Geophysical Journal International

SummarySeismic source classification often relies on a single best-fitting moment tensor and deterministic rules on source-type diagrams. This study formulates classification directly in terms of the posterior distribution of the moment tensor on the fundamental lune. Moment tensors from waveform inversion are mapped into a uniform (v, w) source-type domain and used to build empirical source-type probability density functions p(v, w). We evaluate five hypotheses (±ISOCAL, DC, ±CLVD) plus a residual baseline, where ISOCAL denotes a calibrated isotropic kernel relative to the canonical ISO endmember. Classification probabilities P(Hi) are obtained by integrating p(v, w) against the hypothesis kernels; a residual probability Presidual is included as a sixth baseline (uniform) model that absorbs probability when the posterior is diffuse or poorly matched by the prototypes. The framework is applied to published regional moment tensor ensembles for 32 events, including 23 nuclear tests at the Nevada National Security Site and the DPRK Punggye-Ri test site, plus nearby earthquakes and collapses. Well-recorded underground nuclear tests yield tightly localized +ISOCAL probabilities with small Presidual, while events recorded by sparse, complex source-receiver paths show broader probability vectors and larger Presidual. Tectonic earthquakes map primarily to double-couple and CLVD hypotheses, while collapses map primarily to negative isotropic (–ISOCAL), with secondary CLVD in some events. The method is applicable wherever source-type posteriors are available, given a declared hypothesis family, and produces explicit hypothesis probabilities with uncertainty and a residual diagnostic that flags diffuse or weakly matched posteriors, and can feed directly into monitoring and decision pipelines.

SummaryThe Marine Self-Potential (SP) method has proven to be a useful geophysical tool for localizing seafloor massive sulfide (SMS) deposits. Conventional geophysical inversion approaches face significant limitations in mid-ocean ridge environments with complex topography, including high computational requirements due to high degrees of freedom in grid discretization and the complex computation of sensitivities. Meanwhile, data-driven neural networks, while powerful, often suffer from poor generalizability and interpretability in complex deep-sea environments, owing to their reliance on large-scale training datasets. In this study, we propose a simple-formulated intelligent SP inversion framework, called SP-PINN, based on the Physics-Informed Neural Network (PINN). By embedding the governing equation of SP as a physical constraint into the loss function, SP-PINN effectively couples the intrinsic correlation between SP signals and subseafloor causative sources. To address the characteristics of topographic relief in mid-ocean ridges and discrete distribution of SMS ore bodies, 2D and 3D synthetic case studies were constructed to verify the inversion performance of SP-PINN. Two SMS deposit bodies on the Southwest Indian Ridge (SWIR) were successfully imaged and delineated through the application of SP-PINN to the field data. Our results demonstrate that accurate imaging of SMS can be achieved by introducing physical constraint terms into the neural network, properly optimizing hyperparameters, and incorporating a small amount of field data. This study provides a straightforward and physically interpretable SP imaging method for the assessment of SMS under data-scarce conditions.

SummaryShallow-water depth data serve as the essential foundation for marine spatial planning and coastal engineering projects. However, due to the attenuation of spectral signals, traditional retrieval of shallow-water seafloor topography struggles to maintain high accuracy in highly turbid waters and deeper areas. Furthermore, existing machine learning models incorporating spatial information are highly susceptible to spatial overfitting, which severely limits their generalization ability in complex waters. To address these issues, this study proposes a depth-stratified Extreme Gradient Boosting (XGBoost) approach combining Sentinel-2 remote sensing data and multi-source gravity data. This model innovatively incorporates gravity features, which are unaffected by water turbidity, effectively overcoming the signal attenuation limitations of spectral-only models in deep and turbid waters. In addition, the stratification strategy effectively mitigates spatial overfitting, which commonly affects machine learning models. Application results in the Pearl River Estuary, China, showed that compared to the unstratified spatially uncorrelated XGBoost global model, the root mean square error (RMSE) of the proposed model was significantly reduced by 1.362 m. Ablation experiments further confirmed that incorporating gravity features reduced the RMSE in deeper-water areas (depth > 30 m) by 1.033 m, and dropped the RMSE in highly turbid waters from 1.548 m to 0.955 m. Finally, validation in an adjacent area showed that the model maintained useful local spatial generalization under similar environmental conditions, with an RMSE of 0.591 m in the 30–35 m interval.

