Abstract Melt re-injection after a giant caldera eruption was quantitatively investigated for the Kikai Caldera Volcano in Japan, which erupted 7300 years ago (Kikai-Akahoya eruption). Our seismic refraction survey revealed a low-velocity anomaly directly beneath the Kikai Caldera Volcano, indicating the existence of a large magma reservoir at a shallow depth of 2.5–6 km. The reservoir can be approximated by a trapezoidal shape in this 2D section, with its width being at least the same as the width of the inner caldera, and its melt fraction was estimated as 3–6%, but could be limited to 10% at most. We propose a melt re-injection model in which new melt is re-injected into this large magma reservoir at the shallow depth just beneath the caldera, which is the same magma reservoir for the Kikai-Akahoya eruption. This model may demonstrate a common feature of volcanoes that have experienced a giant caldera eruption.

The 2024 Noto Peninsula Earthquake (MJ 7.6) was a major crustal event with a rupture zone extending from the Noto Peninsula to the offshore region. To clarify the relationship between the seismogenic fault and the deep offshore crustal structure, we conducted a marine seismic survey across the eastern aftershock zone in August and September 2024, using a multichannel seismic reflection (MCS) system and ocean bottom seismographs (OBSs). The crustal structure directly beneath the survey line was imaged by deriving a velocity model from the acquired OBS data and applying it to prestack depth migration (PSDM) processing of the MCS data. In the sedimentary sequence, an asymmetric anticline associated with a northwest-dipping reverse fault was identified, and the location of this fault spatially corresponds to the aftershock distribution determined from OBS observations. In the crust, although no prominent reflectors were observed within the primary aftershock zone at depths of 5–15 km, three distinct reflectors (R1, R2, and R3) were detected beneath the aftershock zone at depths of 15–33 km and are interpreted to originate from the lower crust, the Moho, and the uppermost mantle, respectively. Notably, the shallowest of these deep reflectors (R1) is located at a depth of approximately 15 km, coinciding with the lower limit of the aftershock activity. This spatial correlation points to a close link between the deep crustal structures and the seismogenesis of the 2024 Noto Peninsula Earthquake.

Accurate 3D seismic velocity modeling through seismic travel-time tomography using both active- and passive-source data provides critical underpinning models for seismicity monitoring and hazard assessment. Because travel-time tomography is an inherently ill-posed inverse problem, UQ of the estimated models using Bayesian methods is also important for reliable downstream interpretations and analyses. However, Bayesian inference for 3D tomography based on conventional grid-based representations faces the “curse of dimensionality” and severe computational bottlenecks. Consequently, rigorous Bayesian UQ for margin-wide 3D travel-time tomography has remained largely unexplored. In this study, we propose a meshless 3D Bayesian travel-time tomography method that combines PINNs with a neural representation of the velocity structure, enabling tractable and data-efficient Bayesian inference through function-space particle-based variational inference. To efficiently integrate passive-source data into the Bayesian estimation of the velocity structure, we conduct analytical marginalization treating uncertain source parameters as nuisance parameters, with passive-source relocation carried out in post-processing. We validated the capability of our approach for 3D problems through synthetic experiments. Furthermore, we applied the method to a real-world dataset from marine active-source surveys and natural earthquakes off the Kii Peninsula, Nankai Trough. Our probabilistic 3D ensemble successfully resolves key geological features and provides data-consistent uncertainty maps. The posterior mean hypocenters shifted mainly in the vertical direction by 10-15 km, consistent with a previous relocation result. Finally, the neural representation drastically reduces storage requirements for the entire ensemble velocity model, highlighting the scalability and data efficiency of the proposed framework.

