Geophysical Journal International

SummaryReliable automatic phase picking is important for many seismic applications. With the development of machine learning approaches, many algorithms are proposed, evaluated and applied to different areas. Many of these algorithms are single station based, while recent proposed methods start to combine surrounding stations into consideration in the problem of phase picking. Among these algorithms, the Phase Neural Operator (PhaseNO) shows promising results on regional datasets comparing to existing algorithms. But there are many use cases for the local seismic networks in our community, therefore in this paper we evaluate the performance of PhaseNO on 4 different local datasets and compare the results to PhaseNet and EQTransformer. We used both individual phase picking metrics as well as association metrics to illustrate the performance of PhaseNO. By manually reviewing the newly detected events, we find the PhaseNO model outperforms the single station-based approaches in the local-scale use cases due to its consideration of coherent signals from multiple stations. We also explored PhaseNO’s behaviors when only using one station, as well as gradually increasing the number of stations in the seismic network to better understand its behavior. Overall, using the off-the-shelf machine learning based phase pickers, PhaseNO demonstrated its good performance on local-scale seismic networks.

SummaryEarthquake fault slip arises from nonlinear coupling among frictional evolution, elastic loading, and pore-pressure changes. When pore pressure evolves dynamically, the resulting hydro-mechanical rate-and-state models can be stiff and strongly coupled, making parameter inversion computationally demanding. Here we develop a physics-informed neural network (PINN) solver for a coupled spring–slider system that combines rate-and-state friction with pore-pressure/porosity evolution. The network approximates the time-dependent state variables and is trained by enforcing the governing differential equations together with initial conditions and, for inverse problems, observational constraints. To improve training stability, we employ adaptive inverse-residual weighting and a two-stage optimization schedule (Adam followed by L-BFGS). In forward simulations, PINN predictions closely match a Runge–Kutta reference solution across steady sliding and slow-slip transients, with normalized mean squared error below 0.08 and Pearson correlation coefficient above 0.975 for block velocity and frictional shear stress in the cases tested. In inverse experiments, the framework recovers the applied normal stress from noisy shear-stress observations; uncertainty increases with noise amplitude, but the ensemble mean remains stable, and at the highest noise level considered (q = 1) the inferred normal stress deviates by less than ~1% from the reference value. These results suggest that PINNs provide a differentiable alternative for forward modeling and parameter inversion in coupled hydro-mechanical rate-and-state fault models.

SummaryWe estimated seismic moment tensors (MTs) for the 2017 magnitude M6 Hojedk, Central Iran, earthquake triplet and their aftershocks, employing 1D and 3D regional and global velocity models to evaluate source parameter stability and resolution fitting in-country waveform data. We used the Moment Tensor Uncertainty Quantification (MTUQ) software, which performs a grid search for MT estimation and uncertainty analysis. For the regional 3D velocity model, we used MEAD-M20, a full-waveform inversion model of the Middle East derived from fitting 15-s body waves with 30-s body and surface waves. We compared the regional 1D- and 3D-based results with an existing database of deviatoric MT solutions, and for both regional velocity models, we found good agreement. However, for periods T ≥ 25–30 s and events with moment magnitudes Mw≥ 4.5, the 3D regional synthetic seismograms outperformed the 1D regional model, reducing waveform misfits, time shifts, and non-double-couple contributions. We consider non-double-couple contributions spurious and their reduction an improvement, as previous studies of the sequence found predominantly shear faulting on reverse faults. Furthermore, uncertainty analysis shows that the moment tensor, non-double-couple component, magnitude, and depth are more tightly constrained for the 3D model. The current 3D model shows no clear improvements relative to the 1D model in terms of misfit, time-shifts, and non-double-couple contributions at short periods T ≈ 15-25 s relevant for modeling smaller events. Using the regional models results in lower misfits and tighter constraints on the MT solutions than with the global 1D PREM and the 3D S2.9EA models. Improved MT estimation and parameter resolution for moderate-to-large events using in-country data validate the recently developed 3D Middle East velocity model. Further model refinements are needed to model shorter-period data required to analyze and improve the resolution of smaller (M ≤ 4) seismic events. Such improvements are within reach by using available in-country data and well-constrained MT solutions from a regional moment tensor database.