SummaryAccurate determination of rate-and-state friction parameters from experiments is crucial for physics-based modeling of the earthquake source. In this study, we develop a Bayesian inversion framework based on a spring-slider model of the experiment to infer rate-and-state frictional properties from stable velocity-step experiments of a fine-grained quartz gouge that is close to instability. The inversion framework infers frictional parameters with superior precision to conventional methods based on non-linear least squares. We identify that logarithmic sampling of the characteristic slip DRS is an essential step for accurate inversion. Testing the same gouge material under the same effective normal stress but different pore pressure additionally shows that the presence of pressurized pore fluids changes the inferred rate-and-state properties of the gouge material by increasing rate-and-state parameters a, b, and DRS. The Bayesian inversion framework provides a valuable tool for inferring accurately rate-and-state parameters of realistic gouge materials close to instability, with a small and quantifiable range of uncertainty in cases where the steady-state behavior quantified by a − b can be constrained independently. We also find that the steady-state friction properties evolve over the course of the experiment, adding additional uncertainty to the rate-and-state parameter estimates but not altering the conclusion about significant differences between the rate-and-state parameters in the presence of fluids.

SummaryEvaluating earthquake forecasts is a crucial step in understanding and improving the capabilities of forecasting models. The use of specific metrics to assess the consistency between forecasts and data on one particular aspect of the process is important to understand which aspects of seismicity a model is failing to describe, and, consequently, highlight where new versions of the model should improve. This can be done effectively only by metrics unaffected by inconsistencies in other aspects of the process. The Collaboratory for the Study of Earthquake Predictability (CSEP), which organises earthquake forecasting experiments around the globe, has developed different tests targeting different aspects of the process, such as the number N-test or the magnitude M-test, assessing (in)consistency between the observed and forecasted number and magnitude distributions of events, respectively. We find that the results of the recently proposed M-test for catalog-based forecasts (composed of a collection of synthetic catalogs from the model) depend on the N-test, i.e. the two tests do not isolate the desired aspects appropriately, rendering uninformative forecast results. Here, we address this problem using simulated data and provide a solution based on resampling of simulated catalogs. We implement this new M-test with resampling in the pyCSEP software toolkit, conduct two analyses using actual earthquake forecasts for Europe (1990-2015) and Switzerland (1933-1962, 1962-1992, 1993-2022) that provide inconsistent earthquake counts compared to observations, and analyse how the test results change using this proposed test. Lastly, we investigate alternative metrics, namely an unnormalised M-test, two Chi-square formulations, the Hellinger distance, the Brier score, and a novel Multinomial Log-Likelihood (MLL) score, and compare them based on the ability to find (in)consistency between data and forecast in various synthetic scenarios. We find that there are scenarios in which the alternative metrics outperform the resampled M-test. In particular, the MLL test outperforms the M-test in all the scenarios considered. We also study how the ability of finding inconsistency changes with the number of observations, and the cutoff magnitude, comparing one of the Chi-square metrics, the MLL, and the M-test against classical statistical methods such as the Kolmogorov-Smirnoff (KS) test, the Wilcoxon test, and the Anderson-Darling (AD) test. The MLL and the AD test are the ones providing the highest probability of finding inconsistencies for all combinations of number of observations and cutoff magnitude. This study shows how realistic synthetic examples can be used to compare the ability of different metrics in finding inconsistencies between data and forecasts, and shows that the MLL is the metric providing the best tradeoff between interpretability and probability of finding inconsistencies and, therefore, should be used.