On January 1, 2024, a large-magnitude (M7.6) reverse-fault earthquake struck the Noto Peninsula, Japan, generating a tsunami that caused extensive coastal damage. The source fault, inferred from aftershock distributions by the Japan Meteorological Agency, extends approximately 150 km from the peninsula’s western coast to the northeastern offshore region. High-resolution pre-stack depth migrated seismic profiles reveal a large deformation zone (LDZ) within the 2024 coseismic rupture area. This ~ 2.5–3.8 km wide and ~ 30 km long LDZ consists of steeply southeast-dipping reverse faults (~ 50°–75°), which may represent shallow extensions of the deeper seismogenic listric fault, as well as branching faults suggestive of a local strike-slip component. Numerical simulations of tsunami generation suggest that a coseismic slip of 6–7 m along the reverse fault, followed by reactivation of the LDZ, could have produced up to 3 m of seafloor uplift and triggered subsequent tsunamis. In addition, we identified northwest-dipping reverse faults (~ 50°–55°) that appear to be active but exhibited little coseismic slip during the 2024 event. Our findings provide the first detailed characterization of tsunamigenic fault structures within the rupture area of the 2024 Noto earthquake.

The 2024 Noto Peninsula earthquake (M7.6), which occurred in central Japan on January 1, is thought to have ruptured multiple fault segments with different strike and dip angles. Seismicity along these segments remained active for over one year after the mainshock. To understand the fault zone in terms of fluid and frictional properties, this study estimates the stress drop of small- to moderate-sized earthquakes. We used seismic waveform data from onshore and offshore stations for approximately one month, beginning in mid-January 2024. The estimated stress drop was nearly scale-independent and gradually decreased from southwest to northeast across the offshore source area. The possible causes of this lateral variation were fault orientation, seismic velocity, pore pressure, low initial stress and frictional properties. Among these, fault orientation, seismic velocity and low initial stress were considered unlikely, indicating that frictional properties, potentially influenced by pore pressure, were the likely cause. This study showed that the lateral variation in stress drop is not due to differences in focal depth. Within the 4–10 km depth range, earthquakes on the northeastern segments exhibited significantly smaller stress drops than those on the southern segments. We also found a statistically significant increase in stress drop with increasing depth between 2 and 10 km, whereas no clear depth dependence was observed at depths greater than 10 km. The results of this study may contribute to investigations of frictional properties and fluid effects in the offshore source area.

High pore pressure ratio (λ*) has been associated with the occurrence of slow earthquakes. Many studies have estimated the λ* using taper angle, P- and S-wave velocities, and drilling but the extent is not yet clearly defined. This study utilized a recently published high-resolution P-wave velocity model off Muroto derived from a two-step tomographic inversion of ocean-bottom seismograph data to determine the λ* using empirical relationships between velocity, porosity, and effective mean stress. We determined an extensive zone of high λ* (> 0.4) from the frontal thrust up to ~ 60 km landward and to a depth of 8 km with three characteristic observations. First, the underthrust sediments in the outer wedge show patches of overpressured aquifers where λ*>0.6, consistent with previous drilling results. Second, the high λ* (> 0.6) region in the inner wedge coincides with previously reported underplated sediments composed of fluid-rich trench-fill sediments dragged down by seamounts. The high λ* may be caused by tectonic compression from the newly subducted seamounts. Lastly, vertical columns of high λ* were observed in many of the thrust faults with some faults showing a negative polarity, which are interpreted to be evidence of fluid flow.

Large earthquakes and tsunamis are major hazards at subduction zones, where segmentation governs rupture extent and magnitude. However, the mechanisms that define segment boundaries remain unclear. In the eastern Nankai Trough, which is anticipated to host future megathrust earthquakes, the forearc is divided into the Tonankai and Tokai segments. Here we use newly acquired 3D seismic reflection data (2021-2022), to image structures across the segment boundary. High-resolution profiles reveal three seafloor topographic highs associated with active folding and thrusting in forearc sediments, the accretionary wedge, and along the plate interface. These features trend NNE-NE in Tonankai but ENE in Tokai, across an arcuate syntaxis. Collision of the Paleo-Zenisu Ridge, a remnant of the Izu-Bonin arc on the Philippine Sea Plate, appears to impose the syntaxis structure of the upper plate. Our results demonstrate that the forearc structural complexity exerts a primary control on margin segmentation and associated hazards.