SummaryMonitoring the activity of non-natural seismic events is crucial for constructing an accurate seismic catalog, overseeing the safety of industrial operations, and mitigating the potential threats to local residents. However, recent studies have shown that neural network models trained on local data sets may not generalize well to a different region. Here, we leverage the Siamese neural network (SNN) to enhance the generalization of neural network models in discriminating between blasts, collapses, and natural earthquakes under regional shifts. Two distinct data sets are analyzed. The model is trained on a data set from northeastern China and tested on an out-of-region data set from the Inner Mongolia Autonomous Region and Gansu Province. We evaluate the prediction performance of the SNN model against the Convolutional neural network (CNN) model using the K-fold cross-validation technique. Results show that both the CNN and the SNN models achieve highly comparable performance on the in-domain validation data set. However, when applied to the out-of-region test data set, the SNN model with target-region anchors can improve the predicted AUPRC values by 7% and 4% compared with that of the out-of-region CNN model and the CNN model with transfer learning using target-region anchors, respectively. Furthermore, Grad-CAM importance weight analysis shows that the SNN model mainly relies on early-arrival P- and S-wave trains. The study suggests that SNN model with target-region anchors can deliver better generalization and flexibility than the conventional CNN model under regional shifts, which is particularly valuable for regions lacking labeled data sets.

SummaryAfter the 2015 Ms 6.5 Pishan earthquake, three moderate-magnitude earthquakes, the Ms 5.4 Pishan earthquake on September 4, 2021, the Ms 5.4 Yecheng earthquake on September 5, 2021 and the Ms 5.4 Pishan earthquake on October 23, 2022, occurred in the seismically active western Kunlun Range foreland thrust system. The seismogenic structures responsible for the three most recent earthquakes and their relationships with the 2015 Ms 6.5 Pishan event are still not understood. Integrated analysis of relocation results of the three main events and their aftershock sequences, focal mechanism solutions, and regional geology has identified the seismogenic fault and structural geometry near the earthquake source. Our results recognize a gentlely S-dipping Kuoshi fault ramp, which is the frontal structure at the west part of the WKFTS and is responsible for the three most recent moderate-magnitude earthquakes. The 2015 Ms 6.5 Pishan earthquake and the recent moderate-magnitude events were all generated by the frontal fault ramp, indicating a deformation pattern characterized by simple outward thrusting. In the western Kunlun Range foreland, the most active deformation and topographic growth have migrated northward relative to those of the higher terrace folds as the rear ramp slip ceases. Our results provide new insights into deformation pattern and seismic hazard in the region.

SummaryWe present a new, regionally adjusted local magnitude (ML) model for Switzerland and surrounding regions. The model is derived based on Wood-Anderson displacement amplitudes (AWA) calculated from 150,000 high-quality waveforms from 15,000 earthquakes between 2000 and 2025, recorded by more than 700 seismic instruments. This dataset is substantially richer than those used in previous ML studies in Switzerland, with a large number of near-source recordings and data from low-magnitude events, which were notably sparse in earlier works. AWA attenuation over hypocentral distance is parametrised through linear and logarithmic distance terms along with hinge distance points, which allow proper modelling of the attenuation characteristics at long distances and changes in attenuation associated with post-critical reflected phases. Regional differences in attenuation between the Alpine region in southern Switzerland and the northern Foreland are smoothly modelled through a ray-path-specific regional adjustment parameter, allowing the model coefficients and the hinge distances to vary spatially. The coefficients of the parametric attenuation curves are estimated using mixed-effects regressions, and the model is anchored to yield a magnitude 3 for an AWA of 10 mm measured at a hypocentral distance of 17 km. The station terms are calculated with respect to Swiss reference rock conditions. The new ML model reduces uncertainty by 33 per cent compared to the current ML scale used by the Swiss Seismological Service and does not exhibit any residual trends with respect to hypocentral distance, earthquake depth, local site conditions, or event magnitude. Empirical radiation pattern corrections are derived, further reducing the uncertainty by 8 per cent for strike-slip events. Alternative models, based on non-parametric and cell-based 2D approaches, are derived independently to validate the parametrisation of the parametric model. The new model – MLS26 – yields lower magnitudes for smaller events (with catalogue magnitudes lower than about 2.5) and for events located in the northern Foreland, whereas the magnitudes of the larger Alpine events remain similar. The reduced magnitudes of smaller events decrease the b-value of the input earthquake catalogue from 1.00 to 0.93, corresponding to a reduction of about 7 per cent. MLS26 scales one-to-one with moment magnitude (MW) for MLS26 > 4, while for smaller events, it scales with the logarithm of the seismic moment.

