Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

Rapid wavefield forecasting for earthquake early warning via deep sequence to sequence learning.

Dongwei Lyu1,2,3, Rie Nakata1,2,4, Pu Ren2

  • 1International Computer Science Institute, Berkeley, CA, USA.

Nature Communications
|November 23, 2025
PubMed
Summary

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Learning earthquake ground motions via conditional generative modeling.

Nature communications·2026
Same author

LossLens: Diagnostics for Machine Learning Through Loss Landscape Visual Analytics.

IEEE computer graphics and applications·2025
Same author

Upper-plate conduits linked to plate boundary that hosts slow earthquakes.

Nature communications·2023
Same author

The capture of extracellular vesicles endogenously released by xenotransplanted tumours induces an inflammatory reaction in the premetastatic niche.

Journal of extracellular vesicles·2023
Same author

Response.

Chest·2021
Same author

Predicting trends in the quality of state-of-the-art neural networks without access to training or testing data.

Nature communications·2021
This summary is machine-generated.

WaveCastNet, a new deep learning model, accurately forecasts seismic wavefields and ground motion intensity in real time. This efficient model generalizes well to critical earthquake scenarios without needing magnitude or epicenter data.

Area of Science:

  • Geophysics
  • Computational Seismology
  • Artificial Intelligence

Background:

  • Accurate forecasting of seismic wavefields is crucial for earthquake early warning systems.
  • Traditional methods often struggle with complex wave propagation and require error-prone pre-estimation steps.
  • Deep learning models offer potential for improved seismic wavefield prediction.

Purpose of the Study:

  • To introduce WaveCastNet, a novel deep learning model for high-dimensional wavefield forecasting.
  • To evaluate WaveCastNet's performance in predicting seismic wave intensity and timing, especially in critical scenarios.
  • To demonstrate the model's efficiency and generalization capabilities compared to existing methods.

Main Methods:

  • Developed WaveCastNet, integrating a convolutional long expressive memory architecture into a sequence-to-sequence framework.

Related Experiment Videos

  • Utilized simulated seismic data from the San Francisco Bay Area for training and validation.
  • Evaluated the model on real earthquake data, showcasing zero-shot capabilities.

    • Main Results:

    • WaveCastNet successfully predicts the intensity and timing of destructive ground motions in real time.
    • The model demonstrates superior generalization to rare seismic events, such as high-magnitude earthquakes.
    • WaveCastNet achieves faster inference times and requires fewer parameters than transformer models due to weight sharing.

    Conclusions:

    • WaveCastNet offers an efficient and accurate deep learning approach for seismic wavefield forecasting.
    • The model's ability to predict ground motion without magnitude or epicenter estimation, or empirical models, represents a significant advancement.
    • WaveCastNet shows promise for real-time earthquake impact assessment and early warning systems.