SummaryFull waveform inversion (FWI) has significant potential to reconstruct the subsurface structure at high resolution, but it is a nonlinear and ill-posed problem. Traditional physics-based FWI methods that employ gradient-based optimization algorithms to minimize the discrepancy between observed and synthetic data face several challenges. The aforementioned methods are susceptible to becoming trapped in local minima as a result of cycle-skipping problems and require substantial computational resources, particularly with the increasing size and required resolution of seismic datasets. Nowadays many deep-learning methods are proposed to solve the above challenges. FWIGAN, a recently proposed unsupervised seismic velocity inversion method based on the Wasserstein generative adversarial network with gradient penalty (WGAN-GP), has demonstrated certain success in overcoming these challenges. However, FWIGAN enforces the Lipschitz constraint on the critic through a gradient penalty evaluated at dynamically sampled interpolation points between the observed and synthetic data. This strategy introduces additional computational and hyperparameter tuning costs. Motivated by these limitations, we introduce the spectral normalization into the FWIGAN framework, directly constraining the spectral norm of each network layer to control the Lipschitz behavior of the critic. It can enhance the stability of network training. The proposed method mitigates the dependence on having large labeled data and the pretraining of the network. The new scheme is applied to the Marmousi and Sigsbee models to verify its robust performance. The experimental results clearly indicate that our method outperforms common methods in terms of inversion accuracy.

SummaryPolar motion (PM) provides an independent, observation-based constraint on global mass redistribution, but interpretation of its long-term variability is primarily limited by uncertainty in the glacial isostatic adjustment (GIA) contribution, for which conventional model-based estimates exhibit substantial differences. Here we develop a refined approach that uses primarily GRACE and GRACE Follow-On (GRACE-FO) observations to estimate the secular PM associated with non-surface-loading processes, predominantly GIA. We restore the pole tide contribution removed during standard GRACE processing to reconstruct the total degree-2, order-1 spherical harmonic coefficients and the corresponding PM excitation, then isolate the non-surface-loading component as the difference between this excitation and the GRACE/GRACE-FO-derived excitation from contemporary surface loads and pole tides. This strategy avoids uncertainties associated with atmospheric winds, ocean currents, and core–mantle processes present in geodetic PM records. The resulting secular trends are 1.77 mas/yr in χ1 and − 2.49 mas/yr in χ2, differing markedly from commonly used GIA model predictions (e.g. 1.01 and − 4.67 mas/yr from Peltier et al., 2018). We validate our estimate by reconstructing the PM excitation for 1992–2016 using a comprehensive set of cryospheric, hydrologic, atmospheric, and oceanic contributions together with our inferred secular term; the reconstruction agrees closely with geodetic observations. Compared with the previous PM-based approach of Seo et al. (2021), our method yields a substantially smaller 95 per cent confidence region, providing an improved characterization of the long-term secular contribution to PM and a stronger foundation for using multi-decadal PM observations to diagnose twentieth-century ice mass and terrestrial water storage changes.

SummaryIndustrial activities in active foreland fold-and-thrust belts complicate conventional seismic hazard assessment, yet how tectonic deformation and industrial perturbations affect fault stability remains poorly understood. The Hutubi underground gas storage facility, situated within the Hutubi anticline at the deformation front of northern Tian Shan, has experienced a marked increase in seismicity since the start of gas injection and extraction. Here, we integrate seismic-reflection interpretation, growth strata analysis, geomorphic surface survey and optically stimulated luminescence dating, and area-depth-strain analysis to reconstruct the structure and growth history of the Hutubi anticline and quantify its late Quaternary shortening rate. The Hutubi anticline is a detachment fold growing above an 8–9 km‐deep detachment. The shortening and uplift rates of anticline have undergone a marked acceleration, rising from long-term averages of ∼0.06 and ∼0.03 mm/yr initiated in the Pliocene to ∼1.1 and ∼1.3 mm/yr since ∼0.4 Ma, respectively. This acceleration, consistent with other foreland structures, reflects ongoing basinward propagation of the Tian Shan deformation front in the late Quaternary. Within this tectonic setting, we further modeled the Coulomb stress evolution on secondary faults within the anticline driven by tectonic deformation. Our results indicate that tectonic deformation drives secondary faults to the seismic triggering threshold over a decade, and governs where ruptures nucleate and how they propagate, consistent with relocated earthquake observations in previous studies. Given sustained late Quaternary activity and pronounced injection-extraction perturbations, the potential for moderate-to-strong earthquakes (∼Mw5.3–7.0) on these faults cannot be ignored in the future. Thus, in active tectonic regions, seismic hazard assessments must incorporate deformation rates to accurately evaluate tectonic deformation and industrial perturbation effects on fault stability.