Abstract Seismic and aseismic slip has been reported in the Nankai Trough. In this study we apply ambient noise differential adjoint tomography to obtain a high‐resolution S‐wave velocity (Vs) model in the Nankai Trough. The Vs in the very shallow imbricate thrust zone is higher than that on both the seaward and landward sides. This region may prevent the upward migration of fluids from below, which may lead to an elevated pore fluid pressure beneath the region and facilitate slow earthquake occurrence. There is a large‐scale anomaly with low Vs and high P‐wave velocity (Vp)/Vs in the Kumano basin. This may be caused by the ascent of deep‐sourced fluids from the plate boundary and indicate that an impermeable condition is absent in the seismogenic zone. In subduction zones the permeability property above fluid‐source zones may be important for the strength of the plate interface and thus the types of earthquakes.

Recent seismological and geodetic observations have revealed shallow slow earthquakes on subduction megathrusts, offering an opportunity to explore the factors controlling megathrust slip. Identifying these controlling factors helps constrain the potential updip extent of future megathrust ruptures. Here, we investigate the factors controlling the spatial extent of shallow slow earthquakes by integrating tremor locations with uncertainty estimates, detailed bathymetry data, and multichannel seismic data for the 2020–2021 slow earthquake sequence in the Kumano-nada region of the Nankai Trough. Tremor probability maps reveal that the updip, downdip, and lateral limits of tremor activity coincide with distinct structural boundaries. These include the kink or branching of the décollement, a deep-seated strike-slip fault, and the inner–outer wedge boundary. These structures likely act as geometrical or mechanical barriers, further influenced by variations in pore-fluid pressure and by the subducted Paleo–Zenisu ridge. The results indicate that megathrust geometry, material properties of the overriding prism, fluid distribution, and ridge subduction jointly govern the spatial extent of slow earthquakes, emphasizing the need to account for these factors in assessing the potential extent of future megathrust ruptures. This study shows that slow earthquake slip in the Nankai Trough is confined by structural barriers, including fault geometry, material contrasts, and subducted ridges, revealing key controls on the potential extent of future megathrust ruptures.

Slow earthquakes are crucial for understanding the nature of subducting plate boundaries and their seismic coupling. Although many very low-frequency earthquakes (VLFEs) have been recently recorded in the central part of the Nankai Trough, there have been only rare cases reported between the Kii Channel and Cape Shionomisaki. To investigate the structural features and their potential relationship with the slow events, in November 2018, JAMSTEC deployed a dense array of 96 ocean-bottom seismometers (OBS) to acquire an active-source seismic refraction dataset from the seaward side of the subduction trough to the accretionary prism off Cape Shionomisaki. We applied travel time and waveform tomography to the OBS data and reconstructed the seismic P-wave velocity model down to the upper mantle. The subsurface velocity model reveals some new discovered features, and describes the accretionary prism as consisting of a weak outer wedge and a strong inner wedge. The velocity anomaly bodies inside the wedge possibly link to specific deformation processes and the evolution of the accretionary prism. As compared to a transect across the adjacent Kumano Basin, where there are densely distributed VLFEs, the distinct results from these two locations reveal potential structural and physical differences that cause variability in seismic activities.

The southern Kuril Trench subduction zone experienced a M9‐class megathrust earthquake in the 17th century, and another is highly probable within the next 30 years. This earthquake likely exhibited large coseismic slip at the shallower plate boundary fault, causing a devastating tsunami, while the deeper plate boundary experienced smaller slip, similar to the 2011 Tohoku‐oki earthquake. To investigate the structural factors behind the contrasting slip behaviors, we conducted a controlled‐source seismic survey across the Kuril Trench axis. We found that the boundary between the largely‐slipped‐shallow and small‐slipped‐deeper faults aligns with a Vp transition zone in the overriding plate, characterized by sharp landward Vp increase from the low‐Vp frontal wedge (0–30 km from the trench) to the high‐Vp island arc crust (>60 km), and associated with discontinuous near‐horizontal reflectors zone (RZ) above the plate boundary, exhibiting negative polarity. These findings suggest that the slip behavior boundary correlates with sudden trenchward rigidity reduction along the plate boundary, as found in other areas where seismic rupture extended to the trench. The negative‐polarity reflections in the RZ, together with the plate boundary, imply high‐pore‐pressure created by abundant fluid supply from accreted or underplated sediments consolidation and/or dehydration, and oceanic crustal crack cementation. This suggests a weak fault strength at the plate boundary beneath the transition zone, indicating weak mechanical coupling similar to the shallow fault in Tohoku‐oki. Similar to the Tohoku earthquake, a slip‐to‐the‐trench rupture may be triggered by releasing large strain energy accumulated along the deeper fault beneath the high‐rigidity crust.

