Gaoshan Guo
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Gaoshan Guo has a total of 49 co-authors. Here are the top co-authors by number of shared publications:
- Haiqiang Lan
State Key Laboratory of Lithospheric and Environmental Coevolution, Institute of Geology and Geophysics, Chinese Academy of Sciences, Beijing, China
8 shared publications - Xiaole Zhou
Chinese Academy of Sciences 1 , Institute of Geology and Geophysics, State Key Laboratory of Lithospheric Evolution, Beijing, China and , Beijing, China
3 shared publications - Jean-Xavier Dessa
Sorbonne Universités Observatoire de la Côte d'Azur Université Côte d'Azur CNRS, IRD, Géoazur Valbonne France
2 shared publications Ling Chen
Chinese Academy of Sciences 2 , Institute of Geology and Geophysics, State Key Laboratory of Lithospheric Evolution, Beijing, China; , Beijing, China; and , Guangzhou, China
2 shared publications- Ling Chen
State Key Laboratory of Lithospheric and Environmental Coevolution, Institute of Geology and Geophysics, Chinese Academy of Sciences, Beijing, China
2 shared publications - Qinya Liu
Department of Earth Sciences, University of Toronto, Toronto, Canada
2 shared publications
Latest Publications
- 4 November 2026
- Published by Springer Nature in Nature Geoscience
- p. 1-8
Seismic evidence for double slab tears at the Neo-Tethyan subduction–collision transition
- 4 November 2026
- Published by Springer Nature in Nature Geoscience
- p. 1-8
Seismic evidence for double slab tears at the Neo-Tethyan subduction–collision transition
- Open Access
- 15 October 2026
- Published by American Geophysical Union (AGU) in Journal of Geophysical Research: Solid Earth
- Vol. 131 (10)
Seismic Imaging of the Crust From Long‐Offset Ocean‐Bottom Seismometer Data Using Time‐Domain Extended‐Source Full‐Waveform Inversion: Application to the Eastern Nankai Trough, Japan
High‐resolution crustal‐scale seismic imaging from long‐offset ocean‐bottom seismometer (OBS) data using full‐waveform inversion (FWI) remains challenging because the large number of propagated wavelengths and waveform complexity make the inversion highly sensitive to time‐shift errors (cycle skipping). To mitigate this nonlinearity, we assess the recently developed time‐domain extended‐source (ES‐) FWI with a well‐documented OBS case study from the eastern‐Nankai subduction zone, Japan. ES‐FWI first reconstructs wavefields that fit the recorded data well using extended sources defined as the sum of the experiment and scattering sources. These scattering sources, which depend on the unknown true wavefields and the sought perturbation model, are estimated at each iteration by fitting in the least‐squares sense the associated data residuals, namely the virtually measured scattered data. The model is subsequently updated by minimizing the scattering sources, keeping the reconstructed wavefields fixed to transfer the perturbation model hidden in these sources toward the current iteration background model. This alternating‐direction optimization for wavefield reconstruction and model updating is iterated until convergence, where both data residuals and scattering sources should vanish. We perform ES‐FWI with a multiscale workflow starting from a model obtained by first‐arrival traveltime tomography. We show that ES‐FWI achieves more accurate results than FWI. Moreover, we refine the final model of the extended‐space approach by classical FWI to further decrease residual data misfit once the linear domain has been reached. Finally, several key structures resolved in the final ES‐FWI model, such as the subduction megathrust overlain by the décollement , provide a first validation of the approach.