SummaryThis study aims to understand the recurrent seismicity that occurs in a limited volume along a major fault in the French western Alps, the Vuache Fault, which crosses the geological Jura in a flat-and-ramp zone. In 1996, an M5.3 earthquake occurred near Annecy (France), located at a depth of approximately 2 km. In this article, we analyze the seismicity that has occurred since then and calculate the seismic velocity variations at local permanent seismic stations. The magnitude and focal mechanism of the 1996 M5.3 earthquake indicate that the process was tectonic in origin. However, the duration of the Omori decay of its aftershocks, their migration, the variations in spring flow, the existence of repeated swarms, the seasonal variations in seismic velocity and their relationship to rainfall and seismicity show that the continuation of this seismicity is linked to the pressurization of fluids in a deep, fractured aquifer connected to the surface. This aquifer appears to be limited to the sedimentary formations. Earthquakes, especially during the M5.3 aftershock migration, reach the depth of the Triassic gypsum. The aftershock sequence occurred in two different phases: an initial phase lasting around ten days, during which the earthquakes did not show any migration but rather a random spatial distribution, and a second phase showing a clear migration, suggesting a fluid diffusion process. During this last phase, the hydraulic diffusivity of the aquifer was calculated and estimated at around 2 m²/s. The same order of magnitude was obtained by using the correlation between seismic velocity variations and rainfall. This is a high value, close to those found during man-made fluid injections. This aquifer forms a confined, pressurized reservoir between two thrusts in the regional flat-and-ramp structure. The scenario described in this article could be found, at various scales, in flat-and-ramp regions where tectonic stresses are sufficient to generate seismicity.

SummaryModeling seismic wave attenuation and dispersion in fluid-saturated porous media is essential for reservoir characterization; however, significant challenges remain in accurately capturing the effects of anisotropy. Unified theoretical frameworks that combine Biot and squirt flow mechanisms have often been limited by two key factors: they are typically based on the isotropic assumptions and rely on oversimplified physical models for squirt flow. Consequently, a time-domain numerical implementation for advanced, physically-based anisotropic squirt models has been lacking. This study presents a unified theoretical and numerical framework that, for the first time, integrates Biot’s theory of anisotropic poroelasticity with a state-of-the-art model for anisotropic squirt flow based on one-dimensional fluid pressure diffusion in cracks partially connected to spherical pores. The core innovation is a time-domain implementation achieved through a semi-analytical conversion of the complex, frequency-dependent frame moduli into a Generalized Zener Model representation. With parameters optimized via a genetic algorithm, the system is expressed as a set of differential equations with memory variables, enabling efficient finite-difference time-domain (FDTD) simulations in complex heterogeneous media. Our numerical results demonstrate that the model accurately captures frequency- and angle-dependent velocity dispersion and attenuation in VTI media due to squirt flow in the seismic-to-sonic frequency band. The FDTD algorithm is rigorously validated against analytical solutions, and simulations in heterogeneous media highlight its capability to capture spatially-varying anisotropic attenuation effects. This framework bridges a critical gap between advanced rock physics theory and practical wavefield simulation, providing an accurate forward modeling tool for interpreting seismic data to characterize complex reservoir rocks.

