Search NASA⌕ Search

DOE OSTI · 3005980

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

Abstract

We propose a deep learning model, WaveCastNet, to forecast high-dimensional wavefields. WaveCastNet integrates a convolutional long expressive memory architecture into a sequence-to-sequence forecasting framework, enabling it to model long-term dependencies and multiscale patterns in both space and time. By sharing weights across spatial and temporal dimensions, WaveCastNet requires significantly fewer parameters than more resource-intensive models such as transformers, resulting in faster inference times. Crucially, WaveCastNet also generalizes better than transformers to rare and critical seismic scenarios, such as high-magnitude earthquakes. Here, we show the ability of the model to predict the intensity and timing of destructive ground motions in real time, using simulated data from the San Francisco Bay Area. Furthermore, we demonstrate its zero-shot capabilities by evaluating WaveCastNet on real earthquake data. Our approach does not require estimating earthquake magnitudes and epicenters, steps that are prone to error in conventional methods, nor does it rely on empirical ground-motion models, which often fail to capture strongly heterogeneous wave propagation effects.

Explore related subjects

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Lyu, Dongwei, Nakata, Rie, Ren, Pu, Mahoney, Michael W, Pitarka, Arben, Nakata, Nori, Erichson, N Benjamin. 2025-01-01. Rapid wavefield forecasting for earthquake early warning via deep sequence to sequence learning. https://doi.org/10.1038/s41467-025-65435-2

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

KEEP EXPLORING

Related reports

Investigating Buried Ice at Askja Volcano, Northern Iceland using Ground Penetrating Radar: A Planetary Analog Perspective

Shallow subsurface ice deposits on Mars and the Moon have been identified as potential targets for in situ resource utilization by future crewed missions to the surfaces of these bodies. On Mars, such deposits have been identified through a variety of orbital measurements at the poles, mid-latitudes, and even preserved by volcanic ash near the Tharsis Montes. At some locations, massive ice deposits have been documented to be 100 m in thickness at depths as shallow as 1 m. On the Moon, the presence of thick, massive ice has not been confirmed; however, orbital observations indicate that the upper 1-2 m of regolith may contain a few weight percent in thermally favorable locations. Eruptions of the Askja Volcano in March, 1875 and November, 1961 deposited abundant pyroclasts, which blanketed and insulated fresh snowfall that later densified into massive ice and is preserved today. The pyroclasts consist of a buff colored pumice from the eruption in 1875 and basaltic lapilli and ash from the eruption in 1961. The largely unvegetated, unconsolidated nature of the pyroclasts and their stratigraphic relationships with shallow subsurface ice make them potentially analogous to some ice deposits within regolith and pyroclasts at the Moon and Mars. Ground penetrating radar(GPR) can be used by future human or robotic missions to identify water ice similarly preserved in the shallow subsurface of these bodies. Our team conducted over 66 GPR surveys inside and surrounding the Askja caldera in August 2019 at 200 and 400 MHz to map both the volcanic deposits and the subsurface ice deposits. We also used a hammer drill augur to take boreholes in order to confirm subsurface stratigraphy and the presence of ice down to 1-1.5 m and aerial surveys using a Mavic 2 Pro quadcopter at each site for additional context. We observed shallow ice deposits within pore spaces of the pyroclasts at depths of ~15-30 cm and pure ice deposits at varying depths (0.6-1 m) as thick as 2-3 m. We also plan to use the data to characterize the frequency-dependent attenuation of the radar signal as it travels through volcanoclastic material and ice in the subsurface. Our investigation of Askja as a planetary analog will provide insight into analytical methods that can be used to investigate subsurface water ice from surface operations at other terrestrial bodies.

Geophysics↗