- 15 October 2026
- Published by American Geophysical Union (AGU) in Journal of Geophysical Research: Solid Earth
- Vol. 131 (10)
Open AccessSeismic Imaging of the Crust From Long‐Offset Ocean‐Bottom Seismometer Data Using Time‐Domain Extended‐Source Full‐Waveform Inversion: Application to the Eastern Nankai Trough, Japan
High‐resolution crustal‐scale seismic imaging from long‐offset ocean‐bottom seismometer (OBS) data using full‐waveform inversion (FWI) remains challenging because the large number of propagated wavelengths and waveform complexity make the inversion highly sensitive to time‐shift errors (cycle skipping). To mitigate this nonlinearity, we assess the recently developed time‐domain extended‐source (ES‐) FWI with a well‐documented OBS case study from the eastern‐Nankai subduction zone, Japan. ES‐FWI first reconstructs wavefields that fit the recorded data well using extended sources defined as the sum of the experiment and scattering sources. These scattering sources, which depend on the unknown true wavefields and the sought perturbation model, are estimated at each iteration by fitting in the least‐squares sense the associated data residuals, namely the virtually measured scattered data. The model is subsequently updated by minimizing the scattering sources, keeping the reconstructed wavefields fixed to transfer the perturbation model hidden in these sources toward the current iteration background model. This alternating‐direction optimization for wavefield reconstruction and model updating is iterated until convergence, where both data residuals and scattering sources should vanish. We perform ES‐FWI with a multiscale workflow starting from a model obtained by first‐arrival traveltime tomography. We show that ES‐FWI achieves more accurate results than FWI. Moreover, we refine the final model of the extended‐space approach by classical FWI to further decrease residual data misfit once the linear domain has been reached. Finally, several key structures resolved in the final ES‐FWI model, such as the subduction megathrust overlain by the décollement , provide a first validation of the approach.
- Open Access
- 3 June 2026
- Published by California Digital Library (CDL) in EarthArXiv
Time domain full waveform inversion with decomposed Gauss-Newton Hessian
Full waveform inversion (FWI), a nonlinear data fitting approach for parameter estimation, is generally implemented with local optimization methods. The convergence of the iterations can be slow when the steepest descent direction provided by the negative gradient of the data misfit function is not preconditioned by the inverse full Newton Hessian or its linear approximation, namely the Gauss-Newton (GN) Hessian. Here, we present a novel generalized (or weighted) form of time-domain FWI allowing for a robust and computationally-efficient implementation of Gauss-Newton FWI. This weighted form of FWI is inspired by the extended-source formulation of FWI. In this framework, the GN Hessian of FWI can be decomposed into a source-side diagonal pseudo-Hessian in the model domain and a receiver-side Hessian in the data domain. The pseudo-Hessian corresponds to the auto-correlation of the so-called virtual sources, which is classically used to precondition the descent direction of FWI by scaling the amplitudes of the gradient. The receiver-side data-domain Hessian represents the Hessian of an underdetermined source problem. This source problem aims at estimating the volume scattering sources that would be generated by the interaction of the true incident wavefields with the unknown perturbation model at the current FWI iteration. The so-called adjoint fields of FWI are the adjoint approximations of these scattering sources. The pseudo-Hessian does not generate computational overhead, while the inverse data-domain Hessian is approximated by a multi-dimensional non-stationary matching filter for each source. Finally, the computational cost of one decomposed GN descent direction is two times that of the steepest descent. Numerical tests validate that the decomposed GN Hessian provides a reliable approximation of the GN Hessian. Alternatively, the weighted gradient by the inverse data-domain Hessian can be used in quasi-Newton methods with the diagonal pseudo-Hessian as preconditioner. Synthetic tests show that the decomposed GN Hessian not only accelerates the convergence of classical FWI but can also reconstruct more accurate velocity models when the initial model is not accurate.
- 3 June 2026
- Published by California Digital Library (CDL) in EarthArXiv
Open AccessTime domain full waveform inversion with decomposed Gauss-Newton Hessian
Full waveform inversion (FWI), a nonlinear data fitting approach for parameter estimation, is generally implemented with local optimization methods. The convergence of the iterations can be slow when the steepest descent direction provided by the negative gradient of the data misfit function is not preconditioned by the inverse full Newton Hessian or its linear approximation, namely the Gauss-Newton (GN) Hessian. Here, we present a novel generalized (or weighted) form of time-domain FWI allowing for a robust and computationally-efficient implementation of Gauss-Newton FWI. This weighted form of FWI is inspired by the extended-source formulation of FWI. In this framework, the GN Hessian of FWI can be decomposed into a source-side diagonal pseudo-Hessian in the model domain and a receiver-side Hessian in the data domain. The pseudo-Hessian corresponds to the auto-correlation of the so-called virtual sources, which is classically used to precondition the descent direction of FWI by scaling the amplitudes of the gradient. The receiver-side data-domain Hessian represents the Hessian of an underdetermined source problem. This source problem aims at estimating the volume scattering sources that would be generated by the interaction of the true incident wavefields with the unknown perturbation model at the current FWI iteration. The so-called adjoint fields of FWI are the adjoint approximations of these scattering sources. The pseudo-Hessian does not generate computational overhead, while the inverse data-domain Hessian is approximated by a multi-dimensional non-stationary matching filter for each source. Finally, the computational cost of one decomposed GN descent direction is two times that of the steepest descent. Numerical tests validate that the decomposed GN Hessian provides a reliable approximation of the GN Hessian. Alternatively, the weighted gradient by the inverse data-domain Hessian can be used in quasi-Newton methods with the diagonal pseudo-Hessian as preconditioner. Synthetic tests show that the decomposed GN Hessian not only accelerates the convergence of classical FWI but can also reconstruct more accurate velocity models when the initial model is not accurate.