SummaryDuring the Early Medieval Ages, unusually strong and rapid geomagnetic field variations have been reported in several European regions; however, archeomagnetic data from Central Europe remain scarce. To help filling this gap, we present new archeointensity results from eight archeological sites in Germany, Austria, and Poland, dated between 500 and 1200 AD. The investigated materials mainly consist of potsherds, together with two in situ baked clay structures that also provided archeodirectional information. Archeointensities were determined using the MT4 protocol, a Thellier-type technique including cooling-rate and anisotropy corrections. For two sites, the multi-specimen domain-state-corrected paleointensity protocol was additionally applied. Rock magnetic experiments indicate that the main remanence carriers are low-coercivity magnetite and high-coercivity ε-maghemite and hematite. The presence of these phases suggests incomplete transformation to hematite during firing. To further assess the archeointensity determinations, the Bias Corrected Estimation of Paleointensity (BiCEP) method was applied, particularly for specimens showing curved Arai plots. This analysis confirmed the reliability of the Thellier results for most investigated structures, whereas no reliable BiCEP outcome could be obtained for one structure from Chobienia (Poland). For four structures, the classical Arai plot evaluation agrees with the BiCEP results and, for one site, also with the independently obtained multi-specimen results. In another case, the comparison between Thellier and BiCEP estimates allowed a more realistic assessment of intensity uncertainty. One site mean value (~ 50 µT) around 600 AD yields a lower geomagnetic field intensity than other contemporaneous European records. Overall, the data suggest an increase in field intensity between 600 and 800 AD, with values becoming more consistent with previously published regional results after this period. However, given the relatively large uncertainties and the still limited number of available studies, additional archeointensity data from 500-800 AD are needed to determine whether the observed regional differences reflect genuine geomagnetic field heterogeneity during this period. Furthermore, new chronological constraints were obtained through archeomagnetic dating approaches applied to the two available full-vector records.

SummaryHuang et al. (2025) reported the detection of 67 teleseismic marsquakes identified by P and S wave arrivals. The authors used a deep learning phase picker trained on local earthquake data and applied it to narrow-bandpass filtered seismic data recorded by NASA’s InSight seismometer, making use of similarities between local earthquake and teleiseismic marsquake recordings when adjusting for sampling rate and S-P time scaling relations. We review all detections as similarly done for the Marsquake Service catalogue and other studies on this data set, using the complementary wind and pressure data recorded by InSight. As these auxiliary data were not recorded in the second half of the mission, we also infer wind contamination from bandwidths in the seismic data that contain wind-sensitive lander modes. Additionally, we analyse the signal polarisation to compare it with the expected characteristics of P and S waves and the background noise. Our review indicates that all 67 detections reported by the authors correspond to atmospheric noise. In most cases, the detections relate to the seismic signature of small wind bursts followed by larger wind bursts, onsets of which are interpreted as P and S waves by the authors. Further, we show that if these events were interpreted as genuine marsquakes, their inferred epicentral distance distribution would not match typical marsquake distances, while their magnitudes would make them the largest events of the catalogue. For future studies that deal with seismic event detection and interpretation from InSight, we recommend a careful consideration of the established event and noise signal markers described in this comment and in the literature to avoid misinterpretation of noise as event signals.

SummaryHigh-speed train (HST) signals offer an abundant and eco-friendly seismic source along the railway, but their application to full waveform inversion (FWI) presents unique challenges due to complex source characteristics and field constraints. We develop an HST-based elastic FWI framework that models the train as a distributed pier-force: a chain of time-delayed excitations applied continuously along deeply embedded bridge piers. This source mechanism generates predominantly S- and Rayleigh-wave energy, providing enhanced shallow subsurface illumination. The synthetic anomaly experiment demonstrates that the distributed pier-force enables superior multi-parameter reconstruction, particularly achieving higher density accuracy than the explosive and single-force sources. Additional sparse acquisition tests validate the robustness of the method under limited receiver coverage. The field data application successfully reconstructs subsurface layering through simultaneous source and parameter inversion, supporting the physical plausibility of the distributed source model. Overall, this study establishes the distributed pier-force as an effective and computationally efficient HST source for shallow subsurface characterisation, while also highlighting limitations under realistic field conditions.

