Abstract Soils undergo wetting‐drying and freezing‐thawing cycles, yet their water retention and freezing behaviors are typically described within separate frameworks. This separation raises a fundamental question: do desaturation and freezing share comparable pore‐structure controls? A fractal formulation for the Soil Freezing Characteristic Curve (SFCC) is developed from a fractal pore‐size distribution in which characterizes pore‐size scaling. The proposed model provides good fits to 129 experimental SFCC data sets over wide temperature ranges. Combined with an existing fractal description of the Soil Water Characteristic Curve (SWCC), paired SWCC‐SFCC measurements for 14 materials show that the fractal dimensions, , inferred from the two curves are generally comparable. These results support a pore‐structure‐based connection between desaturation and freezing and provide a basis for describing water retention and freezing behavior within a common structural framework.
Abstract Subglacial sedimentary basins in Antarctica are hypothesized to modulate ice flow and biogeochemical cycles via groundwater and geothermal feedbacks, yet their properties remain poorly constrained. Here, we present a joint quantitative analysis of new magnetotelluric (MT) data acquired at Thwaites Glacier (TG) and WAIS Divide, alongside legacy data sets from Whillans Ice Stream, Ross Ice Shelf, and South Pole. The new MT data reveal a shallow 1‐D crustal structure, around the ice bed interface, at TG, overlying deeper 3‐D structures. Our constrained 1‐D transdimensional Bayesian inversion of the upper crust highlights that the sedimentary basin beneath TG exhibits relatively high resistivity (>10 Ωm), distinct from the lower‐resistivity (<10 Ωm) basins at other sites. Sensitivity analysis reveals that the TG basin is horizontally heterogeneous, with conductive signatures in thicker sections and resistive signatures at GHOST Ridge, a subglacial topographic high which has been identified as a potential future stabilizing point. Conversely, basins beneath Subglacial Lake Whillans and the South Pole exhibit vertical stratification, with relatively resistive upper layers, up to 600 m thick, above conductive deeper layers (<5 Ωm). We hypothesize that complex, spatially variable groundwater regimes are widespread in Antarctica. These contrasting hydrological environments imply continent‐scale variability in subglacial thermodynamics and possible modulation of inherent ice dynamics.
Abstract The Kashmir “seismic gap” in NW Himalaya has accumulated sufficient elastic strain to cause a great earthquake. Constraining the lateral and depth‐dependent variations in seismic attenuation across this region is essential for ground‐shaking estimation from earthquakes. Recently acquired waveform data from precisely‐located‐local earthquakes are used to compute: (a) 3D S‐wave attenuation using coda‐normalization; (b) 3D intrinsic absorption using coda quality‐factor; and (c) 2D scattering from peak delay‐time measurements. Our results, interpreted in light of the structure and seismicity, reveal frequency‐dependent variations in attenuation, controlled by the crustal structure, lithology, geometry of major thrusts and active faults. High attenuation/absorption characterize the sedimentary/meta‐sedimentary rocks of the Sub‐Himalaya and the Lesser Himalayan Duplex in the Kishtwar Window (KW). Enhanced attenuation and scattering within the KW and its surroundings are attributed to concentrated seismicity associated with the frontal ramp of the Main Himalayan Thrust (MHT). In contrast, the reentrants of the Main Frontal Thrust and Reasi Thrust, representing surface expressions of MHT lateral‐ramp, exhibit relatively low attenuation/absorption. This is possibly due to shallowing of the underthrust Indian crust across the lateral ramp. The Higher Himalayan crystallines show low attenuation and weak scattering. The Kashmir Valley shows lateral contrasts in attenuation/absorption, indicating varying sedimentary thicknesses. The MHT demarcates the contrast between the high absorption/attenuation Himalayan Wedge above and the low absorption/attenuation underthrusting Indian crust below. The Indian upper‐crust shows high attenuation in patches, possibly related to fluids from metamorphic dehydration reactions or inherited heterogeneity.