- 8 April 2026
- Published by Agence Bibliographique de l'Enseignement Supérieur
Time-domain extended-source full waveform inversion : theory, algorithm and application
Inversion de la forme d'onde complète à source étendue dans le domaine temporel : théorie, algorithme et application: La Full waveform inversion (FWI) est devenue la méthode d'imagerie de référence en exploration géophysique. FWI utilise les formes d'ondes complètes pour imager le sous-sol avec une résolution d'un demi longueur d'onde. Etant donné la dimension de l'espace des données et des modèles, la FWI est implémentée avec des méthodes d'optimisation locale sur un espace de recherche réduit où l'équation d'onde est résolue exactement à chaque itération. Cela requiert des modèles initiaux précis pour que les données simulées prédisent les données enregistrées sans saut de phase. Pour relâcher cette condition, plusieurs variantes de la FWI ont été proposées telles que les approches sur des espaces de recherche étendus. Parmi ces approches, la 'Wavefield Reconstruction Inversion (WRI)' implémente l'équation d'onde comme une contrainte faible pour ajuster les observables avec des champs d'onde calculées avec des sources étendues. Les extensions de sources sont estimées en ajustant au sens des moindres carrés la différence entre les données enregistrées et simulées traitées comme les données diffractées enregistrées. Il en résulte que ces sources volumiques sont calculées par renversement temporel (retro-propagation) des résidus déconvolués par le Hessien dans l'espace des données. Cette approche est nommée 'extended-source' FWI (ES-FWI).Dans cette thèse, je développe un algorithme opérationnel pour la ES-FWI. Le premier problème est le calcul des sources volumiques où la déconvolution des résidus par le Hessien est coûteuse. Des études précédentes approximent ce Hessien avec une matrice diagonale ce qui peut suffire dans des contextes favorables mais sujet à des minimums secondaires dans des milieux complexes. Je propose d'approximer l'inverse du Hessien par des filtres de Wiener/Gabor. Des tests numériques sur le modèle Marmousi II démontrent les améliorations apportées par ces filtres comparativement à l'approximation diagonale. Les champs d'ondes calculés avec l'assimilation des données ont une précision qui diminue loin des points de mesure ce qui peut piéger l'inversion dans des minimums secondaires. Pour améliorer la robustesse de la méthode, j'ai implémenté des opérateurs de pondération dans l'espace des données pour injecter progressivement des donnés plus complexes dans l'inversion et reconstruire le milieu de la surface vers les niveaux profonds. Cette approche de 'layer stripping' est illustrée avec les géomodèles complexes 2004 BP Salt et GO3DOBS.La ES-FWI est une forme généralisée de la FWI où l'inverse du Hessien du problème de source est utilisé comme une matrice de pondération dans l'espace des données. Cela engendre une décomposition du Hessien en un opérateur diagonal dans le domaine des sources et un opérateur par source dans l'espace des données représentant le Hessien du problème de source évoqué ci dessus. Je montre comment ré-utiliser cette décomposition dans la FWI pour développer une approximation du Hessien Gauss-Newton qui puisse être calculée efficacement tout en accélérant la convergence de la FWI. Alternativement, l'approximation proposée peut être utilisée comme préconditioneur pour des algorithmes de quasi-Newton.Finalement, j'étends l'application de la reconstruction des champs d'onde avec assimilation des données au problème de ‘redatuming'. Cette application requiert des champs d'ondes de haute précision si bien que j'implémente la déconvolution des données diffractées avec le solveur itératif MINRES plutôt qu'avec des filtres de Gabor. L'approche consiste simplement à calculer les champs d'onde avec l'assimilation des données et à les échantillonner sur la surface d'acquisition virtuelle. Cette approche est précise lorsqu'on connaît le milieu situé entre les surfaces définies par les acquisitions réelle et virtuelle. Le ‘redatuming' des sources et des capteurs peuvent être couplés. Cette approche est illustrée avec des géomodèles marins et terrestres et avec un jeu de données réels de fond de mer.