SummaryThe Laji Shan-Jishi Shan Tectonic Belt (L-JTB), situated at a crucial tectonic transfer position on the northeastern margin of the Tibetan Plateau, is important for understanding regional deformation partitioning and seismic hazard. Although it has traditionally been regarded as a zone of relatively modest deformation, the 2023 Jishishan Mw 6.0 earthquake highlights the need to reassess its present-day kinematics and mechanical role. In this study, we integrate GNSS and Sentinel-1 InSAR data to derive a high-resolution three-dimensional interseismic crustal deformation field. We then apply an enhanced dip-aware velocity-profile model within a Bayesian framework to estimate the slip rates and locking characteristics of the major faults, and we further calculate the regional strain-rate field using a multiscale spherical wavelet approach. The results indicate that: (1) The L-JTB exhibits a modest average regional strain level relative to the first-order boundary faults to its north and south, but also displays pronounced along-strike heterogeneity and localized deformation focusing within the belt; (2) The Laji Shan Fault is characterized by a comparatively weak interseismic signal and is dominated by shortening and uplift, with only a minor fault-parallel component, whereas the Jishi Shan Fault is more strongly active, accommodating dextral strike-slip at 1.7 ± 0.3 mm/yr together with shortening-related dip-slip at 3.5 ± 1.2 mm/yr; and (3) the L-JTB is better interpreted as an internal strain transfer belt between the dextral Riyue Shan Fault and the sinistral West Qinling Fault, in which part of the transferred deformation is converted into crustal shortening, uplift, and localized internal strain. Our study reveals that even within a tectonic transfer belt characterized by only modest average regional strain, moderate earthquakes may still nucleate where deformation becomes localized through favorable stepover geometry, structural segmentation, and local strain focusing.

SummaryWe investigate the 410 km discontinuity (thereafter the ‘410’) beneath Europe using teleseismic S-to-P converted waves. This discontinuity—associated with the olivine–wadsleyite phase transition at the top of the mantle transition zone (MTZ)—is widely used as an indicator of thermal variations linked to mantle upwellings and subducting slabs. Our results reveal complex structures of the ‘410’, including closely spaced discontinuities above and below 410 km depth, within laterally confined regions (∼100-200 km wide). These features are predominantly aligned along a corridor extending from the western Alpine front (Western Alps, Rhone Graben) through the Eifel volcanic fields and the Eger Graben to the eastern Alpine front (Western Carpathians, northern Pannonian Basin, Eastern Carpathians), along the European Cenozoic Rift System (ECRIS). In several locations, the disturbed ‘410’ is associated with an elevated lithosphere–asthenosphere boundary (LAB), suggesting a link between lithospheric processes and MTZ structure. We interpret these observations as evidence for small-scale lithospheric dripping, potentially initiated during or following the Alpine orogenesis. Additional occurrences of apparent ‘410‘ doubling are identified beneath the southern Scandes and the northern Adriatic region; beneath the Scandes, this feature spatially correlates with extensional tectonics. Overall, our results indicate that the ‘410’ beneath Europe is strongly influenced by smaller-scale mantle dynamics, closely related to the ECRIS, rather than reflecting solely large-scale thermal anomalies.

