{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,17]],"date-time":"2026-10-17T03:21:07Z","timestamp":1792207267425,"version":"4.4.0"},"reference-count":52,"publisher":"American Association for the Advancement of Science (AAAS)","issue":"18","funder":[{"DOI":"10.13039\/501100001711","name":"Swiss National Science Foundation","doi-asserted-by":"publisher","award":["186858"],"award-info":[{"award-number":["186858"]}],"id":[{"id":"10.13039\/501100001711","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100001711","name":"Swiss National Science Foundation","doi-asserted-by":"publisher","award":["186858"],"award-info":[{"award-number":["186858"]}],"id":[{"id":"10.13039\/501100001711","id-type":"DOI","asserted-by":"publisher"}]},{"name":"European Union\u2019s Horizon 2020 research and innovation programme","award":["101003469"],"award-info":[{"award-number":["101003469"]}]},{"name":"European Union\u2019s Horizon 2020 research and innovation programme","award":["101003469"],"award-info":[{"award-number":["101003469"]}]}],"content-domain":{"domain":["www.science.org"],"crossmark-restriction":true},"short-container-title":["Sci. Adv."],"published-print":{"date-parts":[[2026,5,1]]},"abstract":"<jats:p>Artificial intelligence (AI)\u2013based models are revolutionizing weather forecasting and have surpassed leading numerical weather prediction systems on various benchmark tasks. However, their ability to extrapolate and reliably forecast unprecedented extreme events remains unclear. Here, we show that for record-breaking weather extremes, the physics-based numerical model High RESolution forecast (HRES) from the European Centre for Medium-Range Weather Forecasts still consistently outperforms state-of-the-art AI models GraphCast, GraphCast operational, Pangu-Weather, Pangu-Weather operational, and Fuxi. We demonstrate that forecast errors in AI models are consistently larger for record-breaking heat, cold, and wind than in HRES across nearly all lead times. We further find that the examined AI models tend to underestimate both the frequency and intensity of record-breaking events, and they underpredict hot records and overestimate cold records with growing errors for larger record exceedance. Our findings underscore the current limitations of AI weather models in extrapolating beyond their training domain and in forecasting the potentially most impactful record-breaking weather events that are particularly frequent in a rapidly warming climate. Further rigorous verification and model development is needed before these models can be solely relied upon for high-stakes applications such as early warning systems and disaster management.<\/jats:p>","DOI":"10.1126\/sciadv.aec1433","type":"journal-article","created":{"date-parts":[[2026,4,29]],"date-time":"2026-04-29T17:58:27Z","timestamp":1777485507000},"update-policy":"https:\/\/doi.org\/10.34133\/aaas_crossmark","source":"Crossref","is-referenced-by-count":12,"title":["Physics-based models outperform AI weather forecasts of record-breaking extremes"],"prefix":"10.1126","volume":"12","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-2070-1993","authenticated-orcid":true,"given":"Zhongwei","family":"Zhang","sequence":"first","affiliation":[{"name":"Institute of Statistics, Karlsruhe Institute of Technology, Karlsruhe, Germany."},{"name":"Research Institute for Statistics and Information Science, Geneva School of Economics and Management, University of Geneva, Geneva, Switzerland."}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-1931-6737","authenticated-orcid":true,"given":"Erich","family":"Fischer","sequence":"additional","affiliation":[{"name":"Institute for Atmospheric and Climate Science, Department of Environmental Systems Science, ETH Zurich, Zurich, Switzerland."}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-6045-1629","authenticated-orcid":true,"given":"Jakob","family":"Zscheischler","sequence":"additional","affiliation":[{"name":"Department of Compound Environmental Risks, Helmholtz Centre for Environmental Research\u2014UFZ, Leipzig, Germany."},{"name":"Department of Hydro Sciences, TUD Dresden University of Technology, Dresden, Germany."