SummaryWhile well-log data provide direct measurements of subsurface properties essential for building lithofacies models, their sparse distribution and limited availability in exploration areas introduce significant uncertainty. This study presents an approach that integrates seismic data, rock physics, machine learning, and geostatistics to construct a regional-scale three-dimensional (3D) lithofacies model in the Jeju Basin, located offshore the southern Korean Peninsula. The proposed workflow comprises three integrated steps: performing Simultaneous Pre-Stack inversion on processed seismic data to derive rock properties, training neural networks for lithofacies prediction, and expanding the two-dimensional (2D) predictions into a 3D lithofacies model using Sequential Indicator Simulation. The prediction networks reproduced the interpretation-derived lithofacies labels with reasonable accuracy for the primary sedimentary classes, as demonstrated by quantitative evaluation metrics. The resulting 3D lithofacies model captured key geological features indicated by 2D seismic data, including faults and sedimentary structures, exhibiting improved spatial continuity and detail compared to conventional well-log-only geostatistical simulation. Comparison with a borehole outside the training volume demonstrated reasonable agreement between the predicted and well-log-derived lithofacies. This observation preliminarily implies the workflow’s applicability to field data beyond the 3D seismic coverage, though this assessment is based on a single well location. This integrated approach offers a cost-effective solution for lithofacies characterisation in frontier exploration sites where regional 3D seismic surveys are economically unfeasible and well control is limited.

SummarySeismic inversion plays a critical role in oil and gas exploration and CO2 storage monitoring by providing accurate subsurface structural imaging and quantitative characterization of rock properties, which offer valuable insights into reservoir characteristics, fluid flow, and storage integrity. However, traditional model-driven approaches are computationally expensive and heavily reliant on accurate initial models, which limits their applicability in complex geological settings. Furthermore, the need for manual hyperparameter tuning and expert intervention impedes automation and reduces their suitability for time-sensitive tasks such as real-time monitoring. Recent advancements in deep learning have demonstrated significant potential for seismic inversion by enabling neural networks to capture complex, nonlinear relationships between seismic data and subsurface properties. However, current data-driven methods still face major challenges, including limited generalization across diverse geological settings, poor interpretability due to the black-box nature of neural networks, and inadequate control over resolution and inversion fidelity. To address these limitations, we propose an implicit seismic inversion (ISI) approach that models subsurface physical parameters as a continuous function, implicitly represented by a deep neural network. Specifically, this study focuses on both post-stack impedance inversion, which recovers acoustic impedance as the key physical property, and pre-stack AVO inversion, which simultaneously inverts for P-wave velocity, S-wave velocity, and density. In this way, ISI leverages the learning dynamics of neural network optimization to reduce dependence on prior information, while enhancing interpretability and generalization ability through physics-informed constraints. Moreover, the continuous nature of the representation can significantly reduce the computational burden without compromising inversion accuracy. Experimental results on both synthetic and field datasets demonstrate that the proposed ISI method surpasses existing approaches in terms of accuracy, convergence, and independence from prior information. These advantages position ISI as a promising framework for scalable, automated, and real-time seismic inversion in next-generation geophysical workflows.