SummaryRayleigh wave dispersion curves inversion is an important method for shallow shear wave velocity structure imaging, which can be achieved through different frameworks such as deterministic inversion and Bayesian inversion. The deterministic methods with fixed parametrization usually require pre-set model complexity, and are difficult to directly provide posterior uncertainty estimation. In contrast, trans-dimensional Bayesian method can probabilistically estimate the dimensions of the model during the sampling process and quantify the uncertainty of the inversion results. However, in the inversion of multimode Rayleigh wave dispersion curves, the posterior space usually has high-dimensional, multi-modal, and strongly nonlinear characteristics. How to achieve efficient posterior exploration and stable trans-dimensional mixing is still a key issue in practical applications. In response to this issue, we constructed and evaluated a trans-dimensional Parallel Tempering reversible jump Markov Chain Monte Carlo (PT-rjMCMC) inversion workflow for multi-modal Rayleigh wave dispersion curves in shear wave velocity imaging, denoted as All-Pair-Sweep PT-rjMCMC (APS-PT-rjMCMC). The method integrated Voronoi trans-dimensional parameterization, multi-modal dispersion likelihood function, reversible jump model update, and parallel tempering sampling into a unified framework. In the replica exchange stage, it adopted an all-pair-sweep replica-exchange schedule with randomized ordering to enhance the inter chain information propagation and trans-dimensional mixing ability under a limited number of temperature chains. The inversion results of synthesized model and measured data indicated that compared with the benchmark implementation, the workflow exhibited better performance in convergence behavior, posterior structure recovery, and layer identification. Our technology provides a solution that combines adaptability and reliability for fine survey of shallow geological structures. It is effectively improving the inversion accuracy of shear wave velocity structures under complex geological conditions. It has broad application space and significant application value in fields such as engineering survey, geological hazard assessment, and water resources investigations.

AbstractThe Poisson equation governing a planet’s gravitational field is posed on the unbounded domain, $\mathbb {R}^3$, whereas finite-element computations require bounded meshes. We implement and compare three strategies for handling the infinite exterior in the finite-element method: (i) naive domain truncation; (ii) Dirichlet-to-Neumann (DtN) map on a truncated boundary; (iii) multipole expansion on a truncated boundary. While all these methods are known within the geophysical literature, we discuss their parallel implementations within modern open-source finite-element codes, focusing specifically on the widely-used MFEM package. We consider both calculating the gravitational potential for a static density structure and computing the linearised perturbation to the potential caused by a displacement field—a necessary step for coupling self-gravitation into planetary dynamics. In contrast to some earlier studies, we find that the domain truncation method can provide accurate solutions at an acceptable cost, with suitable coarsening of the mesh within the exterior domain. Nevertheless, the DtN and multipole methods provide superior accuracy at a lower cost within large-scale parallel geophysical simulations despite their need for non-local communication associated with spherical harmonic expansions. The DtN method, in particular, admits an efficient parallel implementation based on an MPI-communicator limited to processors that contain part of the mesh’s outer boundary. A series of further illustrative calculations are provided to show the potential of the DtN and multipole methods within realistic geophysical modelling.

SUMMARYNatural faults commonly contain fluid-saturated gouge layers, in which fluid injection can modify pore pressure and porosity evolution, thereby affecting slip stability and induced seismicity. Here we develop a spring-slider model based on rate-and-state friction (RSF) to investigate fault slip evolution under dry, fluid-saturated, and fluid injection conditions. Our model incorporates gouge dilatancy/compaction and fluid-related pore pressure effects. Our results show pronounced differences in the onset time of the first dynamic instability and the peak slip rate among these cases. Compared to the dry case, fluid saturation without injection delays instability and slightly lowers the peak slip rate, whereas rapid injection-induced pressurization triggers earlier dynamic slip and higher peak slip rates. Without injection, increasing the dilatancy coefficient systematically delays instability. Under rapid injection, however, the onset time becomes much less sensitive to dilatancy, indicating a gradual transition from a dilatancy-influenced to an injection-dominated nucleation regime as the injection-dilatancy competition number increases. Linear stability analysis further suggests that fault stability can be characterized by a generalized critical stiffness that combines the effects of effective normal stress, pressurization rate, and dilatancy/compaction feedback. These results indicate that fluid effects on fault rupture arise from the competition between stabilizing dilatancy hardening and destabilizing time-dependent pressurization, highlighting that injection-induced seismicity can be understood as an injection-rate-driven stability problem. Our findings provide a physical framework for understanding the transition from dilatancy-sensitive to injection-controlled fault slip within the explored net dilatant regime under different fluid environments.