Abstract Corner frequency ( f c ) and seismic moment ( M 0 ) are key parameters derived from seismic signals that are used to characterize earthquake stress drop, rupture area, and slip. These parameters are also affected by fault geometry and boundary conditions. However, the systematic study of these effects in laboratory settings has been challenging. This study presents laboratory earthquake experiments that examine how rupture dynamics are influenced by (a) the aspect ratio of rectangular PMMA velocity‐weakening (VW) asperities surrounded by the Teflon velocity‐strengthening (VS) patches, and (b) whether the sides of a VW asperity are confined with VS patches or are free surfaces. We found that increasing confinement by reducing free surfaces or increasing the VW asperity aspect ratio stabilizes fault slip, so that higher normal stress is required to transition from aseismic to seismic slip. Increased confinement and high aspect ratios also reduced M 0 and increased f c , both of which were determined from the radiated seismic waves. M 0 and f c were primarily controlled by the shorter dimension of the VW asperity. Analysis of high‐frequency acoustic emission signals revealed that ruptures on high‐aspect‐ratio VW asperities propagated more unidirectionally, whereas ruptures on square VW asperities were more complex. Further, the high‐aspect‐ratio asperities were more likely to be eroded by surrounding VS regions while low‐aspect‐ratio asperities were more likely to rupture into the VS surroundings. These results demonstrate that both the confinement from surrounding stable areas and the geometry of the seismogenic patch can affect rupture nucleation, propagation, and seismic source characteristics.
Abstract Seismic studies have suggested widespread thickening of the mantle transition zone (MTZ) beneath the Tibetan Plateau, which is commonly attributed to the foundering of cold lithosphere. However, the spatial pattern of MTZ thickening does not consistently align with the distribution of high‐velocity anomalies, suggesting that other factors, such as hydration, also influence MTZ structure. In this study, we employ a dense broadband seismic array combined with existing regional deployments to obtain improved receiver function images of the MTZ beneath Tibet. After correcting for upper mantle heterogeneity using a recent high resolution tomographic model, we combine MTZ thickness anomalies with mean S‐wave velocity perturbations to quantify lateral variations in temperature and water content. Our results show that the MTZ beneath the plateau contains on average at least 0.31 wt.% H 2 O and is underlain by a low‐velocity layer (LVL) immediately above the 410‐km discontinuity. These features exhibit a pronounced north‐south asymmetry: the northern MTZ is relative warm and water‐rich, whereas the southern MTZ is cooler and drier but overlain by a more extensive LVL. We propose that piece of cold lithosphere sank into the southern MTZ has displaced water‐rich mantle material from the MTZ into the upper mantle, leading to dehydration above the MTZ, formation of the LVL, and giving rise to a water‐rich mantle upwelling. These processes likely played a key role in generating Miocene porphyry copper deposits in the southern Tibetan Plateau.
Abstract Earthquake source parameters are crucial for assessing seismic hazard and understanding earthquake source physics. However, source parameter estimation is challenging for small earthquakes because they suffer notorious trade‐offs with shallow attenuation. Borehole Distributed Acoustic Sensing (DAS) uses fiber‐optic cables to provide unprecedented spatial resolution for attenuation analysis and a closer recording of the earthquake source. In this study, we use DAS in a vertical well in the Cape Modern geothermal field, near Milford, Utah, to characterize shallow attenuation and estimate source parameters of microearthquakes that occurred during a stimulation. We directly correct the DAS instrument response and directional sensitivity. We use along‐depth spectral ratios to obtain a high‐resolution Quality Factor ( Q ) profile for P waves in the top 2,500 m. We find low Q at the near surface, and that attenuation primarily decreases with depth, with less dependence on rock types, consistent with previous studies in California and Switzerland. We use a single‐spectral approach to estimate corner frequency, source radius, slip, magnitude, and spectral stress drop over the limited magnitude range. The spectral stress drops do not show magnitude dependence, while their variability might relate to fault heterogeneity or rupture variability. The DAS response‐correction is channel‐dependent and can introduce biases in source parameters. We reduce potential biases by using channels with similar responses and checking measurement consistency. We use the Empirical Green's Function for two event pairs to benchmark the results. This work shows the potential of using DAS for understanding crustal attenuation and characterizing microearthquakes.
