Technical Report on Adjoint Waveform Tomography of East Asia for Improved Waveform Prediction
Doody, Claire, Simmons, Nathan, Rodgers, Artie · 2026
Read Technical Report on Adjoint Waveform Tomography of East Asia for Improved Waveform Prediction by Doody, Claire, Simmons, Nathan, Rodgers, Artie (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/doe-osti-3021938-technical-report-on-adjoint-waveform-tom.txt.
Introduction
LLNL technical report LLNL-TR-2011986, by Artie Rodgers, Claire Doody, and Nathan Simmons, dated March 2026, presents a 3D seismic image of East Asia built to improve waveform prediction. Adjoint waveform tomography fits full recorded seismograms, not just arrival times, by repeatedly simulating waves and adjusting the Earth model until synthetic and real records match. Better regional models mean better predictions of how shaking propagates, a core monitoring and hazard need.
The resulting model images the subducting Pacific slab as a continuous high-velocity structure down to the 410-kilometer discontinuity. Below that, the slab stagnates and spreads above the 660-kilometer boundary. At the stagnated slab's edge, a narrow low-velocity column rises toward the surface. The authors leave open whether it is deep upwelling, a slab edge effect, or an artifact of limited data coverage. A normalized cross-correlation misfit is planned to sharpen the deep image.
Catalog record
| CREATOR | Doody, Claire, Simmons, Nathan, Rodgers, Artie |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 26 KB |
| LICENSE | Public Domain |
| RIGHTS | PUBLIC DOMAIN |
| SOURCE | DOE OSTI |
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
LLNL-TR-2011986 Technical Report on Adjoint Waveform Tomography of East Asia for Improved Waveform Prediction A Rodgers, C Doody, N Simmons March 2026 This document was prepared as an account of work sponsored by an agency of the United States government. Neither the United States government nor Lawrence Livermore National Security, LLC, nor any of their employees makes any warranty, expressed or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed, or represents that its use would not infringe privately owned rights. Reference herein to any specific commercial product, process, or service by trade name, trademark, manufacturer, or otherwise does not necessarily constitute or imply its endorsement, recommendation, or favoring by the United States government or Lawrence Livermore National Security, LLC. The views and opinions of authors expressed herein do not necessarily state or reflect those of the United States government or Lawrence Livermore National Security, LLC, and shall not be used for advertising or product endorsement purposes. This work performed under the auspices of the U.S. Department of Energy by Lawrence Livermore National Laboratory under Contract DE-AC52-07NA27344. L L NTechnical Report on Adjoint Waveform LTomography of East Asia for Improved - XWaveform Prediction X XClaire Doody, Nathan Simmons, Artie Rodgers X - X X X X X September 30, 2025 Lawrence Livermore National Laboratory is operated by Lawrence Livermore National Security, LLC, for the U.S. Department of Energy, National Nuclear Security Administration under Contract DE- AC52-07NA27344. LLNL-TR-2011986 Abstract We present a preliminary version of the East Asia Tomography (EAT) model, an adjoint waveform tomography model of East and Southeast Asia. We used SPiRaL (Simmons et al., 2021) as our starting model and source parameters for 250 earthquakes from the Global Centroid Moment Tensor catalogue (Ekström et al., 2012). Over 198 iterations, we have iterated the EAT model down to a minimum period of 35 seconds. We plan on continuing our iteration technique down to 30 seconds period before updating our misfit function to use a normalized cross- correlation-based misfit functions (e.g., Tao et al., 2018) to better constrain Earth structure. We hope to iterate the current extent of the model to 25 seconds minimum period before iterating to shorter periods for a smaller subregion of the full model. Introduction The Eurasian continent is a heavily studied region in seismic tomography due to the complex tectonic processes that have shaped the region. East Asia hosts numerous subduction zones along the Ring of Fire, spanning from the Kamchatka Peninsula in northeastern Russia to the Philippine Islands near the Equator. In central Eurasia, the Indian subcontinent collides with southwestern China, creating the Himalaya Mountains and the extrusion of the Tibetan Plateau to the east. Intraplate volcanism in northeastern China has debated origins