Skip to content
DraconiticResearch
Search Sources API
203,234 studies31 sourcesUpdated 18 min ago
Peer-reviewedCC BYNature Machine Intelligence

Regional climate risk assessment from climate models using probabilistic machine learning

  • Zhong Yi Wan1
  • Ignacio Lopez‐Gomez1
  • Robert W. Carver1
  • Tapio Schneider2,1
  • John Anderson1,3
  • Fei Sha1,4
  • Leonardo Zepeda-Núñez1
4 affiliations
  1. 1Google Research, Mountain View, CA, USA
  2. 2California Institute of Technology, Pasadena, CA, USA
  3. 3Now at General Motors Sunnyvale Tech Center, Sunnyvale, CA, USA
  4. 4Now at Meta, Menlo Park, CA, USA
Open at sourcePDFDOICite

Abstract

Effective climate risk assessment is hindered by the resolution gap between coarse global climate models and the fine-scale information needed for regional decision making. Existing downscaling methods struggle to bridge this gap: physics-based methods are too computationally expensive for large ensemble sampling of extreme events, while traditional statistical methods fail to capture multivariate spatiotemporal dependencies crucial to compound risk estimation. Furthermore, current machine learning approaches rely on temporally aligned training pairs, which are unavailable for free-running climate projections. Here we introduce GenFocal, an AI framework that generates statistically accurate, fine-scale weather from coarse climate projections without requiring paired simulated and observed events during training. GenFocal synthesizes complex and long-lived hazards, such as heatwaves and tropical cyclones, even when they are not well represented in the coarse climate projections. It also samples high-impact, rare events more accurately than leading methods. By translating large-scale climate projections into actionable localized information, GenFocal provides a powerful paradigm to improve climate adaptation and resilience strategies.

Related

  1. Climate-stratified assessment of PM2.5 using machine learning: Geographic controls dominate meteorological factors
    Someyo Kamal Utsho, Md Rawjatul Fahim, Ruhaidah Samsudin
    PLOS OnePeer-reviewedCC BY 4.0Full textAtmospheric science & meteorologyClimate sciencePLOS
  2. Advancing Near-Term Water Quality Forecasting with Explainable Machine Learning and Probabilistic Uncertainty Quantification
    Rohit Shukla, Cayelan C. Carey, Adrienne Breef-Pilz +1
    ACS ES&T waterPeer-reviewedCC BYFull textEnvironmental science & pollutionEurope PMC
  3. Machine-learning modelling of regional landslide susceptibility in eThekwini, South Africa
    Cher Petersen, Paidamwoyo Mhangara, Laven Naidoo +1
    Scientific ReportsPeer-reviewedCC BY-NC-ND 4.0Full textEnvironmental science & pollutionCrossref
  4. Butterfly Effect and the Kinetic Energy Cascade in Probabilistic Machine Learning Weather Prediction Models
    Jiakai Chen, Joel Oskarsson, Simon Driscoll +1
    arXivPreprintCC BY 4.0Full textAtmospheric science & meteorologyarXiv
  5. Quantifying nonlinear thresholds of grey-green space morphology on urban carbon metabolism across local climate zones: a machine learning study in Changchun city
    Yuqing Gong, Wenjun Qin
    Scientific ReportsPeer-reviewedCC BY-NC-ND 4.0Full textEnvironmental science & pollutionCrossref

Cite

BibTeXRIS
BibTeX
@article{wan2026regional,
  title = {{Regional climate risk assessment from climate models using probabilistic machine learning}},
  author = {Zhong Yi Wan and Ignacio Lopez‐Gomez and Robert W. Carver and Tapio Schneider and John Anderson and Fei Sha and Leonardo Zepeda-Núñez},
  journal = {Nature Machine Intelligence},
  year = {2026},
  month = sep,
  doi = {10.1038/s42256-026-01308-7},
  url = {https://doi.org/10.1038/s42256-026-01308-7},
  note = {CC BY},
}
RIS
TY  - JOUR
TI  - Regional climate risk assessment from climate models using probabilistic machine learning
AU  - Zhong Yi Wan
AU  - Ignacio Lopez‐Gomez
AU  - Robert W. Carver
AU  - Tapio Schneider
AU  - John Anderson
AU  - Fei Sha
AU  - Leonardo Zepeda-Núñez
PY  - 2026
DA  - 2026/09/28
JO  - Nature Machine Intelligence
DO  - 10.1038/s42256-026-01308-7
UR  - https://doi.org/10.1038/s42256-026-01308-7
AB  - Effective climate risk assessment is hindered by the resolution gap between coarse global climate models and the fine-scale information needed for regional decision making. Existing downscaling methods struggle to bridge this gap: physics-based methods are too computationally expensive for large ensemble sampling of extreme events, while traditional statistical methods fail to capture multivariate spatiotemporal dependencies crucial to compound risk estimation. Furthermore, current machine learning approaches rely on temporally aligned training pairs, which are unavailable for free-running climate projections. Here we introduce GenFocal, an AI framework that generates statistically accurate, fine-scale weather from coarse climate projections without requiring paired simulated and observed events during training. GenFocal synthesizes complex and long-lived hazards, such as heatwaves and tropical cyclones, even when they are not well represented in the coarse climate projections. It also samples high-impact, rare events more accurately than leading methods. By translating large-scale climate projections into actionable localized information, GenFocal provides a powerful paradigm to improve climate adaptation and resilience strategies.
SN  - 2522-5839
ER  -