- 8 April 2026
- Published by Agence Bibliographique de l'Enseignement Supérieur
Time-domain extended-source full waveform inversion : theory, algorithm and application
Inversion de la forme d'onde complète à source étendue dans le domaine temporel : théorie, algorithme et application: La Full waveform inversion (FWI) est devenue la méthode d'imagerie de référence en exploration géophysique. FWI utilise les formes d'ondes complètes pour imager le sous-sol avec une résolution d'un demi longueur d'onde. Etant donné la dimension de l'espace des données et des modèles, la FWI est implémentée avec des méthodes d'optimisation locale sur un espace de recherche réduit où l'équation d'onde est résolue exactement à chaque itération. Cela requiert des modèles initiaux précis pour que les données simulées prédisent les données enregistrées sans saut de phase. Pour relâcher cette condition, plusieurs variantes de la FWI ont été proposées telles que les approches sur des espaces de recherche étendus. Parmi ces approches, la 'Wavefield Reconstruction Inversion (WRI)' implémente l'équation d'onde comme une contrainte faible pour ajuster les observables avec des champs d'onde calculées avec des sources étendues. Les extensions de sources sont estimées en ajustant au sens des moindres carrés la différence entre les données enregistrées et simulées traitées comme les données diffractées enregistrées. Il en résulte que ces sources volumiques sont calculées par renversement temporel (retro-propagation) des résidus déconvolués par le Hessien dans l'espace des données. Cette approche est nommée 'extended-source' FWI (ES-FWI).Dans cette thèse, je développe un algorithme opérationnel pour la ES-FWI. Le premier problème est le calcul des sources volumiques où la déconvolution des résidus par le Hessien est coûteuse. Des études précédentes approximent ce Hessien avec une matrice diagonale ce qui peut suffire dans des contextes favorables mais sujet à des minimums secondaires dans des milieux complexes. Je propose d'approximer l'inverse du Hessien par des filtres de Wiener/Gabor. Des tests numériques sur le modèle Marmousi II démontrent les améliorations apportées par ces filtres comparativement à l'approximation diagonale. Les champs d'ondes calculés avec l'assimilation des données ont une précision qui diminue loin des points de mesure ce qui peut piéger l'inversion dans des minimums secondaires. Pour améliorer la robustesse de la méthode, j'ai implémenté des opérateurs de pondération dans l'espace des données pour injecter progressivement des donnés plus complexes dans l'inversion et reconstruire le milieu de la surface vers les niveaux profonds. Cette approche de 'layer stripping' est illustrée avec les géomodèles complexes 2004 BP Salt et GO3DOBS.La ES-FWI est une forme généralisée de la FWI où l'inverse du Hessien du problème de source est utilisé comme une matrice de pondération dans l'espace des données. Cela engendre une décomposition du Hessien en un opérateur diagonal dans le domaine des sources et un opérateur par source dans l'espace des données représentant le Hessien du problème de source évoqué ci dessus. Je montre comment ré-utiliser cette décomposition dans la FWI pour développer une approximation du Hessien Gauss-Newton qui puisse être calculée efficacement tout en accélérant la convergence de la FWI. Alternativement, l'approximation proposée peut être utilisée comme préconditioneur pour des algorithmes de quasi-Newton.Finalement, j'étends l'application de la reconstruction des champs d'onde avec assimilation des données au problème de ‘redatuming'. Cette application requiert des champs d'ondes de haute précision si bien que j'implémente la déconvolution des données diffractées avec le solveur itératif MINRES plutôt qu'avec des filtres de Gabor. L'approche consiste simplement à calculer les champs d'onde avec l'assimilation des données et à les échantillonner sur la surface d'acquisition virtuelle. Cette approche est précise lorsqu'on connaît le milieu situé entre les surfaces définies par les acquisitions réelle et virtuelle. Le ‘redatuming' des sources et des capteurs peuvent être couplés. Cette approche est illustrée avec des géomodèles marins et terrestres et avec un jeu de données réels de fond de mer.