in the deep mantle or from deep dehydration of the subducted Pacific Slab. These features and more have led to extensive imaging studies of Eurasia. While traditional imaging studies have used body wave tomography (e.g., Fu et al., 2023; Fukao and Obayashi, 2013) and surface wave tomography (e.g., Li et al., 2013; Li et al., 2022; Liu and Zhao, 2016; Yang et al., 2021), recent imaging studies of East Asia have focused on using adjoint waveform tomography (AWT) to produce detailed models of the region (e.g., Liu et al., 2024; Ma et al., 2022; Simute et al., 2016; Tao et al., 2018; Xi et al., 2024). Compared to traditional techniques, AWT uses more of the seismic waveform by windowing data around certain phases (e.g., Xi et al., 2024) or by windowing based on similarity between the observed and synthetic data (e.g., Ma et al., 2022). AWT methods have been shown to improve waveform predictions over traditional models (Rodgers et al., 2022) and produce more defined subsurface features, particularly of subducting slabs (Rodgers et al., 2024). In this preliminary study, we present a long period model of greater East and Southeast Asia. Even at long periods, we see improvement in slab structure compared to SPiRaL, our starting model, and reduce the misfit by 45% over 50 iterations. As we begin including shorter period information, we aim to harness methodological advances of recent studies to improve subsurface structure by considering relative amplitudes in our misfit function and to simulate seismic waves through the ocean layer. Though these choices increase computational expense, we hope that they will improve both waveform predictions and subsurface structure. Data Because of its tectonic complexity, East Asia experiences significant moderate-magnitude seismicity. The Global Centroid Moment Tensor (GCMT) Catalogue (Dziewonski et al., 1981; Ekström et al., 2012), a global moment tensor catalogue for events over Mw 5 includes over 2400 events between Mw5.0-7.0 between 1 January 2010 and 31 March 2021. Because using all available events is infeasible, we spatially binned the available events. We used bins that are 100 km by 100 km laterally and 25 km in depth. Because events are scarce in the deep Earth, we included all events with depths greater than 500 km. Binning the events reduced the number of events to around 600. We simulated data for all binned events and picked windows for all events using a modified version of the FLEXWIN algorithm (Maggi et al., 2009). If an event had less than 20 stations contributing data or if less than 40% of stations contributed windows to an event, they were removed from the dataset due to their limited data or poor data quality. After quality control, we converged on a final dataset of 250 events (Figure 1). Data is downloaded from the EarthScope Data Management Center using the Obspy MassDownloader tool. Around 800 stations from both permanent and temporary networks contributed data to the inversions. Figure 1. Map of the 250 events used in the construction of the EAT model. Moment tensors are taken from the Global Centroid Moment Tensor (GCMT) catalogue and are colored by the depth of the event. Red triangles represent seismic stations used in the inversion of the EAT model. Methods We used Salvus (Afanasiev et al., 2019) to run inversions of the EAT model. Forward simulations are run using the spectral element method (Komatitsch & Tromp, 1999; Komatitsch & Vilotte, 1998). Simulations were run on Lawrence Livermore National Laboratory’s Lassen high-performance computing system on up to twelve NVIDIA Tesla V100 GPUs. After we calculated synthetics, we picked windows using a modified version of the FLEXWIN algorithm (Maggi et al., 2009) proposed in Krischer et al. (2015). Windows were not restricted to given phases; instead, data were windowed based on a variety of constraints on the similarity between the observed and synthetic data (see Doody et al., 2023a for a detailed description). After window picking, we weighted windows based on the event-dependent station weighting scheme from Ruan et al. (2019). Station weights were calculated based on distance from the source and station density in a given region (i.e., clustered stations are downweighted while isolated stations are upweighted). We calculated the misfit between the observed and synthetic data using the time-frequency phase misfit function (Fichtner et al., 2008). The time-frequency phase misfit function first transforms the time domain data into the time-frequency domain (Kristeková et al., 2006, 2009) then calculates the weighted phase difference and envelope