SummaryThis study leverages a new, improved and densified GNSS velocity field of the western Aegean region to quantify slip rates, strain localisation, and interseismic loading within the upper plate of the Hellenic subduction zone, including the deformation systems associated with the subduction–collision transition around Cephalonia, the Hellenic forearc extension, and the southwestern termination of the North Anatolian Fault system. We examine several active tectonic domains, comprising four extensional regions (Corinth–Patras Rift, Evia Gulf, Argolic Gulf, southwestern Peloponnese) and three strike-slip systems (Cephalonia Transform Fault, Katouna–Stamna Fault System, Movri Fault Zone). Across the Corinth Rift, NS extension increases westward from ≈ 7 to ≈ 15 mm.yr$^{\hspace{1.0pt}\text{--}1}$, of which up to ≈ 6 mm.yr$^{\hspace{1.0pt}\text{--}1}$ is accommodated offshore. Velocity profiles indicate combined elastic loading and aseismic creep on a limited number of crustal-scale faults. The Evia–Boeotia sector undergoes NS extension at up to 8 mm.yr$^{\hspace{1.0pt}\text{--}1}$, but deformation is distributed across multiple structures, each accommodating creep or elastic loading at rates <2 mm.yr$^{\hspace{1.0pt}\text{--}1}$. In the southern Peloponnese, diffuse EW extension of up to 6 mm.yr$^{\hspace{1.0pt}\text{--}1}$ occurs alongside 1–2 mm.yr$^{\hspace{1.0pt}\text{--}1}$ of NS extension in the central Peloponnese. A portion of this EW deformation may be accumulating interseismically as elastic strain on prominent structures, such as the Sparti and East Messenia faults. No measurable strain is detected across the Argolic Gulf, suggesting substantially lower present-day loading rates on the Astros Fault than previously inferred. Strike-slip systems display contrasting behaviours. The Katouna segment accommodates transtensional left-lateral creep of ≈ 13 mm.yr$^{\hspace{1.0pt}\text{--}1}$ within a zone <6 km wide, whereas slip decreases to ≈ 8 mm.yr$^{\hspace{1.0pt}\text{--}1}$ on the Stamna segment, consistent with strain transfer through the Trichonida pull-apart basin and the Nafpaktia diffuse shear zone. In contrast, the Movri Fault appears locked down to at least 10 km depth, accumulating ≈ 4 mm.yr$^{\hspace{1.0pt}\text{--}1}$ of right-lateral strain. Onshore velocities near the Cephalonia Transform Fault indicate an onshore half-rate of elastic loading of ≈ 8 mm.yr$^{\hspace{1.0pt}\text{--}1}$, suggesting that the full transpressional right-lateral motion (≈ 16 mm.yr$^{\hspace{1.0pt}\text{--}1}$) accumulates interseismically, highlighting considerable seismic hazard potential.

SummaryDistributed Acoustic Sensing (DAS) data often contain various types of noise, including random noise, coherent noise (e.g., coupling or linear noise), and common mode noise, which significantly degrade seismic signal quality. Conventional denoising methods struggle to effectively suppress diverse noise components while preserving important seismic signals. To address this issue, we propose a denoising self-supervised cascade network (DAS-DSCnet), a multi-stage neural network designed to progressively denoise DAS data without requiring external labels or synthetic training data generation. The network consists of three stages: Stage 1 targets random noise using a Noise2Noise-based approach; Stage 2 suppresses dataset-specific coherent noise using a denoising convolutional neural network (DnCNN)-based network trained with internally extracted noise patches; and Stage 3 predicts and removes common mode noise through trace shuffling and a Noise2Noise-based model. Training data for each stage are generated directly from the input DAS data by exploiting the data’s inherent characteristics, enabling efficient learning that reflects field-specific noise features. The model was evaluated using two distinct field DAS datasets with different noise patterns. The results demonstrate that DAS-DSCnet achieves superior noise suppression compared to conventional approaches, enhancing signal continuity while minimizing leakage. The denoising performance remains stable across different stacking configurations and hyperparameters, confirming the model’s robustness. Therefore, DAS-DSCnet offers a scalable and practical framework for improving seismic data quality in DAS applications, demonstrating the potential for fully automated, data-driven denoising in large-scale seismic monitoring.