SummaryGeomagnetic data assimilation provides a framework for estimating the Earth’s core dynamics by combining geomagnetic observations with numerical geodynamo models. While the ensemble Kalman filter (EnKF) has been widely used as a sequential assimilation method, whether the ensemble Kalman smoother (EnKS), which additionally exploits future observations, improves estimates of the core state in fully nonlinear geodynamo models remains unclear. Here we compare the EnKS with the EnKF using synthetic observation experiments in three-dimensional geodynamo simulations under identical localization and inflation strategies. Under perfect-model conditions, the EnKS converges more rapidly and recovers additional flow modes, yielding more accurate estimates of both the internal velocity field and the core-surface flow. In contrast, the EnKS exhibits filter divergence in the magnetic field during the late stages of assimilation, reflecting its greater sensitivity to sampling errors. We then examine the influence of model error by introducing viscosity differences between the truth and assimilation models while maintaining otherwise nearly identical dynamics. Viscosity-induced model error degrades the EnKS, leading to reduced accuracy in recovering the dominant velocity modes and renewed filter divergence, although the large-scale core-surface flow remains well recovered. These results demonstrate that the EnKS can outperform the EnKF when the ensemble adequately represents the forecast-error covariance, but its more extensive use of observational information also makes it more susceptible to sampling and model errors. Improving the description of error covariances is therefore essential for realizing the full potential of EnKS-based geomagnetic data assimilation.

SummaryPassing trains act as correlated sources in both space and time, generating powerful waves that illuminate the near-surface below critical infrastructure. We propose an automated workflow for surface wave imaging below railway tracks that uses passing trains as the seismic source, in a context where train signals are far from an ideal isotropic source distribution. We exploit a dense 2D seismic array situated a few meters from the tracks for 3D imaging. To accommodate the complex moving source, we retain only correlations that form a limited angle with the moving train. This extends existing methods for a single line of receivers, either parallel or perpendicular to the tracks. We split the data into two cases, where the train is either far from or close to the acquisition setup. When the train is far from the array, sensor pairs mainly parallel to the tracks are important, while sensor pairs perpendicular to the tracks are decisive otherwise. In each case, the signals are time-windowed before computing the cross-correlations. Local dispersion panels are obtained through a Matched Field Processing algorithm. The dispersion curves are computed for various subsets of the array and referred to as patches. Extracting the dispersion curve of the surface wave below many patches results in a three-dimensional model of the Rayleigh wave phase velocity of the near-surface. Analyzing correlations for trains both near and far from the array yields broadband Rayleigh wave dispersion curves with high-frequency content (above 70 Hz), which is uncommon in passive methods. The velocity model is robust for wavelengths as short as 3 m and sensitive to soils at around 1 meter depth, allowing to image what may be a thicker ballast layer formed by previous track settlement issues.

SummaryConvection in Earth’s liquid outer core is powered by a combination of thermal and compositional buoyancy released at the inner-core boundary (ICB). The interaction between these two buoyancy sources, known as thermochemical convection, governs the strength of core flow and ultimately controls magnetic field generation. We present three-dimensional simulations of rotating convection in spherical shells that explicitly incorporate this coupling through a flux condition linking temperature (T) and composition (χ) at the ICB, expressed as $\partial \chi / \partial r = \mathcal {C}\, \partial T / \partial r$, where $\mathcal {C}$ is the coupling coefficient and r is the radial coordinate. By systematically varying $\mathcal {C}$ and the aspect ratio ζ = ri/ro (with ri and ro denoting the inner and outer radii respectively), two distinct flow regimes are identified: a viscous-dominated regime at weak coupling, primarily driven by thermal buoyancy, and an inertia-dominated regime at strong coupling, driven by compositional variations. Extrapolation of these dynamical regimes from the parameter values explored here to Earth’s actual Ekman number requires caution. We find that an effective coupling parameter $\mathcal {C}_{\mathrm{eff}} = \mathcal {C}\, \Lambda \, [3\zeta ^2(1-\zeta )/(1-\zeta ^3)]$, which combines the effects of buoyancy flux coupling and geometry through the buoyancy ratio Λ (ratio of compositional to thermal forcing across the domain), collapses the kinetic-energy $(\mathcal {K})$ of all models, which span a range of radius ratios, into universal scaling relationships: $\mathcal {K} \propto \mathcal {C}_{\mathrm{eff}}$ in the viscous regime and $\mathcal {K} \propto \mathcal {C}_{\mathrm{eff}}^{2/3}$ in the inertial regime. Diagnostics of the local Rossby number and of the momentum budget corroborate a transition in the dominant force balance when $\mathcal {C}_{\mathrm{eff}}$ exceeds unity. These results suggest that the geodynamo transitioned from thermally to compositionally dominated convection soon after the onset of inner-core growth (0.5–1.1 Gyr ago): for the midline material parameters the $\mathcal {C}_{\rm eff}=1$ crossing occurs within ~10–20 Myr of nucleation (inner-core radius ~85 km), with the full uncertainty band depending primarily on the poorly constrained effective compositional diffusivity. This transition may be apparent in the paleomagnetic record. The derived scaling laws establish a quantitative link between inner-core evolution, ICB thermochemical coupling, and the vigor of core convection, providing physically grounded benchmarks for models of planetary dynamos and core evolution.