Abstract The anatomy of orogens is difficult to decipher as structural information is usually limited to surface and seismic data. Seismic properties are directionally dependent on anisotropic minerals, especially phyllosilicates. Due to tectonic complexity, the effect of anisotropy in seismic data is difficult to determine. Here, we present an approach to determine the anisotropy in representative volumes of deformed phyllosilicate‐rich rocks at the margin of the Tauern Window in the European Alps. Samples from the northern half of the Brenner Base Tunnel were selected within the lithotectonic units Innsbruck Quartzphyllite, Matrei Zone and Bündner schists. Microstructure types were determined by optical microscopy. In order to determine the largest non‐folded volume within phyllosilicate‐rich layers, the crystallographic preferred orientations (CPOs) in millimeter‐thick cylinders were measured using synchrotron X‐ray diffraction. Seismic velocities were computed by using compositional data, single crystal stiffness tensors and the measured CPO. Based on the largest homogeneous–microstructurally undisturbed–volume, an upper bound for a non‐folded state of the velocity anisotropy was estimated. The P‐wave anisotropy varies from 6% to 17%, while the maximum shear wave splitting varies from 0.2 to 0.7 km/s. Fast and slow velocity directions are parallel and perpendicular to the foliation, respectively. Hence, the anisotropy depends on phyllosilicate content and CPO, while other rock‐forming minerals of the crust are largely irrelevant for the elastic anisotropy of schists. Microstructural deviations from planar fabrics lead to a decrease in CPO and anisotropy. These findings need to be considered for the geotectonic interpretation of seismic cross sections in orogens.
Abstract The recognition that relay zones along faults can act as barriers to earthquake propagation implies that rupture size may not be random and earthquakes are constrained by fault geometry. Resolution and observational biases can, however, complicate the integration of actual fault zone geometries into deterministic modeling strategies, thereby negatively impacting earthquake magnitude assessment. This paper describes a statistical modeling strategy as an alternative. The chosen method relies on a stochastic approach combining quantitative fault zone characteristics with a kinematic model linking realistic fault segmentation with expected rupture termination. A sensitivity analysis highlights the effect of fault geometry on rupture sizes, including an increase in the probability of large ruptures, and corresponding changes in the earthquake magnitude‐frequency distribution and associated b ‐values, with decreasing degrees of segmentation and increasing fault maturity. Three case studies confirm these effects for the Yushu‐Ganzi‐Xianshuihe fault system, the North Anatolian fault system, and hydraulic fracturing in Ohio. For each case, simulations show first‐order agreement with documented seismicity, including observed magnitude‐frequency distributions, and support the notion that fault characteristics such as relay zone size and internal segmentation are structural controls determining earthquake b ‐values and maximum magnitudes. These findings suggest that the incorporation of more realistic fault zone geometries and their impact on rupture cascading could improve seismic risk assessment across various regions.
Abstract Silicate melt has been proposed to exist near the base of the mantle and has been invoked to explain seismic and electrical conductivity anomalies observed near the core‐mantle boundary; however, its presence and stability under lowermost‐mantle conditions remain uncertain. Here we report high‐pressure electrical conductivity measurements of Fe 2+ ‐ and Fe 3+ ‐bearing pyroxene glasses (Fe 3+ /ΣFe ≈ 0.5), used as analogs of silicate melts, up to megabar pressures at room temperature. At lower pressures, conductivity increases with pressure and iron content, consistent with electron‐hole hopping between Fe 2+ and Fe 3+ . Above 77–85 GPa, all samples show a marked conductivity decrease, suggesting a pressure‐induced spin transition of Fe 3+ . Previous studies have suggested that lower‐mantle silicate melts may be enriched in Fe 3+ , and iron spin‐state changes may modify iron partitioning between silicate melts and coexisting crystalline phases. Therefore, the Fe 3+ spin transition inferred here may affect the evolution of deep‐mantle melts through pressure‐dependent changes in iron partitioning behavior.
