Geophysical Journal International

SummaryAmbient noise based monitoring of subsurface velocity changes is possible without the explicit retrieval of Green’s functions by correlation. Velocity variations can directly be observed from the fluctuations in the spectrograms of ambient noise time series or their cross-spectra. This approach is more resource efficient than the conventional Green’s function based monitoring and ideally suited for edge processing and the analysis of large volumes of data for example from fibre-optic records. The spectral fluctuations result from wave propagation in the heterogeneous subsurface and interference between different scattering paths. The fluctuations are directly related to the occurrence of coda waves in the Green’s function. In contrast to coda wave interferometry in the time domain, monitoring the evolution of spectral fluctuations allows for measurements of velocity changes that are unbiased by changes in the source spectrum. Recognizing that the imprint of the subsurface heterogeneity is directly encoded in the spectral fluctuations opens a new perspective on the investigation of subsurface scattering and attenuation.

SummaryWe develop a discrete event modeling framework that captures the progression of geophysical systems toward catastrophic failure through sequences of distinct damage events. By representing geophysical system evolution as a succession of temporally accelerating and amplitude-varying events, the framework reveals how finite-time singularities, both logarithmic and power law types, naturally emerge from the interplay between shrinking interevent intervals and growing event magnitudes. This event-based perspective, which can be viewed as a discrete representation of progressive damage accumulation in heterogeneous geomaterials, provides an intuitive physical understanding of rupture processes, highlighting how precursory signals such as accelerating strain rate, event frequency, and energy release can be traced back to simple underlying mechanisms. A mean-field damage-based formulation further links the observed power law exponents to the evolving stiffness of the geophysical system under constant or time-varying stress. Incorporating stochastic fluctuations, the model captures the inherent randomness of natural systems leading to the emergence of stochastic finite-time singular behavior. Together, these results establish a simple yet powerful framework for interpreting the dynamics of catastrophic events, providing a common event-based perspective on observations from landslides, glacier breakoffs, volcanic eruptions, and other related processes, and strengthening the physical foundations of early warning and hazard forecasting.

SummaryThe Accurately Controlled Routinely Operated Signal System (ACROSS) is an artificial seismic source that generates highly repeatable and stable seismic waves for high-resolution temporal monitoring. While direct P- and S-waves from ACROSS have been widely used to monitor earthquakes, volcanic activity, and environmental changes, the scattered waves that arrive later, known as coda waves, have received limited attention. In this study, we used coda waves generated by ACROSS to monitor subsurface seismic velocity changes (dv/v) over 10 months. The ACROSS signals were recorded by a seismic array of 14 seismometers deployed approximately 3 km from the source in Morimachi, central Japan. By deconvolving the records with a known source function, we obtained transfer functions corresponding to band-limited Green’s functions. Coda wave interferometry applied to the coda portions of these transfer functions detected temporal variations in dv/v, exhibiting both seasonal long-term variations and rainfall-induced short-term variations. Comparative analyses using direct P- and S-wave travel-time changes from the same ACROSS data and ambient-noise interferometry show that, under the present observational conditions, these conventional methods primarily detect long-term variations; however, they do not resolve the rapid, transient short-term variations. In contrast, the stable ACROSS source combined with the broader sampling of coda waves provides superior sensitivity to environmental changes across both timescales over kilometer-scale distances, highlighting its effectiveness for high-temporal-resolution monitoring. To investigate the mechanisms responsible for the observed dual-timescale variations, we used a poroelastic model with precipitation input and considered thermoelastic effects and groundwater-induced changes as possible contributors to the long-term variations.