The heterogeneous distributions of large and slow earthquakes in subduction zones are caused by multiple uncertain conditions in the source regions, and the basement topography is considered one of the major controlling factors. We revealed the topography of the subducting basement along the entire Nankai Trough on the basis of seismic reflection profiles compiled from 1997 to 2024. We interpreted the reflection profiles in the time domain to ensure consistency among multiple-generation datasets. The new surface model captured the detailed topography on a scale of several kilometers, over 730 km long and 150 km wide, and down to depths of 15–20 km. The basement topography is characterized by past tectonic activity and is divided into three domains, which affect the present heterogeneity in geological structures and physical properties along the Nankai Trough. While the three domains correspond to megathrust seismogenic zones along the Nankai Trough, the lack of correspondence between the significant topographic relief and the slow earthquake distribution in some areas suggests the need for additional factors controlling spatiotemporal variations in seismicity.

The Nankai Trough is a major subduction zone in southern Japan capable of generating the next M7 or larger earthquake off Kii Peninsula. In addition to earthquakes characterized by sharp P and S arrivals, there are numerous slow earthquakes without clear P phases at the subducting plate boundary corresponding to transient slips that are related to fluids at the plate boundary and high pore pressure. In this study, we perform ambient noise differential adjoint tomography to derive the S-wave velocity model beneath a linear Ocean Bottom Seismometer (OBS) array, which was previously difficult based on airgun active source study. We find that S-wave velocities in the southern (seaward) Kumano Basin are significantly lower than in the northern (landward) Kumano Basin, suggesting weak upper plate and abundant fluids in the southern Kumano Basin. The fluid-rich southern Kumano Basin lies above a weak interplate coupling zone with shallow slow earthquakes and pressurized fluids due to plate interface dehydration. Conversely, the northern Kumano Basin overlies a strong interplate coupling zone with few slow earthquakes, suggesting fluid migration from the weak-coupling interface to the forearc basin above.

The 2024 Noto-Hanto earthquake with a magnitude of 7.6 occurred in the Noto Peninsula on January 1, 2024. The mainshock had a reverse fault focal solution and direction of compression axis was the northwest–southeast. In the Noto Peninsula, earthquake swarms have been observed since December 2020. In contrast to this swarm activity, the mainshock had extending to the marine area. Therefore, we performed a rapid response seafloor seismic observation in the source region and its vicinity. We deployed 34 free-fall pop-up type ocean bottom seismometers (OBSs) in January 2024, recovered 26 short-period OBSs (SPOBSs) after a month. The arrival times of the P- and S-waves were manually read from the data of SPOBSs and land seismic stations based on the event list by a land seismic network. We relocated the hypocenters of the events by combination of the location programs using absolute travel times with station corrections and the double-difference method. A velocity model was derived from the velocity structure by the marine seismic survey. Focal mechanisms were estimated using the grid search method based on the polarities of the first P-wave arrivals. The aftershock depths mostly ranged from 0.2 km to 17 km. Although the aftershock activity seems to be confined in the upper crust, relatively deep events occurred in the easternmost source region. The aftershocks formed several dipping planes corresponding to the multiple faults described in the offshore active fault model constructed before the mainshock. The upper boundaries of the planes of the hypocenter distribution coincide with the upper edges of the modeled faults, and the lower boundary of the aftershock distribution also aligns well with the lower edges of the faults. This consistency indicates that the rupture at the mainshock propagated to faults with different geometries. Although 70 events a reverse fault focal mechanism similar to the mainshock, we identified 87 strike-slip events. Most of the events involving both reverse and strike-slip faults had P-axes perpendicular to the fault strike. This finding suggests that the aftershock activity was affected by a northwest–southeast compressional stress.