SUMMARYWe present a comprehensive dataset of 920 coordinates and 509 velocities for geodetic points in central Greece and the Peloponnese, an area characterised by intense tectonic deformation. The points, with observation periods within the 1990–2024 range, are organised into three categories: permanent stations, triangulation pillars, and markers. The latter two categories are subdivided according to whether or not they feature self-centring. Most of the triangulation pillars belong to the Greek national network originally surveyed in the 1960s–70s. The GNSS data were processed using the GIPSY 6.4 software. To assess the secular velocities, we corrected for co-seismic and post-seismic displacements using earthquake parameters constrained by the time series of the permanent stations. Self-centring systems improve precision, reducing the average horizontal coordinate residual variability from 6.15 to 4.45 mm. The velocity uncertainties stabilise below 0.15 mm yr−1 when the time series exceed twenty years. Points with self-centring achieve 0.2 mm yr−1 accuracy after twelve years of data, compared to twenty years for those without self-centring. After twenty-five years, campaign points observed eight to ten times match the precision of permanent stations. The velocities at the campaign points further validate the HELVEL model previously developed using permanent stations only. We calculate a seven-parameter transformation from the original coordinates of 424 triangulation pillars to their GNSS-based ITRF2020 coordinates at epoch 2020.0. The lowest mean scatter after the transformation is 0.134 m when 1965 is used as the mean epoch for the triangulation data. We then apply this transformation to all 9,729 pillars of the study area. At the 424 resurveyed pillars, the GNSS ellipsoidal heights agree with the sum of the levelled heights and the official HG2023 geoid heights to within 0.184 m root-mean-square. Our dataset is entirely referenced to ITRF2020 at epoch 2020.0, which enables interoperability with previous and future geodetic studies. Dense campaign point arrays are critical for resolving the strain distribution at the scale of individual active faults, beyond the reach of arrays of permanent stations alone.

AbstractMantle viscosity remains one of the largest outstanding uncertainties in global geodynamics. Time-dependent mantle circulation models that assimilate tectonic histories (MCMs) provide a way to test viscosity by assessing their present-day predictions against observations. This approach allows for the influence of viscosity on mantle density structure to be accounted for, which is not possible using instantaneous modelling approaches. Here we present the first systematic test of lower mantle viscosity against dynamic topography, the geoid, and seismic heterogeneity using high-resolution MCMs. Model density structure depends strongly on the assumed viscosity profile, which in turn controls the fit to seismic heterogeneity. The fit to dynamic topography and the geoid is further influenced by the instantaneous transmission of stresses to the surface. These two effects can either reinforce or counteract each other at different depths, which must be considered when attempting to match dynamic topography and geoid amplitudes. MCMs typically overestimate dynamic topography amplitudes. We find that it is possible to reduce these amplitudes by lowering viscosity in the upper lower mantle (≈660-2000 km), though this comes at the expense of a reduced fit to the geoid and/or seismic heterogeneity. Our preferred viscosity profile provides an excellent fit to observed geoid amplitudes and the seismic heterogeneity of S40RTS. We also tested an alternate tectonic reconstruction with tomography-based refinements around the Pacific which improved the correlation with the observed geoid by ≈20%. Our results show that MCMs can now reach a level of resolution and realism sufficient for comparison to multiple independent data sets, opening the door to systematic assessment of uncertain parameters which govern convection in the mantle.