SummaryThe Alpine-Himalayan orogen preserves geological remnants of subducted lithosphere from the Paleotethys and Neotethys oceans and intervening microcontinents. This orogenic belt displays distinct segments separated by discontinuities aligned with paleo-transform faults, reflecting laterally varying ocean opening and closure histories. We investigate how upper and lower mantle slab remnants imaged by seismic tomography may correlate with Paleo- and Neotethyan subduction zones, focusing on the Anatolian, Aegean, and Iranian segments. Using plate tectonic reconstructions in a mantle reference frame, we predict slab subduction timing, amount, and location and make a semi-quantitative dimensional comparison with seismic tomographic images beneath the Eastern Mediterranean and Middle East. We identify three major Neotethyan slabs: the Pontides and Egypt slabs detached in the Late Cretaceous and now occupy the upper lower mantle. The Cyprus slab remains mainly in the upper mantle in which it overturns. To account for subducted lithosphere that reconstructions predict, we interpret that the Cyprus slab lies overturned in the lower mantle down to ∼1000 km. We interpret a large lower mantle anomaly volume between 2200-1500 km (the Herodotus anomaly) as representing Paleotethyan lithosphere that subducted between ∼240-180 Ma. Our reconstruction-tomography comparison suggests that current slab positions likely reflect past detachment locations, while geometries indicate paleo-trench absolute motions, including a Late Cretaceous same-dip, double subduction configuration. Slabs associated with Aegean, Anatolian, Iranian, and Tibetan segments define stable mantle provinces with boundaries aligned with transform-related orogenic segmentation, implying minimal paleo-longitudinal mantle flow since the Early Mesozoic. Our findings indicate upper and lower mantle structure primarily results from near-vertical slab sinking after detachment since the Triassic, without evidence for deflection by lateral components of mantle convection.

SummaryAs the demand for European domestic mineral resources increases, exploration is increasingly focused on deeper and covered targets, often located in populated regions. In such environments, the application of electromagnetic (EM) techniques is severely challenged by anthropogenic infrastructure. Metal-bearing structures, including power lines, pipelines, railway tracks, and mine shafts, can strongly distort the EM fields, producing significant artefacts in both the measured data and the resulting inversion models, thereby biasing or hindering geological interpretation. These effects currently limit EM investigations in inhabited areas, despite their potential for non-invasive and efficient exploration of challenging subsurface targets. Focusing on semi-airborne electromagnetic (sAEM) data, we develop a data-driven approach to address infrastructure-related effects. By inverting affected data for the distribution of infrastructure currents, we account for the full coupling between the EM transmitter, the conducting Earth, and the metal infrastructure. In sAEM field experiments, involving a grounded dipole transmitter and an airborne receiver system, we study the distortion by self-built infrastructure and investigate both measured currents in infrastructure as well as its EM coupling. We validate the functionality of our current inversion approach by reproducing the measured currents in a simplified 1D scenario. With a synthetic 3D study, mimicking a sAEM campaign in the presence of infrastructure, we find that the current inversion approach is capable of separating and correcting for infrastructure effects even if the true resistivity distribution of the subsurface is unknown. In particular, other conductive subsurface structures can be resolved well, even directly below infrastructure. On a small sAEM field data example, affected by the impact of a metal-built conveyor belt and pipeline, we demonstrate the applicability of this framework for real-world scenarios.

SummaryPermeability is a critical parameter for reservoir characterization and hydrocarbon development, yet its accurate prediction remains a challenge. Pore structure, as the intrinsic factor governing both the elastic and hydraulic transport properties of rocks, serves as a bridge between these properties and facilitates permeability prediction from well logs and seismic data. To accurately describe the variation in the physical properties of tight sandstone reservoirs with pressure, our study aims to construct a physical model that relates rock elastic properties with permeability. We developed a dual-porosity rock physics model by coupling David & Zimmerman’s pore-structure inversion method with Dienes’s percolation theory. Our model divides the pore space into pressure-insensitive stiff pores and pressure-sensitive compliant microcracks. By inverting the microcrack density and aspect ratio distribution-which evolve with pressure-from elastic wave velocities, we quantitatively predict permeability variations using Dienes statistical percolation model. To validate this model, we measured porosity, permeability, and P- and S-wave velocities on four tight sandstone samples under effective pressures of 5 to 50 MPa. The results show that the proposed model accurately captures the evolution of both elastic parameters and permeability with effective pressure, demonstrating strong predictive capability for the experimental data. The significance of this study lies in achieving a quantitative relation between elastic and transport properties by explicitly characterizing the pore structure and integrating percolation theory, which is then calibrated with real experiment data, thereby providing solid physical basis for better prediction of reservoir permeability using acoustic logs and seismic data.