We investigate the potential fluid migration process in a plate subduction zone on the basis of integrated active-source seismic imaging from multichannel seismic (MCS) reflection and ocean-bottom seismograph (OBS) wide-angle seismic surveys along a 100 km-long line in the area offshore of Cape Muroto in the central Nankai Trough. First, to overcome poor reflection imaging in a thick part of the accretionary wedge and discrepancies between the results of conventional analyses from different datasets, we apply a signal enhancement technique in addition to standard preprocessing for the MCS reflection data and prestack depth migration based on a high-resolution seismic velocity model derived by full waveform inversion (FWI) using wide-angle OBS data. The improved MCS reflection profiles clarify the detailed geological architecture consistent with the seismic velocity variations with significant low-velocity zones in the accretionary wedge. A low-velocity band elongated immediately above the oceanic crust implies possible high pore fluid pressure in the underthrust sequence. A low-velocity column passing through the thick accretionary wedge could suggest the existence of gas-rich fluid in fault/fracture zones caused by subsequent deformation due to the oblique subduction of the Philippine Sea plate accompanied by multiple seamounts and topographic relief. We propose a schematic model of potential fluid migration in the central Nankai Trough subduction zone. The hemipelagic muddy sediments and turbidites containing organic carbon deposited in the Shikoku Basin were transported downward by underthrusting. High pore fluid pressure, which is caused by rapid loading and clay mineral dehydration within the underthrust sequence, may lead upward fluid migration through fault zones. Thermogenic methane generated at depth is a driving force of buoyancy for upwelling through permeable parts of fractures or dipping strata. In addition to biogenic methane, migrated thermogenic methane is also a source of widely distributed gas hydrates.

Subsurface seismic velocity structure is essential for earthquake source studies, including hypocenter determination. Conventional hypocenter determination methods ignore the inherent uncertainty in seismic velocity structure models, and the impact of this oversight has not been thoroughly investigated. Here, we address this issue by employing a physics-informed deep learning (PIDL) approach that quantifies uncertainty in two-dimensional seismic velocity structure modeling and its propagation to hypocenter determination by introducing neural network ensembles trained on active seismic survey data, earthquake observation data, and the physical equation of wavefront movement. An analysis of an earthquake in southwest Japan using our method revealed that accounting for such uncertainty propagation significantly reduced the bias and uncertainty underestimation in the hypocenter determination, enabling quantitative evaluation of the focal depth relative to the plate boundary. Our results highlight the potential of PIDL for various geophysical inverse problems, such as investigating earthquake source parameters, which inherently suffer from uncertainty propagation.

Abstract Mud volcanoes, common seafloor features in subduction zone forearcs, provide crucial insights into deep hydrogeological systems associated with plate subduction. Despite their widespread occurrence, the factors governing their uneven distribution remain unclear, and their internal physical properties are poorly understood. Using multiple seismic reflection profiles and high-resolution P-wave velocity (Vp) models, we examined the distribution and physical properties of mud diapirs in the Hyuga-nada subduction zone, southwest Japan, impacted by the Kyushu-Palau ridge seamount chain entering the junction of the Nankai Trough and the Ryukyu Trench. Our seismic reflection data revealed more than 60 diapiric structures, some breaching the seafloor, while others remain buried beneath shallow sediments. These diapirs predominantly occur where the subducting slab is 10–20 km deep, mainly on the leading edge of a large subducting seamount. Notably, there is a significant absence of diapirs directly above the seamount and within approximately 50 km of the trench axis. The findings suggest that once the slab reaches depths greater than 10 km, thermogenic hydrocarbon gas production, which gains significant buoyancy, drives upwelling fluid flows in the overriding plate. The ridge subduction also contributes to diapir distribution by enhancing dehydration of underthrust sediments on the downdip side and creating a porous fracture network in the upper plate. The Vp structure of an exposed mud volcano indicates that it contains methane levels similar to those in the Kumano-nada region of the central Nankai Trough. Buried mud diapirs, rooted in fluid- and gas-rich reservoirs at 1–5 km depths below the seafloor, exhibit significantly lower seismic velocities (Vp < 3.0 km/s), indicating higher fluid and hydrocarbon gas concentration at depth. These observations suggest the forearc wedge in the Hyuga-nada area contains more extensive fluid/gas reservoirs than previously recognized.