- Open Access
- 23 March 2026
- Published by American Geophysical Union (AGU) in Geophysical Research Letters
- Vol. 53 (6)
Teleseismic Radial Anisotropy Reveals a Sill‐Dominated Magma Reservoir Beneath the Valles Caldera, New Mexico
The Valles Caldera (VC), one of the largest Quaternary silicic calderas in North America, formed by explosive rhyolitic eruptions. Seismic studies suggest a crustal magmatic reservoir beneath the caldera with low‐velocity anomalies, but resolving the detailed geometry of localized melt requires constraints from seismic anisotropy. To image P‐wave velocity and radial anisotropy using dense nodal array data, we develop a teleseismic tomography method that integrates an eikonal solver with the adjoint‐state approach. Our results reveal a pronounced low‐velocity anomaly (>20% reduction) extending laterally across the resurgent dome and down to 15–20 km depth, consistent with a crustal magma chamber. We also identify a colocated zone showing a previously unrecognized pattern of strong positive P‐wave radial anisotropy (up to 8%) where horizontally polarized P‐waves travel faster than vertically polarized ones. This anisotropy indicates a laterally extensive magmatic sill complex and provides new constraints on magma distribution and reservoir architecture beneath the VC.
- 23 March 2026
- Published by American Geophysical Union (AGU) in Geophysical Research Letters
- Vol. 53 (6)
Open AccessTeleseismic Radial Anisotropy Reveals a Sill‐Dominated Magma Reservoir Beneath the Valles Caldera, New Mexico
The Valles Caldera (VC), one of the largest Quaternary silicic calderas in North America, formed by explosive rhyolitic eruptions. Seismic studies suggest a crustal magmatic reservoir beneath the caldera with low‐velocity anomalies, but resolving the detailed geometry of localized melt requires constraints from seismic anisotropy. To image P‐wave velocity and radial anisotropy using dense nodal array data, we develop a teleseismic tomography method that integrates an eikonal solver with the adjoint‐state approach. Our results reveal a pronounced low‐velocity anomaly (>20% reduction) extending laterally across the resurgent dome and down to 15–20 km depth, consistent with a crustal magma chamber. We also identify a colocated zone showing a previously unrecognized pattern of strong positive P‐wave radial anisotropy (up to 8%) where horizontally polarized P‐waves travel faster than vertically polarized ones. This anisotropy indicates a laterally extensive magmatic sill complex and provides new constraints on magma distribution and reservoir architecture beneath the VC.
- Open Access
- 4 February 2026
- Published by Wiley
Teleseismic Radial Anisotropy Reveals a Sill-Dominated Magma Reservoir Beneath the Valles Caldera, New Mexico
The Valles Caldera, one of the largest Quaternary silicic calderas in North America, formed by explosive rhyolitic eruptions. Seismic studies suggest a crustal magmatic reservoir beneath the caldera with low-velocity anomalies, but resolving the detailed geometry of localized melt requires constraints from seismic anisotropy. To image P-wave velocity and radial anisotropy using dense nodal array data, we develop a teleseismic tomography method that integrates an eikonal solver with the adjoint-state approach. Our results reveal a pronounced low-velocity anomaly (>20% reduction) extending laterally across the resurgent dome and down to 15–20~km depth, consistent with a crustal magma chamber. We also identify a colocated zone showing a previously unrecognized pattern of strong positive radial anisotropy (up to 8%) where horizontally polarized P-waves travel faster than vertically polarized ones. This anisotropy indicates a laterally extensive magmatic sill complex and provides new constraints on magma distribution and reservoir architecture beneath the Valles Caldera.