SummarySatellite formation configuration plays a crucial role in formulating the next generation gravity mission aiming at improving the accuracy and spatiotemporal resolution of Earth’s gravity field modeling. For the proposed GRACE-Pendulum (GRACE-P) formation, technical challenges and limited accuracy gains from the third observation link led to the adoption of only two baselines, without incorporating the third link perpendicular to the GRACE-type baseline, resulting in potential limitations in modeling accuracy. With future technological advancements overcoming these constraints, this study first proposes a complete triangular GRACE-Pendulum-full (GRACE-PF) formation, which includes the omitted third link, to explore the upper limit of accuracy achievable with this formation geometry. Building on this, we further introduce a novel satellite formation, namely the Three-Observation-Link (Three-OL) formation, inspired by the Chinese TianQin gravitational wave detection project. While geometrically similar to GRACE-PF, Three-OL does not include a GRACE-type baseline. To systematically evaluate the performance of these formations, a closed-loop simulation experiment was conducted, followed by a comparative analysis of GRACE-P, GRACE-PF, and Three-OL formations in the spectral-spatial domain. Specifically, spatial-domain results show that GRACE-PF outperforms GRACE-P by 28.7 per cent globally, confirming the accuracy contribution of the third link. Three-OL exhibits superior performance to GRACE-PF, with higher accuracy for mid-high degree signals in the spectral domain, and an 11.3 per cent accuracy improvement over land areas in the spatial domain. Furthermore, compared with the mainstream Bender and GRACE formations, Three-OL reduces the residual RMS over land areas by 28.4 per cent and 75.9 per cent, respectively, while also demonstrating improved capability for recovering sub-monthly gravity signals. In summary, this study reveals the accuracy potential of the complete GRACE-P configuration and proposes a novel formation capable of supporting high-precision temporal gravity field modeling. This work provides a feasible and efficient formation design for gravity satellite missions dedicated to Earth’s gravity field detection.

SummaryThe North Maladeta fault system, reactivated after the Alpine orogeny as a normal fault (Ortuño et al., 2008; Ortuño and Viaplana-Muzas, 2018) extends for almost 75 km across the border between Spain and France. This system is considered the most probable source of the historical M6.3 Ribagorça earthquake (1373). Within the system, the Coronas Fault is a secondary fault subparallel to the North Maladeta Fault and located in the southern slope of the Maladeta massif. Recent seismic activity in the massif has markedly increased since 2020, with previously unknown clusters of shallow earthquakes detected. Using a Template Matching approach, we verified that this apparent increase is not an artifact of network evolution, identifying additional clusters during 2014–2016, with far fewer events however. Relocation and focal mechanism analysis suggest that most of the seismicity is associated with the Coronas Fault, which appears to be less vertical and flatter than previously assumed. The seismicity is distributed on both sides of the fault, surrounding a 5–6 km wide aseismic segment. Another cluster, active since 2022 (Gerbosa cluster), likely corresponds to a previously unknown parallel fault.Temporal correlations indicate that cluster activation coincides with periods of low surface water levels. In particular, one cluster at the Western part of the Coronas fault, beneath the most rapidly shrinking part of the Aneto glacier, reactivated in late June 2020 and 2024, suggesting a possible influence of glacier melting. In the eastern Coronas fault, the two largest clusters were triggered within days of major rainfall events recorded in the past five years, consistent with a combination of short-term infiltration processes, and longer-term surface unloading from reduced snow and water storage since 2019. These observations raise the possibility of a climate–seismicity link in the Maladeta Massif, whereby rainfall and snowmelt infiltrations, and/or surface unloading may contribute to triggering shallow seismicity. Longer-term monitoring and detailed studies are required to confirm these processes and assess their potential implications for seismic hazard in the Pyrenees.