Abstract Through bedrock deformation, solid Earth dynamics (specifically, glacial isostatic adjustment or GIA) influences ice sheet mass balance, especially for marine‐based ice sheets like the West Antarctic ice sheet (WAIS). As ice sheets retreat, crustal uplift can help stabilize a marine based ice sheet by reducing the ice exposed at the grounding line. Predicting bedrock deformation requires accurate understanding of rheology. The mantle beneath WAIS is anomalously hot and exhibits low viscosity. In such cases, even on decadal time scales, viscoelastic deformation of the mantle must be considered. GIA studies often consider 3D Maxwell viscoelasticity but transient rheology has been shown to be important on decadal‐centennial timescales. No study has considered the simultaneous modeling of 3D viscoelasticity and transient viscoelasticity. Under several ice sheet histories that span the Last‐Interglacial (122 ky BP) to several hundred years into the future (with a suite of WAIS projections), we show that transient viscoelasticity can have a greater effect than 3D Maxwell viscoelasticity. In particular, in the WAIS region 3D transient viscoelastic uplift rates roughly double 3D Maxwell viscoelastic uplift rates. This has important implications for ice sheet dynamics and to explore the full extent of 3D transient viscoelasticity on such dynamics, a fully coupled system must be considered. While this study focuses primarily on WAIS, we also explore the effects of 3D transient/Maxwell viscoelasticity in the nearfield of the ancient Laurentide ice sheet, which exhibits opposite rheological trends compared with the West Antarctic region due to the relatively cold mantle beneath northern North America.
Abstract Normal faults in southern Tibet have long suffered from limited ground‐based seismic and geodetic observations, constraining our understanding of both interseismic and coseismic processes and their interactions. The 2025 Mw 7.1 Dingri earthquake provides an opportunity to address these gaps. Here, we integrate interseismic slip deficit rate inversion, finite‐fault slip analyses, and 3D dynamic rupture simulations constrained by geodetic and seismic data to investigate fault behavior. The earthquake ruptured two asperities, with a shallow slip gap located above the hypocentral region, and exhibited a slow initiation followed by rapid strain release. A fault bend may have arrested the southward rupture, while the northward termination appears to have been controlled by a low slip deficit rate barrier. The shallow slip gap above the hypocenter may represent a high‐strength segment requiring elevated fracture energy for rupture initiation, resulting in the relatively slow rupture onset during the event. These results indicate that geometric complexity and heterogeneity of on‐fault stress and strength jointly governed the normal faulting. In addition, we quantified seismic moment accumulation rates on 132 normal faults in southern Tibet, finding that the accumulated moments over 500 years correspond to earthquakes of Mw 5.3–7.2. However, geometric and stress‐strength heterogeneities may reduce the likelihood of large cascading ruptures compared to large block‐bounding strike‐slip and thrust faults. Overall, our findings highlight rupture complexities of the Dingri earthquake, while underscoring substantial seismic hazards posed by rift systems in southern Tibet.
