_2","doi-asserted-by":"publisher","DOI":"10.3389\/fclim.2022.789641"},{"key":"e_1_3_4_61_2","doi-asserted-by":"crossref","unstructured":"H. Zhang J. Finkel D. S. Abbot E. P. Gerber J. Weare Using explainable AI and transfer learning to understand and predict the maintenance of Atlantic blocking with limited observational data. arXiv [Preprint] (2024). https:\/\/arxiv.org\/abs\/2404.08613 (Accessed 16 April 2024).","DOI":"10.1029\/2024JH000243"},{"key":"e_1_3_4_62_2","doi-asserted-by":"publisher","DOI":"10.3847\/2515-5172\/ad18a6"},{"key":"e_1_3_4_63_2","doi-asserted-by":"publisher","DOI":"10.1038\/s41467-023-40112-4"},{"key":"e_1_3_4_64_2","doi-asserted-by":"publisher","DOI":"10.1038\/s42256-024-00897-5"},{"key":"e_1_3_4_65_2","doi-asserted-by":"publisher","DOI":"10.1002\/qj.3803"},{"key":"e_1_3_4_66_2","unstructured":"Z. Li Fourier neural operator for parametric partial differential equations. arXiv [Preprint] (2020). https:\/\/arxiv.org\/abs\/2010.08895 (Accessed 1 February 2023)."},{"key":"e_1_3_4_67_2","unstructured":"J. Guibas Adaptive Fourier neural operators: Efficient token mixers for transformers. arXiv [Preprint] (2021). https:\/\/arxiv.org\/abs\/2111.13587 (Accessed 12 February 2023)."},{"key":"e_1_3_4_68_2","unstructured":"Y. Q. Sun Gray Swan Tropical Cyclone Forecasting with AI? Github. https:\/\/github.com\/envfluids\/FourCastNet. Deposited 13 October 2024."},{"key":"e_1_3_4_69_2","unstructured":"Y. Q. Sun UChicago Hurricane AI Project - Results 1. Zenodo. https:\/\/zenodo.org\/records\/13835657. Deposited 24 September 2024."},{"key":"e_1_3_4_70_2","unstructured":"Y. Q. Sun UChicago Hurricane AI Project - Model Weights. Zenodo. https:\/\/zenodo.org\/records\/13834149. Deposited 24 September 2024."}],"container-title":["Proceedings of the National Academy of Sciences"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/pnas.org\/doi\/pdf\/10.1073\/pnas.2420914122","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,5,20]],"date-time":"2025-05-20T13:40:39Z","timestamp":1747748439000},"score":1,"resource":{"primary":{"URL":"https:\/\/pnas.org\/doi\/10.1073\/pnas.2420914122"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,5,20]]},"references-count":70,"journal-issue":{"issue":"21","published-print":{"date-parts":[[2025,5,27]]}},"alternative-id":["10.1073\/pnas.2420914122"],"URL":"https:\/\/doi.org\/10.1073\/pnas.2420914122","relation":{},"ISSN":["0027-8424","1091-6490"],"issn-type":[{"value":"0027-8424","type":"print"},{"value":"1091-6490","type":"electronic"}],"subject":[],"published":{"date-parts":[[2025,5,20]]},"assertion":[{"value":"2024-10-18","order":0,"name":"received","label":"Received","group":{"name":"publication_history","label":"Publication History"}},{"value":"2025-04-08","order":2,"name":"accepted","label":"Accepted","group":{"name":"publication_history","label":"Publication History"}},{"value":"2025-05-20","order":3,"name":"published","label":"Published","group":{"name":"publication_history","label":"Publication History"}}],"article-number":"e2420914122"}}