SummarySeismic stratigraphic interpretation of shelf-edge clinothems is essential for revealing tectonic evolution, paleoclimate change, depositional dynamic conditions, and hydrocarbon generation and accumulation during basin filling. However, traditional interpretation methods remain labor-intensive, time-consuming, and highly subjective. Although AI-based method offer a potential solution for automated this task, its development has been limited by the scarcity of comprehensive and representative benchmark datasets for shelf-edge clinothems. This limitation primarily arises from limited field data availability, the scarcity of reliable geological labels, and the structural complexity and strong variability of clinothem-dominated systems. To address this gap, we develop a hybrid benchmark dataset through two complementary strategies of field data curation and geological and geophysical forward modeling, ultimately generating 3,000 unlabeled field and 4,000 labeled synthetic seismic data, respectively. We further evaluate several representative baseline deep learning models on these datasets, and the accurate results demonstrate that the curated dataset provides an effective and representative basis for model training, quantitative assessment, and practical application. Finally, we have publicly released this hybrid benchmark dataset (https://doi.org/10.5281/zenodo.20769762) to facilitate the development, validation, and assessment of deep learning methods for automated seismic stratigraphic interpretation.

SummaryAdvances in high-precision gravimetric instrumentation have significantly improved the quality of gravity data and increased the need for refined theoretical modeling of local gravitational fields. Although analytical expressions for vertical gravitation (VG) of variable density two-dimensional (2-D) frustums has been studied earlier, these formulas are relatively complex, and much less attention has been paid to three-dimensional (3-D) frustums. This study derives analytical solutions for VG and vertical gradient of the vertical component of gravitation (VGG) of 2-D and 3-D frustums with a quadratic density function in the z-direction. Validation through comparison with numerical methods and previous computational results confirms the correctness and effectiveness of the proposed algorithm. Analysis of the derived formulas enables the quantitative determination of the relationship between the number of VGG extreme points (possibly three or five), the frustum size, and the observation points location. Based on the multibeam data and gravity model, the RMS misfits are 4.1, 3.5, and 3.1 mGal for the three Necker Ridge profiles, and approximately 10.5 Eötvös for the Horizon Tablemount VGG profile. Additionally, we apply this formula to calculate VGG of Horizon Tablemount, and successfully verify the existence of five extreme points. This characteristic demonstrates that the occurrence of significantly smaller local VGG is an inherent property of the VGG itself, independent of terrain undulation and specific density distribution patterns. This study provides effective computational tools and a theoretical framework for high-precision gravitation modeling.

SummaryThermal conductivity anisotropy influences heat flow in the crust, yet predicting it from rock fabric remains difficult because the effective conductivity of a polycrystalline rock depends not only on mineral thermal properties and crystallographic preferred orientation, but also on grain shape, phase distribution and grain-scale interactions. Here we extend an asymptotic expansion homogenization–finite element (AEH–FE) framework for polycrystals to anisotropic heat conduction in rocks. Single-crystal conductivity tensors are rotated according to measured crystallographic orientations and assigned throughout the microstructure, and 2-D electron backscatter diffraction (EBSD) maps are analysed as representative sections extruded normal to the section. The method yields both the homogenized conductivity tensor and microscale heat-flux fields for prescribed macroscopic temperature gradients. To characterize section-based anisotropy, we define a coordinate-independent in-plane measure from the extrema of directional conductivity. We apply the method to a simple natural microstructure comprising a single mica grain in a quartz matrix and to three natural phyllosilicate-rich rocks representing planar foliation, transitional crenulation cleavage, and fully developed crenulation cleavage. Progressive fabric reorganization reduces in-plane anisotropy (max/min thermal conductivity) from A2D = 2.77 for planar foliation to A2D = 1.22 for fully developed crenulation cleavage. The transitional fabric remains substantially anisotropic (A2D = 2.26) despite similar diagonal tensor components because a large off-diagonal term rotates the principal conductivity directions. Comparisons with Voigt, Reuss, Voigt–Reuss–Hill and geometric-mean estimates show that classical averages capture broad trends but can misrepresent both anisotropy magnitude and principal directions when microstructures are strongly textured or spatially organized. These results demonstrate that EBSD-resolved multiscale homogenization provides a physically grounded route from rock fabric to effective thermal conductivity tensors and microscale heat-flow patterns relevant to geophysical models of anisotropic crustal heat transport.