SummaryThe central United States has experienced a surge in induced seismicity linked to wastewater injection, with Oklahoma as a key study area. While the broader effects of large-scale injection are well documented, the relationship between local injection activities and short-term seismicity fluctuations is less understood. Using deep learning and dense seismic monitoring, we developed an enhanced earthquake catalog and identified four persistent, low-magnitude (ML= – 0.69 to 2.83) seismic clusters within an active saltwater-disposal region. These clusters, several located near mapped faults, exhibit near-daily activity and provide case studies for examining local injection–seismicity dynamics. We find that injection histories for wells within approximately 8–10 km of the clusters generally exhibit stronger temporal correspondence with seismicity than those aggregated over larger spatial scales. Analysis of individual wells further identifies injection histories exhibiting comparatively strong short-term temporal correspondence with observed seismicity, independent of average injection rate or well distance. These relationships are interpreted as statistical associations within a pre-existing regional stress and pore-pressure state rather than evidence of direct hydraulic causality. Compared with higher-moment induced-seismicity cases, the detected clusters exhibit comparable patterns of cumulative seismic moment growth but at predominantly lower magnitudes. These low-moment sequences demonstrate how deep learning–augmented, high-resolution earthquake catalogs can reveal short-term variations in seismic response that are poorly resolved by conventional regional catalogs. Our results suggest that temporal similarity metrics may complement established injection-rate and distance criteria by identifying individual wells for further evaluation in induced-seismicity monitoring and mitigation strategies.

SummaryIn August 2024, Global Navigation Satellite System (GNSS) stations observed changes in the displacement pattern of the Santorini Volcanic Complex, accompanied by microseismic activity, showing a potential sign of unrest. Here, we analyze and discuss surface deformation patterns related to the Santorini caldera unrest, spanning from summer 2024 to January 2025. Synthetic Aperture Radar (SAR) interferometry was performed exploiting the Persistent Scatterer Interferometry (PSI) technique. We used both satellite geometries of the Sentinel-1 mission, producing line-of-sight (LOS) deformation maps and displacement decomposition in vertical and horizontal components. The observed displacements were combined with measurements from seven GNSS stations across the Santorini Volcanic Complex. The results highlight horizontal movements and uplift up to 70 mm/yr with a radial inflation pattern centered around the Kameni Islands, reminiscent of the 2011-2012 inflation episode. The geophysical model for the 2024-2025 period suggests a positive volume rate change at 3.15 km depth. Compared to the 2011-2012 unrest, the volume rate change is smaller; potentially a mix of magma and gas may be causing pressure along Kameni’s volcano-tectonic line, this time with a shallower source than the previous unrest. We address this particular period as the ”first phase” of the 2024-2025 unrest, since subsequent GNSS observations reveal a shift in the deformation, indicating the initialization of a different displacement pattern.

