Engineering PapersSearch

DOE OSTI · 3375649

Learning earthquake ground motions via conditional generative modeling

Ren, Pu [Lawrence Berkeley National Laboratory (LBNL), Berkeley, CA (United States)] (ORCID:000000026354385X)·Nakata, Rie [Lawrence Berkeley National Laboratory (LBNL), Berkeley, CA (United States); Univ. of Tokyo (Japan); University of California, Berkeley, CA (United States). International Computer Science Inst.] (ORCID:0000000291889030)·Lacour, Maxime [University of California, Berkeley, CA (United States). International Computer Science Inst.] (ORCID:0000000283164435)·Naiman, Ilan [Ben Gurion Univ. of the Negev, Beer Sheva (Israel)]·Nakata, Nori [Lawrence Berkeley National Laboratory (LBNL), Berkeley, CA (United States); University of California, Berkeley, CA (United States). International Computer Science Inst.; Massachusetts Inst. of Technology (MIT), Cambridge, MA (United States)] (ORCID:0000000292959416)·Song, Jialin [University of California, Berkeley, CA (United States). International Computer Science Inst.; Simon Fraser Univ., Burnaby, BC (Canada)] (ORCID:0009000313816754)·Bi, Zhengfa [Lawrence Berkeley National Laboratory (LBNL), Berkeley, CA (United States)]·Malik, Osman Asif [Lawrence Berkeley National Laboratory (LBNL), Berkeley, CA (United States)] (ORCID:000000034477481X)·Morozov, Dmitriy [Lawrence Berkeley National Laboratory (LBNL), Berkeley, CA (United States)]·Azencot, Omri [University of California, Berkeley, CA (United States). International Computer Science Inst.; Ben Gurion Univ. of the Negev, Beer Sheva (Israel)]·Erichson, N. Benjamin [Lawrence Berkeley National Laboratory (LBNL), Berkeley, CA (United States); University of California, Berkeley, CA (United States). International Computer Science Inst.] (ORCID:0000000306673516)·Mahoney, Michael W. [Lawrence Berkeley National Laboratory (LBNL), Berkeley, CA (United States); University of California, Berkeley, CA (United States). International Computer Science Inst.] (ORCID:0000000179204652)

Abstract

Predicting high-fidelity ground motions for future earthquakes is crucial for seismic hazard assessment and infrastructure resilience. Conventional empirical simulations suffer from sparse sensor distribution and geographically localized earthquake locations, while physics-based methods are computationally intensive and require accurate representations of Earth structures and earthquake sources. We propose an artificial intelligence (AI) spectrogram generator, Conditional Generative Modeling for Ground Motion (CGM-GM). CGM-GM leverages earthquake magnitudes and geographic coordinates of earthquakes and sensors as inputs, when postprocessed with phase information, capturing spatially continuous Fourier amplitude spectra (FAS) as well as properties such as P and S arrivals, and waveform durations, without explicit physics constraints. This is achieved through a probabilistic autoencoder that extracts latent distributions in the time-frequency domain and variational sequential models for prior and posterior distributions. We evaluate the performance of CGM-GM using small-magnitude earthquake records from the San Francisco Bay Area, a region with high seismic risks. Here, we report that CGM-GM demonstrates potential for complementing physics-based simulations and non-ergodic empirical ground motion models, as well as shows promise in seismology and beyond.

Explore related subjects

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Ren, Pu [Lawrence Berkeley National Laboratory (LBNL), Berkeley, CA (United States)] (ORCID:000000026354385X), Nakata, Rie [Lawrence Berkeley National Laboratory (LBNL), Berkeley, CA (United States); Univ. of Tokyo (Japan); University of California, Berkeley, CA (United States). International Computer Science Inst.] (ORCID:0000000291889030), Lacour, Maxime [University of California, Berkeley, CA (United States). International Computer Science Inst.] (ORCID:0000000283164435), Naiman, Ilan [Ben Gurion Univ. of the Negev, Beer Sheva (Israel)], Nakata, Nori [Lawrence Berkeley National Laboratory (LBNL), Berkeley, CA (United States); University of California, Berkeley, CA (United States). International Computer Science Inst.; Massachusetts Inst. of Technology (MIT), Cambridge, MA (United States)] (ORCID:0000000292959416), Song, Jialin [University of California, Berkeley, CA (United States). International Computer Science Inst.; Simon Fraser Univ., Burnaby, BC (Canada)] (ORCID:0009000313816754), Bi, Zhengfa [Lawrence Berkeley National Laboratory (LBNL), Berkeley, CA (United States)], Malik, Osman Asif [Lawrence Berkeley National Laboratory (LBNL), Berkeley, CA (United States)] (ORCID:000000034477481X), Morozov, Dmitriy [Lawrence Berkeley National Laboratory (LBNL), Berkeley, CA (United States)], Azencot, Omri [University of California, Berkeley, CA (United States). International Computer Science Inst.; Ben Gurion Univ. of the Negev, Beer Sheva (Israel)], Erichson, N. Benjamin [Lawrence Berkeley National Laboratory (LBNL), Berkeley, CA (United States); University of California, Berkeley, CA (United States). International Computer Science Inst.] (ORCID:0000000306673516), Mahoney, Michael W. [Lawrence Berkeley National Laboratory (LBNL), Berkeley, CA (United States); University of California, Berkeley, CA (United States). International Computer Science Inst.] (ORCID:0000000179204652). 2026-03-16. Learning earthquake ground motions via conditional generative modeling. https://doi.org/10.1038/s41467-026-70719-2

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related reports

Advances in geophysical forensic event monitoring

Forensic analysis of man-made, non-nuclear events (such as industrial accidents, explosion experiments and mine collapses) has become more frequent and detailed owing to advancements in geophysical monitoring. Here, in this Technical Review, we demonstrate how geophysical forensic monitoring using seismic, infrasound and hydroacoustic recordings provides insights on events in the solid earth, atmosphere and underwater. Advanced techniques, including machine-learning-based models, have been developed to detect, identify and investigate these events, providing information on location, subevents, sources and explosive yield. The increase in data availability, application of advanced methods and computation and the growth of multitechnology approaches have increased the accuracy of forensic event analysis and enabled more realistic characterization of uncertainties. For example, the 2020 Beirut explosion in Lebanon demonstrated that various seismic, acoustic and other methods could be used to estimate explosive yield (and yield uncertainties) of about 1 ktonne, providing confidence in the application of these methods to smaller events where data are available. However, forensic investigations remain largely limited to known events with identified sources. Increased access to data, sophisticated analysis methods and high-resolution earth models will improve forensic event analysis further, enabling civil and scientific applications, such as localization in the search for the lost ARA San Juan submarine.

geophysics