SummaryReflectivity estimation aims to enhance the resolution of seismic data, providing crucial support for the detailed inversion of reservoir parameters. We propose a method that combines supervised pre-training with synthetic data and physics-guided fine-tuning in field data to estimate reasonable reflectivity from seismic data. Initially, a U-shaped network is pre-trained by supervised learning on a large amount of synthetic seismic data. Subsequently, multiple geophysically meaningful constraints including structure-oriented smoothness, reflectivity sparsity, and data reconstruction, are introduced to formulate a self-supervised or unsupervised learning mechanism to fine-tune the pre-trained model so that it is better adapted to field data for obtaining more reasonable reflectivity. The pre-trained model provides an initial reflectivity that aligns with fundamental structural features. Based on the initial estimate, the model is further optimized through physics-guided fine-tuning to obtain a more reasonable reflectivity estimation that better reflects the characteristics of the field data and geophysical priors. Furthermore, the well-log-based correlation evaluation metric is applied to automatically and adaptively determine the early-stopping point of the fine-tuning process where favorable results are achieved. Experiments on synthetic and field seismic data confirm that the proposed method yields reasonable and high-resolution reflectivity estimation.

SummaryContinental intraplate regions exhibit slow deformation yet host destructive earthquakes, posing fundamental questions about strain partitioning far from plate boundaries. Here we combine a refined GNSS velocity field and elastic block modeling to investigate the Datong basin–range system in North China, a key intraplate deformation zone at the intersection of major fault systems and Quaternary volcanism. We find that left-lateral shear (∼2 mm/yr) and NW–SE extension (∼1 mm/yr) are accommodated across multiple faults and coherent block rotations, indicating broadly distributed deformation rather than localization on a single structure. Integration with mantle tomography reveals a low-velocity anomaly beneath Datong, suggesting that upwelling-driven thermal weakening augments far-field stresses from India–Asia collision and Pacific plate rollback. This deep–shallow coupling implies non-negligible seismic hazard and highlights the role of mantle dynamics in intraplate deformation.

SummaryLaterally distributed conductive bodies and intervening resistive regions within the transient electromagnetic (TEM) sensitivity region can produce anomalously enhanced responses over resistive zones. When interpreted with standard one-dimensional (1D) inversion, this response pattern can lead to laterally misplaced conductive anomalies. We refer to this family of inversion artefacts as array effects, because they arise when a subsurface array of conductive and resistive features jointly contributes to the measured TEM response, causing 1D inversion to recover conductive anomalies at incorrect lateral positions. The underlying physical cause is the conductive–resistive alternation mechanism: spatially separated conductive regions, interleaved with resistive gaps, can jointly influence the measured response, causing 1D inversion to place conductive anomalies into, or towards, intervening resistive regions. This produces severe artefacts in inversion models and can mislead geological interpretation. Such conditions may occur in geological settings characterized by repeated lateral electrical heterogeneity, provided that the electrical contrast and spacing are appropriate. Using smoke-ring theory and 3D numerical modelling, we investigate how these effects arise and propose the Instantaneous Smoke Ring Footprint (InSR-Footprint) as a practical criterion for model design and effect identification. Finally, we present field evidence for array effects in SkyTEM data acquired over a coastal dune environment in the Netherlands.