Abstract Serpentine is a major hydrous mineral present in hydrous asteroids that have evolved through planetary impacts. Understanding its dynamic behavior is essential to elucidate the redistribution and retention of water during high‐velocity impact processes. The shock response of antigorite was examined at pressures up to 106 GPa using laser‐driven shock compression combined with ultrafast time‐resolved X‐ray diffraction measurements. Results show that shock‐compressed antigorite remained crystalline up to 44 GPa (impact velocity ∼4.7 km/s), whereas it transformed to an amorphous state above 66 GPa (∼6.4 km/s). This transformation was completed within a few nanoseconds during compression, and recrystallization from the amorphous state did not occur during subsequent decompression, indicating that shock‐induced amorphization is the dominant structural response of antigorite. Adiabatic release calculations further indicate that the decompression path of shocked antigorite from a peak shock pressure of 60 GPa intersects the stability fields of several high‐pressure nominally anhydrous minerals (NAMs). On the other hand, the decompression path from peak pressures of 80 to 100 GPa is predicted to remain above the liquidus until ambient pressure is reached. These findings have significant implications for the impact thermal histories of serpentine‐rich hydrous asteroids, such as Ryugu and Bennu.
Abstract The Altyn Tagh Fault, as the northern edge of the Tibetan Plateau, is crucial for understanding the plateau evolution and continental convergence processes. Using teleseismic waveforms recorded by a 300‐km‐long broadband seismic array deployed by our team in northern Tibet, we investigate the detailed crustal and upper mantle structure beneath the middle section of the Altyn Tagh Fault system and the eastern Kunlun Fault. Our receiver function results reveal complex intra‐crustal discontinuities and a laterally variable Moho beneath the study region. These features suggest a vertically partitioned shortening mode in northern Tibet. The upper crust accommodates horizontal shortening through brittle faulting and folding, whereas the mid‐lower crust responds through ductile deformation, distributed shear, and basal thickening. Along with lateral extrusion, far‐field compressional stresses from the India‐Asia collision and the blocking effect from the rigid Tarim Basin craton have contributed to the crustal shortening in the brittle middle section between the Altyn Tagh Fault and the eastern Kunlun Fault.
Abstract Seismic attenuation provides sensitivity to fracture density, fluid saturation, and pore geometry that complements velocity, enabling quantitative imaging of fracture architecture and associated fluid‐pathways in near‐surface bedrock. Yet high‐resolution attenuation models remain rare in refraction studies due to amplitude‐fidelity and modeling challenges. We present the first such model, using viscoacoustic full‐waveform inversion (‐FWI) of active‐source seismic refraction data acquired at the Garner Run field site within the Susquehanna Shale Hills Critical Zone Observatory (Pennsylvania). To preserve amplitude fidelity under the viscoacoustic approximation, we process the data to remove unaccounted‐for effects and convert the modeled pressure wavefields into vertical particle velocity. Our ‐FWI method simultaneously inverts for P‐wave velocity and attenuation, and we interpret the results using rock‐physics modeling to estimate porosity and fluid saturation across a range of permeabilities. The results reveal: (a) a low‐velocity zone beneath the valley, consistent with thick colluvial fill; (b) localized high‐velocity zones within weathered sandstone beneath the hillslopes, interpreted as less‐fractured regions; (c) strong attenuation anomalies down to ∼20 m depth beneath the hillslopes, linked to high porosity and partially water‐saturated pores in the vadose zone; and (d) low‐attenuation anomalies in the colluvium and valley fill, mapped to low‐water‐saturation zones. These results demonstrate that ‐FWI of refraction data provides quantitative, meter‐scale constraints on the architecture and physical state of fractured bedrock and their fluid‐pathways in the near surface.
Abstract Open‐vent volcanoes are characterized by the persistent activity of outgassing, lava effusion and mild explosions, which is, however, occasionally interrupted by more violent explosive activity (paroxysm), for example, Stromboli (Italy). In spite of the different (two orders of magnitude) explosive intensities and different petrological magma source, regular explosive activity and paroxysms share the same inflation pattern of ground deformation. We show that this similar ground deformation pattern is also reflecting a common pressurized shallow reservoir of gas‐poor, crystal‐rich stagnant magma source located at 280 m depth beneath the eruptive vent. We demonstrate that this source can also explain the ground deformation recorded during flank effusive eruptions, suggesting that both explosive and effusive dynamics are involving the same shallow magmatic system. We propose a fluid mechanical model of a basaltic conduit that correlates a geodetically‐derived pressure change of the shallow reservoir and eruptive phenomena at the vent. The pressure changes corresponding to the ground inflation before explosive eruptions (regular “Strombolian” explosion and paroxysm) and deflation during lava effusion are compatible with the magma volume discharged during each eruption, whereas a dynamic pressure caused by viscous magma flow in the conduit can be used to infer the final inflation velocity of the magma surface before the paroxysm.