- 4 February 2026
- Published by Wiley
Open AccessTeleseismic Radial Anisotropy Reveals a Sill-Dominated Magma Reservoir Beneath the Valles Caldera, New Mexico
The Valles Caldera, one of the largest Quaternary silicic calderas in North America, formed by explosive rhyolitic eruptions. Seismic studies suggest a crustal magmatic reservoir beneath the caldera with low-velocity anomalies, but resolving the detailed geometry of localized melt requires constraints from seismic anisotropy. To image P-wave velocity and radial anisotropy using dense nodal array data, we develop a teleseismic tomography method that integrates an eikonal solver with the adjoint-state approach. Our results reveal a pronounced low-velocity anomaly (>20% reduction) extending laterally across the resurgent dome and down to 15–20~km depth, consistent with a crustal magma chamber. We also identify a colocated zone showing a previously unrecognized pattern of strong positive radial anisotropy (up to 8%) where horizontally polarized P-waves travel faster than vertically polarized ones. This anisotropy indicates a laterally extensive magmatic sill complex and provides new constraints on magma distribution and reservoir architecture beneath the Valles Caldera.
- 30 January 2026
- Published by Seismological Society of America (SSA) in Seismological Research Letters
- Vol. 97 (4) , 2583-2598
Development and Application of Topography-Dependent First-Arrival Double-Difference Adjoint Tomography to Active-Source 3D OBS Data from the Southwest Indian Ridge
First-arrival tomography (FAT) is widely used for crustal-scale seismic imaging based on active-source ocean-bottom seismometer (OBS) data. However, conventional FAT approaches face challenges, including the ill-posedness and the difficulty in accurately accounting for complex bathymetric variations. To overcome these limitations, we develop a novel 3D topography-dependent first-arrival double-difference tomography (3D TDFADT) method. This approach minimizes a joint cost function that combines both absolute and differential first-arrival travel-time measurements with appropriate weighting. Our method uses a boundary-conforming mesh to accurately discretize models with rugged bathymetry. By introducing a coordinate transformation between Cartesian and curvilinear coordinates, we solve a 3D topography-dependent eikonal equation to compute travel times. The inverse problem is tackled using a preconditioned quasi-Newton optimization algorithm, with the gradient and preconditioner of the data misfit efficiently computed by the adjoint-state method. This formulation enables gradient computation without the need for tedious posterior ray tracing or storage of large Fréchet matrices in curvilinear coordinates. In contrast, the preconditioner improves the stability and accelerates the convergence of the inversion by partially compensating for uneven ray coverage of first arrivals in OBS data, but this may inherently involve a trade-off with resolution. We apply TDFADT to 3D first-arrival data from airgun shots recorded by an OBS array deployed at the ultraslow-spreading Southwest Indian Ridge (SWIR). Resolution tests indicate that TDFADT reduces both absolute and differential travel-time residuals more effectively, yielding velocity models with significantly improved spatial resolution. The final SWIR velocity model from TDFADT reveals several strong anomalies, such as a high-velocity anomaly possibly related to oceanic core complexes and pronounced low-velocity anomalies in the lower crust beneath the central volcano, suggesting active magmatism due to partial melt. This case study demonstrates that the proposed method is a robust and accurate tool for crustal imaging using first-arrival data from active-source OBS surveys.
