Skip to content
DraconiticResearch
Search Sources API
203,948 studies31 sourcesUpdated 16 min ago
Peer-reviewedCC BYScience Advances

Predicting regional gray swans via translocation: AI weather models and Dubai’s unprecedented 2024 rainfall

  • Y. Sun1,2
  • Pedram Hassanzadeh1,3
  • Tiffany Shaw1
  • Hamid A. Pahlavan1,4
  • Adam Marchakitus1
4 affiliations
  1. 1Department of the Geophysical Sciences, University of Chicago, Chicago, IL 60637, USA.
  2. 2State Key Laboratory of Severe Weather Meteorological Science and Technology, Nanjing University, Nanjing 210023, China.
  3. 3Committee on Computational and Applied Mathematics, University of Chicago, Chicago, IL 60637, USA.
  4. 4NorthWest Research Associates, Boulder, CO 80301, USA.
Open at sourceDOICite

Abstract

Artificial intelligence (AI) models have transformed weather forecasting, but their skill for unprecedented weather extremes is unclear. Here, we analyze GraphCast, AIFS, and FuXi forecasts of the unprecedented 2024 Dubai storm, which had twice the training set's highest rainfall in that region. GraphCast and AIFS accurately forecast this event up to 8 days ahead. FuXi forecasts the event but underestimates the rainfall. Fine-tuning and receptive field analyses suggest that these models' success stems from "translocation": learning from comparable/stronger dynamically similar events in other regions during training. Evidence of "extrapolation" (learning from weaker events) is not found. Even events within the global distribution's tail are poorly forecasted, which is not only due to data imbalance (generalization error) but also spectral bias (optimization error). These findings demonstrate the potential of AI models to forecast "regional" gray swans and the opportunity to improve them through understanding the mechanisms behind their successes/limitations.

Related

  1. Comparison of data-driven and numerical weather prediction models for rainfall nowcasting in major cities of Bangladesh
    Sumia Akter, Jayeef Rahman Muaz, Rabiul Awal
    PLOS ClimatePeer-reviewedCC BY 4.0Full textAtmospheric science & meteorologyStatisticsPLOS
  2. Development of a weather radar simulator for the characterisation of radar reflectivity of tropical rainfall
    Temitope Yetunde Ojebisi, Moses Oludare Ajewole, Joseph Sunday Ojo +1
    Science Research Annals - Physical SciencesPeer-reviewedCC BY-NC-NDFull textAtmospheric science & meteorologyDOAJ
  3. LoRA Adaptation Strength in Aurora-WRF: Trade-offs in Regional Heavy-Rainfall Forecasting
    Boyan Liu, Xiaoyuan Zhang
    arXivPreprintarXiv licenceFull textAtmospheric science & meteorologyarXiv
  4. Towards operational data-driven regional weather forecasting: extreme-event skill and near-real-time verification with HRRRCast
    Daniel Abdi, Isidora Jankov, Vanderlei Vargas +9
    arXivPreprintCC BY 4.0Full textAtmospheric science & meteorologyarXiv
  5. HourGlass: A probabilistic data-driven temporal downscaler for global and regional weather forecasting
    Magnus Sikora Ingstad, Mariana Clare, Olav Ersland +16
    arXiv (Cornell University)PreprintCC BYFull text7 citedAtmospheric science & meteorologyOpenAlex

Cite

BibTeXRIS
BibTeX
@article{sun2026predicting,
  title = {{Predicting regional gray swans via translocation: AI weather models and Dubai’s unprecedented 2024 rainfall}},
  author = {Y. Sun and Pedram Hassanzadeh and Tiffany Shaw and Hamid A. Pahlavan and Adam Marchakitus},
  journal = {Science Advances},
  year = {2026},
  month = oct,
  doi = {10.1126/sciadv.ady9406},
  url = {https://www.science.org/doi/10.1126/sciadv.ady9406},
  note = {CC BY},
}
RIS
TY  - JOUR
TI  - Predicting regional gray swans via translocation: AI weather models and Dubai’s unprecedented 2024 rainfall
AU  - Y. Sun
AU  - Pedram Hassanzadeh
AU  - Tiffany Shaw
AU  - Hamid A. Pahlavan
AU  - Adam Marchakitus
PY  - 2026
DA  - 2026/10/07
JO  - Science Advances
DO  - 10.1126/sciadv.ady9406
UR  - https://www.science.org/doi/10.1126/sciadv.ady9406
AB  - Artificial intelligence (AI) models have transformed weather forecasting, but their skill for unprecedented weather extremes is unclear. Here, we analyze GraphCast, AIFS, and FuXi forecasts of the unprecedented 2024 Dubai storm, which had twice the training set's highest rainfall in that region. GraphCast and AIFS accurately forecast this event up to 8 days ahead. FuXi forecasts the event but underestimates the rainfall. Fine-tuning and receptive field analyses suggest that these models' success stems from "translocation": learning from comparable/stronger dynamically similar events in other regions during training. Evidence of "extrapolation" (learning from weaker events) is not found. Even events within the global distribution's tail are poorly forecasted, which is not only due to data imbalance (generalization error) but also spectral bias (optimization error). These findings demonstrate the potential of AI models to forecast "regional" gray swans and the opportunity to improve them through understanding the mechanisms behind their successes/limitations.
SN  - 2375-2548
ER  -