},{"name":"Center for Scalable Data Analytics and Artificial Intelligence (ScaDS.AI), Dresden\/Leipzig, Germany."}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-6356-918X","authenticated-orcid":true,"given":"Sebastian","family":"Engelke","sequence":"additional","affiliation":[{"name":"Research Institute for Statistics and Information Science, Geneva School of Economics and Management, University of Geneva, Geneva, Switzerland."}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"221","reference":[{"key":"e_1_3_2_2_2","doi-asserted-by":"publisher","DOI":"10.1038\/s41467-023-36289-3"},{"key":"e_1_3_2_3_2","doi-asserted-by":"publisher","DOI":"10.1126\/science.1201224"},{"key":"e_1_3_2_4_2","doi-asserted-by":"publisher","DOI":"10.1080\/10643380802238137"},{"key":"e_1_3_2_5_2","doi-asserted-by":"publisher","DOI":"10.1256\/003590002321042036"},{"key":"e_1_3_2_6_2","doi-asserted-by":"publisher","DOI":"10.5194\/nhess-9-405-2009"},{"key":"e_1_3_2_7_2","doi-asserted-by":"publisher","DOI":"10.1038\/s41467-025-57450-0"},{"key":"e_1_3_2_8_2","doi-asserted-by":"publisher","DOI":"10.1038\/nature14956"},{"key":"e_1_3_2_9_2","doi-asserted-by":"publisher","DOI":"10.1038\/s41586-023-06185-3"},{"key":"e_1_3_2_10_2","doi-asserted-by":"publisher","DOI":"10.1126\/science.adi2336"},{"key":"e_1_3_2_11_2","unstructured":"S. Lang M. Alexe M. Chantry J. Dramsch F. Pinault B. Raoult M. C. A. Clare C. Lessig M. Maier-Gerber L. Magnusson Z. B. Bouall\u00e8gue A. P. Nemesio P. D. Dueben A. Brown F. Pappenberger F. Rabier AIFS\u2014ECMWF\u2019s data-driven forecasting system. arXiv:2406.01465 [physics.ao-ph] (2024)."},{"key":"e_1_3_2_12_2","doi-asserted-by":"publisher","DOI":"10.1038\/s41586-025-09005-y"},{"key":"e_1_3_2_13_2","doi-asserted-by":"publisher","DOI":"10.1098\/rsta.2020.0097"},{"key":"e_1_3_2_14_2","doi-asserted-by":"publisher","DOI":"10.1088\/1748-9326\/ac9d4e"},{"key":"e_1_3_2_15_2","doi-asserted-by":"publisher","DOI":"10.1175\/BAMS-D-23-0162.1"},{"key":"e_1_3_2_16_2","doi-asserted-by":"publisher","DOI":"10.5194\/gmd-17-7915-2024"},{"key":"e_1_3_2_17_2","doi-asserted-by":"publisher","DOI":"10.1029\/2025GL117990"},{"key":"e_1_3_2_18_2","doi-asserted-by":"publisher","DOI":"10.1073\/pnas.2420914122"},{"key":"e_1_3_2_19_2","doi-asserted-by":"publisher","DOI":"10.1038\/s41558-021-01092-9"},{"key":"e_1_3_2_20_2","doi-asserted-by":"publisher","DOI":"10.1038\/s41467-023-40112-4"},{"key":"e_1_3_2_21_2","doi-asserted-by":"publisher","DOI":"10.1038\/s41612-025-01090-0"},{"key":"e_1_3_2_22_2","doi-asserted-by":"publisher","DOI":"10.1038\/s41612-025-01198-3"},{"key":"e_1_3_2_23_2","unstructured":"J. Pathak S. Subramanian P. Harrington S. Raja A. Chattopadhyay M. Mardani T. Kurth D. Hall Z. Li K. Azizzadenesheli P. Hassanzadeh K. Kashinath A. Anandkumar FourCastNet: A global data-driven high-resolution weather model using adaptive Fourier neural operators. arXiv:2202.11214 [physics.ao-ph] (2022)."},{"key":"e_1_3_2_24_2","first-page":"240085","article-title":"An operations-based evaluation of tropical cyclone track and intensity forecasts from artificial intelligence weather prediction models","volume":"4","author":"DeMaria M.","year":"2025","unstructured":"M. DeMaria, J. L. Franklin, G. Chirokova, J. Radford, R. DeMaria, K. D. Musgrave, I. Ebert-Uphoff, An operations-based evaluation of tropical cyclone track and intensity forecasts from artificial intelligence weather prediction models. Artif. Intell. Earth Syst. 4, 240085 (2025).","journal-title":"Artif. Intell. Earth Syst."},{"key":"e_1_3_2_25_2","doi-asserted-by":"publisher","DOI":"10.1038\/s41612-024-00638-w"},{"key":"e_1_3_2_26_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. Intell. Earth Syst. 4, e240033 (2025).","journal-title":"Artif. Intell. Earth Syst."},{"key":"e_1_3_2_27_2","unstructured":"Y. Q. Sun P. Hassanzadeh T. Shaw H. A. Pahlavan Predicting beyond training data via extrapolation versus translocation: AI weather models and Dubai\u2019s unprecedented 2024 rainfall. arXiv:2505.10241 [physics.ao-ph] (2025)."