Abstract In this study, we develop a 3D S‐wave velocity model extending to 120 km depth beneath Northeast China, featuring high resolution for the crust and uppermost mantle, based on physics‐informed neural network framework for eikonal tomography. Notably, this model provides a characterization of the transitional lithosphere‐asthenosphere boundary (LAB) beneath the volcanic regions, while also illuminating its connections to the formation of Quaternary intraplate volcanoes in the region. Key observations include a prominent large‐scale high‐velocity anomaly beneath the Songliao Basin (SLB), which suggests a thicker lithosphere in that area. In addition, we identify undulations in the LAB, as well as diverse low‐velocity features underlying the volcanoes. Drawing from a schematic representation of lithospheric and upper mantle convection, we interpret these patterns as stemming from regional variations in the water‐carbon cycle and lithospheric thickening within the SLB. For instance, inputs from the deep Big Mantle Wedge not only promote lithospheric thickening beneath the SLB, but also enhance asthenospheric mobility, facilitating sodic volcanism in the southeastern and western sectors. In contrast, the Wudalianchi (WDLC) and Nuominhe (NMH) volcanoes, which lack intense deep convection, produce potassic lavas instead. Overall, these insights improve our understanding of intraplate volcanic mechanisms in this region.
Abstract Caldera systems commonly experience inner resurgence, which shapes patterns of volcanic activity. However, incomplete reconstructions of past ground deformation can lead to incorrect recognition of precursory signals. The Campi Flegrei caldera (Italy) provides an exceptional opportunity to address this issue as uplifted marine sediments interlayered with pyroclastic deposits preserved repeated Holocene deformation. Here, we reassess the mechanisms and evolution of caldera resurgence at Campi Flegrei by integrating previous data sets with new observations from the submerged sector, providing improved spatial and temporal constraints. We document ∼190 m of permanent uplift over the last 10.5 kyr, expressed as a ∼9‐km‐wide, bell‐shaped pattern centered on the city of Pozzuoli. This pattern is best explained by the incremental growth of a resurgent dome controlled by magma emplacement at ∼3–4 km depth in a stacked‐sill configuration. Our reconstruction shows that, since resurgence began at 10.5 ka, the volume of shallow intruded magma has been approximately three times greater than the erupted volume, implying dampened eruptive activity. Nevertheless, temporal overlaps between uplift phases and eruptions indicate a direct link between magma transfer and unrest. Resurgence has also marked the inward migration of eruptive vents and the eruption of more silicic magmas, reflecting a major reorganization of the magmatic system. Localized deformation anomalies that deviate from the main displacement pattern are related to localized intrusions and fault reactivation. These findings identify Campi Flegrei as a resurgent dome system tightly connected to the dynamics of present‐day unrest, which may represent the early stage of a renewed eruptive period.