- 30 January 2026
- Published by Seismological Society of America (SSA) in Seismological Research Letters
- Vol. 97 (4) , 2583-2598
Development and Application of Topography-Dependent First-Arrival Double-Difference Adjoint Tomography to Active-Source 3D OBS Data from the Southwest Indian Ridge
First-arrival tomography (FAT) is widely used for crustal-scale seismic imaging based on active-source ocean-bottom seismometer (OBS) data. However, conventional FAT approaches face challenges, including the ill-posedness and the difficulty in accurately accounting for complex bathymetric variations. To overcome these limitations, we develop a novel 3D topography-dependent first-arrival double-difference tomography (3D TDFADT) method. This approach minimizes a joint cost function that combines both absolute and differential first-arrival travel-time measurements with appropriate weighting. Our method uses a boundary-conforming mesh to accurately discretize models with rugged bathymetry. By introducing a coordinate transformation between Cartesian and curvilinear coordinates, we solve a 3D topography-dependent eikonal equation to compute travel times. The inverse problem is tackled using a preconditioned quasi-Newton optimization algorithm, with the gradient and preconditioner of the data misfit efficiently computed by the adjoint-state method. This formulation enables gradient computation without the need for tedious posterior ray tracing or storage of large Fréchet matrices in curvilinear coordinates. In contrast, the preconditioner improves the stability and accelerates the convergence of the inversion by partially compensating for uneven ray coverage of first arrivals in OBS data, but this may inherently involve a trade-off with resolution. We apply TDFADT to 3D first-arrival data from airgun shots recorded by an OBS array deployed at the ultraslow-spreading Southwest Indian Ridge (SWIR). Resolution tests indicate that TDFADT reduces both absolute and differential travel-time residuals more effectively, yielding velocity models with significantly improved spatial resolution. The final SWIR velocity model from TDFADT reveals several strong anomalies, such as a high-velocity anomaly possibly related to oceanic core complexes and pronounced low-velocity anomalies in the lower crust beneath the central volcano, suggesting active magmatism due to partial melt. This case study demonstrates that the proposed method is a robust and accurate tool for crustal imaging using first-arrival data from active-source OBS surveys.
- Open Access
- 27 October 2025
- Published by Wiley
Robust Imaging of the Crust From Long-Offset Ocean-Bottom Seismometer Data Using Time-Domain Extended-Source Full-Waveform Inversion: Application to the Eastern Nankai Trough, Japan
High-resolution seismic imaging at crustal-scale from long-offset ocean-bottom seismometer (OBS) data by full-waveform inversion (FWI) is challenging because the large number of propagated wavelengths and waveform complexity make the inversion highly sensitive to time-shift errors (cycle skipping). To mitigate this nonlinearity, we assess the recently developed time-domain extended-source (ES-)FWI with a well-documented OBS case study from the eastern-Nankai subduction zone, Japan. ES-FWI first computes wavefields that fit the data arbitrarily well using extended sources defined as the sum of experimental and scattering sources. These scattering sources, which depend on the unknown true wavefields and the sought perturbation model, are estimated at each iteration by fitting in the least-squares sense the associated data residuals, namely the virtually measured scattered data. Then, the model is updated by minimizing the scattering sources keeping the reconstructed wavefields fixed to transfer the perturbation model hidden in these sources toward the background model. This alternating-direction optimization for wavefield reconstruction and model updating is iterated until convergence where both data residuals and scattering sources should vanish. We perform ES-FWI with a multiscale workflow starting from a model obtained by first-arrival traveltime tomography. We show that ES-FWI achieves more accurate results than FWI after multiscale steps. Moreover, we refine the final model of the extended-space approach by classical FWI to further decrease residual data misfit once the linear domain has been reached. Finally, several key structures recognized in the final ES-FWI model, such as the subduction megathrust overlain by the decollement, provide a first validation of the approach.
- 27 October 2025
- Published by Wiley
Open AccessRobust Imaging of the Crust From Long-Offset Ocean-Bottom Seismometer Data Using Time-Domain Extended-Source Full-Waveform Inversion: Application to the Eastern Nankai Trough, Japan
High-resolution seismic imaging at crustal-scale from long-offset ocean-bottom seismometer (OBS) data by full-waveform inversion (FWI) is challenging because the large number of propagated wavelengths and waveform complexity make the inversion highly sensitive to time-shift errors (cycle skipping). To mitigate this nonlinearity, we assess the recently developed time-domain extended-source (ES-)FWI with a well-documented OBS case study from the eastern-Nankai subduction zone, Japan. ES-FWI first computes wavefields that fit the data arbitrarily well using extended sources defined as the sum of experimental and scattering sources. These scattering sources, which depend on the unknown true wavefields and the sought perturbation model, are estimated at each iteration by fitting in the least-squares sense the associated data residuals, namely the virtually measured scattered data. Then, the model is updated by minimizing the scattering sources keeping the reconstructed wavefields fixed to transfer the perturbation model hidden in these sources toward the background model. This alternating-direction optimization for wavefield reconstruction and model updating is iterated until convergence where both data residuals and scattering sources should vanish. We perform ES-FWI with a multiscale workflow starting from a model obtained by first-arrival traveltime tomography. We show that ES-FWI achieves more accurate results than FWI after multiscale steps. Moreover, we refine the final model of the extended-space approach by classical FWI to further decrease residual data misfit once the linear domain has been reached. Finally, several key structures recognized in the final ES-FWI model, such as the subduction megathrust overlain by the decollement, provide a first validation of the approach.