SummaryAccurate source localization is critical for monitoring volcanic tremors, microseismic events, and earthquakes. Compared to other source localization algorithms, waveform-based interferometric imaging exhibits superior robustness. It proves particularly effective when confronting weak signal phase signatures, noise interference, and inaccurate velocity models. Despite shared methodological commonalities, high-resolution imaging and localization of volcanic tremors prove notably more challenging than other passive seismic tasks. This difficulty primarily arises from sparse monitoring networks, complex wavefield interactions with the subsurface, and the indistinct phase signatures inherent in tremor signals. Consequently, waveform-based interferometric imaging tends to suffer from more severe blurring effects (i.e., poorer spatial resolution) when applied to volcanic tremor localization and compared against other application fields. To address these resolution limitations, we propose a maximum likelihood estimation (MLE)-based iterative peeling interferometric imaging method. This approach synergizes the inherent noise robustness of interferometry with a rigorous iterative framework designed to deblur the source image. By constructing data-driven empirical wavelets and employing an adaptive whitening filter dynamically updated via the Sherman-Morrison formula, the method sequentially strips source responses from the residual wavefield. This adaptive procedure directly counters the energy smearing caused by sparse arrays, effectively suppressing sidelobe interference to achieve a highly focused source resolution. Synthetic tests based on the Katla scenario demonstrate that the proposed method yields significantly higher spatial resolution than existing techniques. Furthermore, it exhibits robust performance even under adverse array geometries, concurrent multi-source interference, and velocity model inaccuracies. Field application to the 2011 Katla data further validates this by recovering source locations consistent with known ice cauldrons.

SummaryEarthquake early warning requires reliable uncertainty estimates from short initial waveforms. We evaluate Split conformal prediction (CP) and conformalized quantile regression (CQR) for single-station peak ground acceleration using event-partitioned K-NET records. Each 3 s waveform is normalized by component while retaining amplitude scales measured before normalization. The convolutional neural network (CNN)–Transformer that retains amplitude information is more accurate and yields narrower intervals than the waveform-only ablation. At the nominal 90 per cent level, CQR attains 90.03 per cent coverage; raw quantile regression and Split CP are also near nominal, whereas heteroscedastic Deep Ensemble and MC-Dropout baselines are more conservative and wider. Event-cluster bootstrap and repeated event-level resplits support marginal-coverage stability, although performance degrades for rare large earthquakes. Upper-bound screening trades increased recall for more false alarms, so the intervals are an uncertainty-quantification and screening layer rather than a standalone alert system. Under exact exchangeability, conformal prediction supplies finite-sample marginal coverage; record-level exchangeability is not established for this event-clustered sample, and event-conditional reliability is not guaranteed.

SummaryInductive electromagnetic (EM) methods induce currents in the ground with a changing magnetic field associated with time-varying current flowing in a transmitter loop. Geophysicists use measurements of the secondary magnetic field to estimate the conductivity structure of the ground. These estimations often assume that conductivity is constant, but in some cases, the conductivity can be dispersive and vary as a function of frequency. In this case, the dispersion is called Induced Polarisation (IP), and with an inductive (EM) source is called Inductive Induced Polarization abbreviated as IIP. IP and IIP phenomena are usually caused by electrochemical diffusion associated with small capacitor-like components in the ground. These components may be metallic particles, graphite, clay or local restrictions in current flow paths. Capacitor-like elements result in induced electrical polarisation or charge separation associated with current flow. When the injected or induced currents have turned off or decayed, the separated charges return to their original equilibrium state. This results in a net flow of current, called polarisation current, in the opposite direction to the original secondary current flow. The most common way of characterising IP and IIP is to estimate the parameters of a mathematical model that fits the measured data. The Cole-Cole model is the most common. Reversed or return polarisation current flow manifests itself in time-domain (transient) EM data as negative amplitudes or as positive decays that decay faster at late-time decays than is possible from a non-polarizable source. Depending on whether excitation is from a current or a voltage source, observed decays may have time constants that differ by orders of magnitude. For inductive sources, characterisation is best achieved by fitting a model that is a perturbation from the high-frequency conductivity rather than a perturbation from a low-frequency resistivity, with the associated Cole-Cole conductivity time-constant most directly associated with grain size. Sign reversals or negative responses from return current flow have been recorded since the 1970s or earlier in ground central-loop EM data, and in airborne EM data since the 1990s. Prior to the development of inversion methods, IIP negative transients were often simply removed from field data, leading to inaccurate conductivity models and interpretation. Since the turn of the century, modelling and inversion of Cole-Cole parameters have been common and applied to measured ground, semi-airborne and airborne transient EM data to parameterise any IIP effects. In order to explain the distortion of decays by IIP phenomena, IP-incorporated modelling and inversion of both the source and its host materials are critical for valid characterisation and interpretation. However, overparameterization and the limited bandwidth can lead to model uncertainty and ambiguous results. This is particularly the case if large-scale 3D inversion is attempted. We review IIP field examples for minerals of economic interest that may have IP associations, including disseminated sulphides, gold-associated alteration and kimberlite pipes under permafrost conditions. Of importance in near-surface, hydrogeological and contaminant mapping are other polarizable sources, and IIP has been applied to mapping clays, hydrocarbons and permafrost.