Abstract We present a comprehensive seismic characterization of the gas hydrate system in the outer continental shelf of the Mackenzie Trough, Canadian Beaufort Sea, based on recently acquired multichannel seismic (MCS) data, P‐wave velocity modeling, and 3D thermal modeling. We identified high‐amplitude, reverse‐polarity Bottom‐Simulating Reflections (BSRs) over an area of ∼406 km 2 , marking the base of the gas hydrate stability zone (GHSZ). Iterative migration velocity analysis reveals high‐velocity zones (2.0–2.5 km/s) indicative of hydrate‐bearing sediments overlying low‐velocity free gas (1.0–1.6 km/s). Effective‐medium modeling indicates gas hydrate saturations of 12%–67% of pore space, with maxima near the shelf break, and free‐gas saturations of up to 9%. Combined, these estimates yield a potential methane mass of approximately 0.17–0.23 Gt within the Trough. By integrating a 3D heat flow model constrained by local ocean temperatures, we find that the theoretical base of the GHSZ closely matches observed BSR depths, except in the western sector where BSRs are 40–60 m deeper than the modeled GHSZ. This discrepancy implies that the gas hydrate system in the Mackenzie Trough exhibits localized disequilibrium. Furthermore, these findings suggest that the extensive deep‐marine gas hydrate province and its associated localized thermodynamic disequilibrium, documented off the US Alaska margin, may extend eastward to the Mackenzie Trough, which may represent the easternmost documented occurrence along the US‐Canadian Beaufort margin. While this deep hydrate system is unlikely to be undergoing destabilization from recent ocean warming, sustained long‐term climate forcing may pose a latent risk for future gas hydrate dissociation.
Abstract Transmission electron microscopy imaging of dislocations in olivine indicates heterogeneous structures and diversity of dislocation types. However, the volumes imaged at high resolution are smaller than individual grains even of fine‐grained samples. Electron backscatter diffraction (EBSD) mapping of large areas provides a statistical view of the distribution of dislocations also in coarse‐grained samples. Maps of a deformed single crystal confirm that conventional EBSD can image mobile dislocations. Dislocation densities in polycrystalline samples deformed under high temperature or wet conditions are heterogeneous, both within grains and from grain to grain. Slip systems with Burgers vector b = [100] and [001] dominate, but dislocations with b = [010] are also present. The most prominent dislocation structures are bands of high dislocation density evolving into subboundaries. Dislocation densities along bands and subboundaries and the resulting misorientation are variable, reflecting non‐uniform grain‐internal strain. Bands of high dislocation density and subboundaries commonly terminate at concave parts of grain boundaries of grains containing the bands. Areas of adjacent grains with matching convex grain boundaries generally have low dislocation density, consistent with dislocation density‐driven grain boundary migration. No characteristic structure distinguishing wet from dry deformed samples was found, but the densities of the dislocation types differ. The observed dislocation structures indicate that formation and migration of bands and subboundaries (requiring both glide and climb of dislocations) are an integral part of deformation in the dislocation creep regime at temperatures >1200°C.
Abstract Ambient noise interferometry is a key seismological method that extracts stable empirical Green's functions (EGFs) from seismic noise for subsurface imaging without requiring earthquakes. However, its performance is often challenged by non‐diffuse and persistent directional noise sources, which hinder the retrieval of EGFs and introduce spurious arrivals. The conventional method with year‐long stacking of cross‐correlation functions (CCFs) is impractical for short‐duration deployments and often fails to suppress the spurious arrivals arising from the uneven distribution of noise energy. To overcome these limitations, we develop a self‐supervised neural network weighting stacking (NNWS) method that learns optimal stacking weights for CCFs by jointly optimizing signal‐to‐noise ratio (SNR) and waveform symmetry, both key proxies of a diffuse wavefield. This physics‐constrained, label‐free approach effectively suppresses acausal signals from strong directional noise. Using synthetic tests, we demonstrate that NNWS successfully mitigates contamination even when interfering noise arrives near the true surface wave. Applied to USArray data, our method recovers stable, high‐quality EGFs using only 10 days of continuous data. The resulting SNRs surpass those achieved by conventional 1‐year linear stacking. The resulting EGFs also exhibit enhanced symmetry and significantly suppressed acausal noise, enabling reliable dispersion measurements from short‐duration data. NNWS thus provides an efficient and robust strategy for rapid retrieval of EGFs under realistic, non‐uniform noise conditions, considerably expanding the application scope of ambient noise interferometry.