SummarySeismic surface wave tomography uses surface wave information to obtain velocity structures in the subsurface. Due to data noise and nonlinearity of the problem, surface wave tomography often has non-unique solutions. It is therefore required to quantify uncertainty of the results in order to better interpret the resulting images. Bayesian inference is the most widely-used method for this purpose. However, the commonly-used Monte Carlo methods require huge computational cost and remains intractable in high-dimensional problems. Variational inference uses optimization to solve Bayesian inverse problems, and therefore can be more efficient in the case of large datasets and high-dimensional parameter spaces. Variational inference has been widely applied to 2-D phase velocity map inversion. In this study, we extend the method to 3-D surface wave tomography by directly inverting for 3-D spatial seismic velocity structures from frequency-dependent travel time measurements. Specifically, we apply three variational methods, mean-field automatic differential variational inference (mean-field ADVI), physically structured variational inference (PSVI) and stochastic Stein varational gradient descent (sSVGD) to surface wave tomographic problems using both synthetic data and real data in the Southwest China. The results show that all methods can provide accurate velocity mean estimates, while sSVGD produces more reasonable uncertainty estimates than mean-field ADVI and PSVI because of Gaussian assumption used in these methods. In the real data case, the variational methods provide more detailed velocity structures than those obtained using traditional linearized methods, along with reliable uncertainty estimates. We therefore conclude that variational surface wave tomography can be applied fruitfully to many realistic problems.

SummaryOn 21 May 2021, a moderate earthquake with a magnitude of M6.4 occurred in western Yunnan Province, China. Several felt earthquakes had occurred in the region in the three days preceding this event. We aimed to investigate whether there are indications of crustal fluid migration in this region and, if so, the extent to which regional seismicity might be influenced by crustal fluids. Using the epidemic-type aftershock sequence (ETAS) model, we analyzed the earthquake triggering process and the nonstationary background rate. In addition, seismicity rates were used to estimate Coulomb stress changes. Specifically, the results suggest that the background rate increased from 0.2 to 15 events per day during the Yangbi earthquake sequence, and the spatial evolution of the background rate was broadly consistent with a fluid diffusion pattern. The evaluated cumulative stress change indicates that the stress rate increased by approximately 50 per cent following the M6.4 earthquake. Spatial stress changes suggest migration and expansion of regions with increased stress. These results imply the presence of crustal fluid flow in the study area, with a tendency to diffuse southward.

SummaryDistributed Acoustic Sensing (DAS) is an emerging technology that turns existing optical-fiber cables into high-density seismic arrays, generating vast amounts of observational data in various contexts. Consequently, large-scale storage, transmission and processing of DAS data present both challenges and opportunities in seismology. In this study, we propose a data compression and seismic detection workflow based on compressive sensing (CS) and apply it to DAS data from the Taiwan Milun Fault Drilling and All-inclusive Sensing (MiDAS) project. Our algorithm achieves a compression ratio of at least 15, with reconstructed data from specific earthquakes compatible with standard seismological algorithms. Additionally, we migrate the detection algorithm to the compressed domain, enabling quasi-real-time, high-accuracy seismic detection. This approach demonstrates the feasibility of directly processing compressed data, reducing computational burdens in DAS processing. Furthermore, we discuss an empirical criterion for determining the maximum compression ratio of a given signal based on CS. Our workflow is compared with other mature compression algorithms across eight different datasets to demonstrate its advantages, limitations, and applicability. Generally, the CS algorithm is more effective for high-SNR event-oriented DAS applications rather than continuous noise-dominated monitoring. Collectively, these results highlight the potential of the CS algorithm in advancing the development of efficient, user-friendly DAS data products.