SummaryWedge subduction represents the final stage of the oceanic plates before terrane-continent collision. However, its influence on collision dynamics is often overlooked. The closure of the Paleo-Tethys Ocean, during which a mid-ocean ridge nearly parallel to the trench was subducted before continental collision, makes it an ideal natural laboratory for studying collision modes. Based on the tectonic background of Paleo-Tethys Ocean closure, this study systematically investigates the effects of wedge subduction on dynamic modes of terrane-continent collision through a finite-element method. The results show that low initial wedge subduction angles facilitate the terrane-continent collision (TCC) mode, whereas steep initial angles promote the continental-terrane subduction (CTS) mode. A slow convergence rate (1 cm/yr) increases the likelihood of the CTS mode, whereas faster convergence tends to promote the TCC mode. The Cimmeria-directed opposite-polarity configuration may cause overthrusting of the trailing oceanic plate onto the continental terrane and promote the TCC mode. Continental terrane strength is another critical factor. Weaker terranes favour the CTS mode, while stronger ones favour the TCC mode. Geological evidence and tomography results are broadly compatible with the tentative inference that the western and central segments of the Paleo-Tethys suture zone were more likely to experience the CTS mode, whereas the eastern segment may have been more likely to experience the TCC mode. These east–west differences may be explained by the modelled effects of convergence rate, depending on factors such as continental terrane strength and wedge subduction angle. This research establishes relationships between wedge subduction and continental collision modes, providing new perspectives for understanding structural diversity within suture zones globally.

SummaryIn the last thirty years, the Northern and Central Apennines (Italy) have been affected by three main destructive seismic sequences: the 1997 Colfiorito (three events ML > 5.5), the 2009 L’Aquila (one event ML > 5.5), and the 2016–2017 Amatrice-Visso-Norcia (three events ML > 5.5). Several studies have analysed the spatial-temporal evolution and processes driving each sequence, focused more on the foreshock-mainshock-aftershocks dynamics. Here, we focus on the 2018–2024 interseismic period, aiming to unravel the long-term seismogenic behaviour of this region. We first relocated the earthquake catalogue and identified clusters through a declustering algorithm. During this phase, background seismicity and most of the clusters were arranged in a 2-3 km thick low- angle layer. We found that (i) most clusters were driven by an aseismic process, (ii) the depth of both clusters and seismicity layer increased toward the southeast, (iii) the volume of clusters decreases to the southeast, (iv) the low-angle layer almost disappeared in the L’Aquila area. Comparing two interseismic phases (2011–2016 and 2018–2024), we found striking similarities with events occurring on same sites and characterized by both foreshock-mainshock-aftershocks and swarm-like behaviour. In addition to this, the L’Aquila area was seismically more “silent” compared to the norther sites during both interseismic phases. We propose that these different long-term seismogenic behaviours may reflect variation in the structure and rheology of the upper crust moving from the Northern to the Central Apennines. This highlights the important role of the structural inheritance in controlling how the active deformation can affect the interseismic period.