},{"key":"e_1_3_2_28_2","doi-asserted-by":"publisher","DOI":"10.1002\/qj.3803"},{"key":"e_1_3_2_29_2","doi-asserted-by":"publisher","DOI":"10.1002\/joc.6850"},{"key":"e_1_3_2_30_2","doi-asserted-by":"publisher","DOI":"10.1126\/sciadv.adi0775"},{"key":"e_1_3_2_31_2","doi-asserted-by":"publisher","DOI":"10.1038\/s41612-023-00512-1"},{"key":"e_1_3_2_32_2","doi-asserted-by":"publisher","DOI":"10.1029\/2023MS004019"},{"key":"e_1_3_2_33_2","doi-asserted-by":"publisher","DOI":"10.1038\/s43017-025-00681-y"},{"key":"e_1_3_2_34_2","doi-asserted-by":"publisher","DOI":"10.1214\/16-STS588"},{"key":"e_1_3_2_35_2","doi-asserted-by":"publisher","DOI":"10.1214\/13-AOAS709"},{"key":"e_1_3_2_36_2","doi-asserted-by":"publisher","DOI":"10.1002\/qj.4351"},{"key":"e_1_3_2_37_2","doi-asserted-by":"publisher","DOI":"10.1038\/s41586-024-08252-9"},{"key":"e_1_3_2_38_2","first-page":"1","article-title":"Domain-adversarial training of neural networks","volume":"17","author":"Ganin Y.","year":"2016","unstructured":"Y. Ganin, E. Ustinova, H. Ajakan, P. Germain, H. Larochelle, F. Laviolette, M. Marchand, V. Lempitsky, Domain-adversarial training of neural networks. J. Mach. Learn. Res. 17, 1\u201335 (2016).","journal-title":"J. Mach. Learn. Res."},{"key":"e_1_3_2_39_2","doi-asserted-by":"publisher","DOI":"10.1038\/s41467-024-50712-3"},{"key":"e_1_3_2_40_2","doi-asserted-by":"publisher","DOI":"10.1038\/s42256-023-00729-y"},{"key":"e_1_3_2_41_2","doi-asserted-by":"publisher","DOI":"10.1029\/2023GL105747"},{"key":"e_1_3_2_42_2","doi-asserted-by":"publisher","DOI":"10.1029\/2023GL107377"},{"key":"e_1_3_2_43_2","doi-asserted-by":"publisher","DOI":"10.1038\/s44387-026-00073-7"},{"key":"e_1_3_2_44_2","doi-asserted-by":"crossref","unstructured":"F. Alet I. Price A. El-Kadi D. Masters S. Markou T. R. Andersson J. Stott R. Lam M. Willson A. Sanchez-Gonzalez P. Battaglia Skillful joint probabilistic weather forecasting from marginals. arXiv:2506.10772 [cs.LG] (2025).","DOI":"10.1038\/s41586-024-08252-9"},{"key":"e_1_3_2_45_2","doi-asserted-by":"publisher","DOI":"10.1186\/s40537-019-0197-0"},{"key":"e_1_3_2_46_2","doi-asserted-by":"publisher","DOI":"10.1038\/s41586-025-08680-1"},{"key":"e_1_3_2_47_2","doi-asserted-by":"publisher","DOI":"10.1016\/j.jcp.2018.10.045"},{"key":"e_1_3_2_48_2","doi-asserted-by":"publisher","DOI":"10.1093\/jrsssb\/qkae108"},{"key":"e_1_3_2_49_2","unstructured":"G. Buritic\u00e1 S. Engelke Progression: An extrapolation principle for regression. arXiv:2410.23246 [stat.ME] (2024)."},{"key":"e_1_3_2_50_2","doi-asserted-by":"publisher","DOI":"10.1017\/eds.2022.4"},{"key":"e_1_3_2_51_2","unstructured":"R. Owens T. Hewson ECMWF Forecast User Guide (ECMWF 2018)."},{"key":"e_1_3_2_52_2","unstructured":"R. Balestriero J. Pesenti Y. LeCun Learning in high dimension always amounts to extrapolation. arXiv:2110.09485 [cs.LG] (2021)."},{"key":"e_1_3_2_53_2","doi-asserted-by":"crossref","unstructured":"A. Agresti An Introduction to Categorical Data Analysis Wiley Series in Probability and Mathematical Statistics (Wiley-Interscience ed. 2 2007).","DOI":"10.1002\/0470114754"}],"container-title":["Science Advances"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/www.science.org\/doi\/pdf\/10.1126\/sciadv.aec1433","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,4,29]],"date-time":"2026-04-29T17:59:02Z","timestamp":1777485542000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.science.org\/doi\/10.1126\/sciadv.aec1433"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,5,1]]},"references-count":52,"journal-issue":{"issue":"18","published-print":{"date-parts":[[2026,5,1]]}},"alternative-id":["10.1126\/sciadv.aec1433"],"URL":"https:\/\/doi.org\/10.1126\/sciadv.aec1433","relation":{},"ISSN":["2375-2548"],"issn-type":[{"value":"2375-2548","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,5,1]]},"assertion":[{"value":"2025-09-08","order":0,"name":"received","label":"Received","group":{"name":"publication_history","label":"Publication History"}},{"value":"2026-03-25","order":2,"name":"accepted","label":"Accepted","group":{"name":"publication_history","label":"Publication History"}},{"value":"2026-04-29","order":3,"name":"published","label":"Published","group":{"name":"publication_history","label":"Publication History"}}],"article-number":"eaec1433"}}