- 30 October 2023
- Published by Society of Exploration Geophysicists in Geophysics
- Vol. 89 (2) , R73-R94
Time-domain extended-source full-waveform inversion: Algorithm and practical workflow
Extended-source full-waveform inversion (ES-FWI) first computes wavefields with data-driven source extensions such that the simulated data in inaccurate velocity models match the observed counterpart sufficiently well to prevent cycle skipping. Then, the source extensions are minimized to update the model parameters toward the true medium. This two-step workflow is iterated until data and sources are matched. It has been recently indicated that the source extensions are the least-squares solutions of the scattered data-fitting problem. As a result, the source extensions are computed by propagating backward in time the deconvolved data residuals by the damped data-domain Hessian of the scattered data-fitting problem. Estimating these weighted data residuals is the main computational bottleneck of time-domain ES-FWI. To mitigate this burden, we approximate the inverse data-domain Hessian by mono- and multidimensional matching filters with two simulations per source. We implement time-domain ES-FWI with the alternating-direction method of multipliers and total-variation regularization. Moreover, we apply ES-FWI with a multiscale approach involving frequency continuation and layer stripping, with the latter being implemented with an offset-time-dependent weighting operator. In this framework, we further regularize the inversions while mitigating their computational burden by matching the grid interval to the frequency bandwidth. Finally, the overall workflow combines ES-FWI and classical FWI during the early and late stages of the multiscale approach, respectively. We illustrate that the sensitivity of ES-FWI to the accuracy of the approximated inverse data-domain Hessian depends on the complexity of the targeted model, the data anatomy, and the accuracy of the starting model. In the case of the 2004 BP salt model, we determine that the layer stripping is necessary when the inverse data-domain Hessian is approximated by a 2D Gabor matching filter and the starting model is crude, whereas this feature is not necessary with the Marmousi II model.
- 30 October 2023
- Published by Society of Exploration Geophysicists in Geophysics
- Vol. 89 (2) , R73-R94
Time-domain extended-source full-waveform inversion: Algorithm and practical workflow
Extended-source full-waveform inversion (ES-FWI) first computes wavefields with data-driven source extensions such that the simulated data in inaccurate velocity models match the observed counterpart sufficiently well to prevent cycle skipping. Then, the source extensions are minimized to update the model parameters toward the true medium. This two-step workflow is iterated until data and sources are matched. It has been recently indicated that the source extensions are the least-squares solutions of the scattered data-fitting problem. As a result, the source extensions are computed by propagating backward in time the deconvolved data residuals by the damped data-domain Hessian of the scattered data-fitting problem. Estimating these weighted data residuals is the main computational bottleneck of time-domain ES-FWI. To mitigate this burden, we approximate the inverse data-domain Hessian by mono- and multidimensional matching filters with two simulations per source. We implement time-domain ES-FWI with the alternating-direction method of multipliers and total-variation regularization. Moreover, we apply ES-FWI with a multiscale approach involving frequency continuation and layer stripping, with the latter being implemented with an offset-time-dependent weighting operator. In this framework, we further regularize the inversions while mitigating their computational burden by matching the grid interval to the frequency bandwidth. Finally, the overall workflow combines ES-FWI and classical FWI during the early and late stages of the multiscale approach, respectively. We illustrate that the sensitivity of ES-FWI to the accuracy of the approximated inverse data-domain Hessian depends on the complexity of the targeted model, the data anatomy, and the accuracy of the starting model. In the case of the 2004 BP salt model, we determine that the layer stripping is necessary when the inverse data-domain Hessian is approximated by a 2D Gabor matching filter and the starting model is crude, whereas this feature is not necessary with the Marmousi II model.