Preprint · 2026 · Open access
Missing the Butterfly and Predicting the Past: Features or Bugs of Accurate AI Weather Models?
arXiv (Cornell University) · Published 25 August 2026 · DOI 10.48550/arxiv.2608.25835
In brief
“Missing the Butterfly and Predicting the Past: Features or Bugs of Accurate AI Weather Models?” is a 2026 preprint by Pedram Hassanzadeh and 6 co-authors, published in arXiv (Cornell University). OpenAlex counts 3 citations as of 20 November 2026. It is open access under a CC BY licence: the full text is free to read at arXiv (Cornell University).
- Citations
- 3
- Most cited in
- 2026 (3)
- References
- 101
- Authors
- 7
Abstract
AI weather prediction (AIWP) models rival physics-based models, yet the sources of their unexpected forecast accuracy and the degree of their physical fidelity remain unclear. Here, across a hierarchy spanning observation-based reanalysis, a general circulation model, and the multi-scale Lorenz system, we show that AI models can be trained to skillfully predict the past (backcast), though backcasts are less accurate than forecasts. However, skillful backcasting appears to violate the second law of thermodynamics, and all these forecasting and backcasting models miss the butterfly effect. We trace the surprising forecast accuracy, missing butterfly, and skillful backcasting to a single cause: inevitable coarse-graining of training data, which removes fast, small scales and/or some variables. From the Lorenz system to official Pangu-Weather models, reducing coarse-graining makes AI predictions more physics-like (arrow of time and butterfly-like effects emerge), but forecast accuracy declines. Results offer an explanation for AIWP models' forecast skill: unlike physics-based models, they implicitly learn how fast, small scales affect large scales without inheriting their rapid error growth. Broader implications are that AI models' proliferation calls for revisiting predictability theories and long-term climate emulations; backcasting offers a useful, new lens for such analyses.
Cite this paper
Formatted from the catalogue record. Check names, capitals and page numbers against the paper before you submit.
APA 7
Hassanzadeh, P., Li, W., Sun, Y., Wang, J., Wikner, A., Finkel, J. M., & Weare, J. (2026). Missing the Butterfly and Predicting the Past: Features or Bugs of Accurate AI Weather Models? arXiv (Cornell University). https://doi.org/10.48550/arxiv.2608.25835
MLA 9
Hassanzadeh, Pedram, et al. “Missing the Butterfly and Predicting the Past: Features or Bugs of Accurate AI Weather Models?” arXiv (Cornell University), 2026. https://doi.org/10.48550/arxiv.2608.25835.
Chicago
Hassanzadeh, Pedram, Weidong Li, Y.Qiang Sun, Jiangdi Wang, Alexander Wikner, Justin M Finkel, and Jonathan Weare. 2026. “Missing the Butterfly and Predicting the Past: Features or Bugs of Accurate AI Weather Models?” arXiv (Cornell University). https://doi.org/10.48550/arxiv.2608.25835.
Harvard
Hassanzadeh, P. et al. (2026) ‘Missing the Butterfly and Predicting the Past: Features or Bugs of Accurate AI Weather Models?’, arXiv (Cornell University). doi:10.48550/arxiv.2608.25835.
IEEE
P. Hassanzadeh et al., “Missing the Butterfly and Predicting the Past: Features or Bugs of Accurate AI Weather Models?,” arXiv (Cornell University), 2026, doi: 10.48550/arxiv.2608.25835.
BibTeX
@misc{hassanzadeh2026missing,
title = {{Missing the Butterfly and Predicting the Past: Features or Bugs of Accurate AI Weather Models?}},
author = {Hassanzadeh, Pedram and Li, Weidong and Sun, Y.Qiang and Wang, Jiangdi and Wikner, Alexander and Finkel, Justin M and Weare, Jonathan},
howpublished = {arXiv (Cornell University)},
year = {2026},
doi = {10.48550/arxiv.2608.25835},
url = {https://doi.org/10.48550/arxiv.2608.25835}
}
Download for a reference manager: BibTeX (.bib) · RIS for Zotero, Mendeley and EndNote (.ris)
Topic and keywords
Physical Sciences › Earth and Planetary Sciences › Atmospheric Science › Meteorological Phenomena and Simulations
- backcasting
- butterfly effect
- coarse-graining
- second law of thermodynamics
- arrow of time
- Lorenz system
- Pangu-Weather
- predictability
Work with this paper
Most-cited papers in Earth and Planetary Sciences
- The ERA5 global reanalysisHans Hersbach, Bill Bell, Paul Berrisford and 40 more · Quarterly Journal of the Royal Meteorological Society · 2020 · 35,424 citations
- The NCEP/NCAR 40-Year Reanalysis ProjectEugenia Kalnay, Masao Kanamitsu, Robert E. Kistler and 19 more · Bulletin of the American Meteorological Society · 1996 · 28,486 citations
- Chemical and isotopic systematics of oceanic basalts: implications for mantle composition and processesShen‐Su Sun, W. F. McDonough · Geological Society London Special Publications · 1989 · 26,603 citations
- The ERA‐Interim reanalysis: configuration and performance of the data assimilation systemDick P. Dee, Sakari M. Uppala, A. J. Simmons and 33 more · Quarterly Journal of the Royal Meteorological Society · 2011 · 25,245 citations
- PAST: PALEONTOLOGICAL STATISTICAL SOFTWARE PACKAGE FOR EDUCATION AND DATA ANALYSISØyvind Hammer, David A. T. Harper, Paul D. Ryan · 2001 · 17,871 citations
Questions about this paper
How do I cite this paper in APA style?
Hassanzadeh, P., Li, W., Sun, Y., Wang, J., Wikner, A., Finkel, J. M., & Weare, J. (2026). Missing the Butterfly and Predicting the Past: Features or Bugs of Accurate AI Weather Models? arXiv (Cornell University). https://doi.org/10.48550/arxiv.2608.25835
How many times has this paper been cited?
3 times, according to OpenAlex on 20 November 2026. Google Scholar and Scopus count different sets of sources, so their numbers differ.
Is this paper free to read?
Yes. It is open access at arXiv (Cornell University): sign in and open the PDF from this page.
Who wrote this paper?
Pedram Hassanzadeh, Weidong Li, Y.Qiang Sun, Jiangdi Wang, Alexander Wikner, Justin M Finkel, Jonathan Weare.
What is the DOI of this paper?
10.48550/arxiv.2608.25835, permanently at https://doi.org/10.48550/arxiv.2608.25835. Sign in and open it from this page.
Bibliographic data from OpenAlex (CC0), last checked 20 November 2026. ResearchGuru does not host this paper; once you are signed in, the links open it at the publisher or the repository that does.