An increase in seismic activity was observed near the Izu-Torishima and Sofugan islands in the Izu-Ogasawara arc of Japan in early October 2023. Several M6-class earthquakes occurred from October 3 to 6, and tsunamis were recorded on October 5 and 6, associated with M6-class normal faulting earthquakes. On October 8, tsunamis were observed without any significant earthquakes but accompanied by distinct T-phase signals. Analyses of T-phases, tsunami waveforms, and bathymetry surveys suggested that the October 8 tsunamis were linked to the volcanic activity at the Sofu Seamount, located west of Sofugan Island. However, the precise location of the entire activity and its connection to the tsunamis on October 5 and 6 remain unclear. We conducted earthquake observations near the Sofu Seamount from November to December 2023 using ocean bottom seismographs (OBSs). These observations revealed that seismic activity extended approximately 50 km in the NNW direction, from the Sofu Seamount to the Torishima Rift, a back-arc rift between the Izu-Torishima and Sofugan Islands. Although it is difficult to constrain the hypocenter depths owing to the limited number of OBSs and their large separations, earthquakes in the Torishima Rift were likely located at depths of 15–20 km or even shallower, at approximately 5 km. The tsunami observed at Hachijo-jima Island on October 5 can be attributed to the rupture of a normal fault in the Torishima Rift, as inferred from OBS observations. Shallow normal faults and intrusions reaching the seafloor were previously imaged in the back-arc rift zone near the Torishima Rift. Moreover, the earthquake activity in early October showed a northward migration from the Sofu Seamount to the Torishima Rift, followed by a rapid decrease after the volcanic activity of the Sofu Seamount, which was accompanied by tsunamis on October 8. The results suggest that the seismic activity in early October 2023, accompanied by tsunamis, is likely related not only to the volcanic activity at the Sofu Seamount but also to the tectono-magmatic activity in the Torishima Rift.

The Lord Howe Rise (LHR), a submerged continental fragment of Zealandia offshore eastern Australia, provides a rare opportunity to investigate the fine‐scale structure of a magma‐poor rifted margin. This study applies full waveform inversion (FWI) of 2D marine wide‐angle seismic data acquired with densely deployed ocean‐bottom seismometers with 800 m spacing, generating a high‐resolution P‐wave velocity model across the central LHR. The inversion reveals localized low‐velocity zones (LVZs) within fault‐bounded sedimentary depocenters, and a prominent low‐angle reflector interpreted as a possible décollement at greater depths. Unlike previous studies in this region, which primarily relied on conventional tomography and reflection data, our results resolve subtle velocity anomalies linked to fluid migration, tectonic fracturing, and diagenetic alteration. Low‐velocity zones exhibit reductions from ∼2.5 km/s to ∼2.1 km/s between ∼2.4 and 2.7 km depth, coinciding with enhanced porosity and opal‐CT precipitation. The possible décollement, observed at ∼6–8 km depth, dips gently and marks a mechanical decoupling horizon consistent with asymmetric rifting. These findings reveal previously unresolved features of post‐rift diagenesis, fluid pathways, and fault‐related strain partitioning within the LHR, offering new insights on the tectonic evolution of Zealandia. When placed within the broader tectonic evolution of Zealandia, the study emphasizes post‐rift overprints as primary controls on the observed seismic anomalies and demonstrates how imaged velocity anomalies can constrain fluid–rock interactions and basin evolution during continental breakup.