SummaryThe pore-crack network, characterized by the distribution of pore aspect ratios and porosity, controls the seismic properties of a rock. The increase in laboratory ultrasonic velocity with effective pressure reflects the closure of microcracks, thereby allowing the pore-aspect-ratio spectrum to be inverted through rock-physics modelling. Previous studies on sandstone samples have suggested that the pore-aspect-ratio spectrum (porosity distribution vs. pore-aspect-ratio distribution) follows a power-law distribution; however, its robustness across different lithologies, model dependence, and origin remain poorly understood. In this study, these questions are further investigated. We compile laboratory ultrasonic velocity-pressure data from 50 rock samples spanning a wide range of lithologies and porosities. Then we invert the power-law exponent (the slope of the power-law distribution) from these data by integrating the pore-aspect-ratio spectrum into two independent rock-physics models: the existing Kuster-Toksöz (KT) model and the multi-crack wave theory (MCWT) that incorporates the squirt-flow mechanism. The inversion results show that both models obtain power-law pore-aspect-ratio spectra across lithologies, with the MCWT showing improved modelling of the squirt-flow effect in saturated P-wave velocities, leading to smaller fitting errors. Theoretical modelling explains the differences between the two models in their fitting behavior and inverted power-law exponent values. Furthermore, we demonstrate that the power-law pore-aspect-ratio spectrum can be derived from widely observed power-law fracture length and aperture distributions of the crack system. The result also establishes a relationship between the power-law exponent and porosity that well explains the global data trend from inverting the velocity-pressure data of rock samples. The significance of the power-law exponent and the applicability of the E-porosity relationship are also discussed. Our findings support a robust power-law pore-aspect-ratio spectrum across different lithologies and highlight its simplicity and practical applicability for analyzing velocity-pressure data in cracked-porous rocks.

SummaryAccurate numerical computation of traveltimes is essential for seismic applications, including traveltime tomography and seismic imaging. Most existing numerical methods have been developed on rectangular grids, whose regular structure limits their ability to capture rugged topography and irregular subsurface interfaces. To address this limitation, we develop a fast sweeping method (FSM) combined with the discontinuous Galerkin (DG) method to efficiently calculate traveltimes on triangular meshes. In the proposed method, the FSM offers an efficient Gauss–Seidel iterative framework, whereas the DG method solves the eikonal equation with second-order accuracy. The combination of these two techniques enables efficient computation of transmitted-wave traveltimes in models with complex geometries. After computing the transmitted-wave traveltimes, those at the reflection interfaces are employed as the initial conditions to further compute the traveltimes of reflected PP and PS waves in the elastic medium. Numerical simulations confirm the validity and accuracy of the proposed method in handling models with complex geometry.

SummaryFull-waveform inversion (FWI) is a key tool for velocity model building. Multiparameter elastic FWI suffers from parameter coupling and unbalanced radiation sensitivities, which generate crosstalk and hinder convergence. The Gauss–Newton (GN) method alleviates these limitations by incorporating second-order curvature information, but its computational cost remains a limiting factor. We present an adaptive local inverse-Hessian preconditioning approach for elastic FWI that approximates the GN update without solving the global linear system. The method constructs a local inverse mapping between curvature responses and model perturbations using reference perturbations and associated curvature, which is defined as the Hessian-vector product in sense of GN. This formulation captures both diagonal and off-diagonal contributions of the inverse Hessian, providing an explicit treatment of parameter coupling. The local systems are solved by singular value decomposition with Tikhonov regularization to improve stability. Numerical experiments on synthetic models, including a crosstalk-sensitive anomaly test and the Marmousi II model, indicate that the proposed method yields update directions consistent with those of a fully converged GN solution at substantially lower cost. Compared with block-diagonal pseudo-Hessian and truncated GN implementations, the method shows reduced crosstalk, faster misfit reduction, and improved reconstruction under comparable computational constraints.