Man-made activities contribute to the ambient seismic wavefield. Locating these anthropogenic seismic noise sources at a high spatiotemporal resolution becomes vital for identifying human activities, understanding source mechanisms, and imaging subsurface geological structures. Existing noise source localization methods based on array analysis face challenges in time-lapse monitoring of the source distribution due to the computational cost. Here, we introduce an efficient approach of using deep learning to monitor the change of seismic noise source locations. The neural network extracts noise source locations from noise cross-correlation functions. The network model is trained with abundant simulated data at a low cost, and then it is sustained for monitoring the changes of noise source distribution in a survey area. We propose a comprehensive framework to ensure prediction accuracy and stability, which allows us to monitor the noise source distribution at a high spatiotemporal resolution. Field data examples acquired by small nodal arrays in the urban area of Hangzhou, China, suggest that the anthropogenic noise sources can be monitored at meter and second scales. The results reveal spatiotemporal variations of the vehicle traffic distribution along a road and two drilling machines working at a construction site. Our study demonstrates an approach to develop deep learning models for real-time monitoring of the distribution of seismic noise sources associated with human activities.

Monterey Canyon, one of the most representative submarine canyons worldwide, remains debated for its evolutionary history due to limited observational coverage and imaging resolution in a complex marine setting. Here we present an advanced seismic velocity model with adaptive resolution, extending around 20 km seaward from the head of Monterey Canyon across the continental shelf to a depth of 1.5 km. This model is constructed using a novel multi-scale ambient noise imaging framework that integrates cross-scale distributed acoustic sensing observations from submarine fiber-optic cable with adaptive shear-wave velocity inversion based on Voronoi tessellation. Our results reveal low velocity zones at multiple spatial scales-from shallow anomalies near 0.1 km to deeper structures approaching 1.5 km-that define nested paleocanyon geometries, including deeply incised sediment pathways overprinted by younger fault-guided conduits. By incorporating existing geophysical observations and canyon evolution models, we suggest that these paleocanyons record a multi-phase evolutionary process: initially conditioned by deep-seated tectonic activity, subsequently reshaped by climate-modulated surface dynamics, and ultimately preserved by successive episodes of sediment transports and fault activities. This work offers new insights into landscape evolution at active continental margins and enables deeper understanding of Earth's multi-layered response to climatic and tectonic forcing. It also underscores the transformative potential of repurposing submarine telecommunication cables as dense, long-term seismic arrays-paving the way for a new era in marine geoscience.

Permafrost degradation along the Qinghai–Tibet Engineering Corridor (QTEC) poses severe risks to infrastructure stability. While current monitoring techniques often lack sufficient spatial continuity or depth resolution, Distributed Acoustic Sensing (DAS) offers a scalable alternative for subsurface characterization. This study introduces a high-resolution imaging framework that leverages existing fiber-optic infrastructure and a Convolutional Neural Network (CNN) to isolate transient traffic-induced vibrations from low Signal-to-Noise Ratio (SNR) DAS records. By selectively stacking these detected signals, we expand the recoverable frequency range of ambient noise interferometry to 45 Hz (vs. 35 Hz for standard stacking), enabling the reconstruction of a 2D shear-wave velocity (Vs) profile near the Kunlun Mountain Pass. The imaging results clearly delineate the permafrost table and base, revealing a permafrost thickness of up to 90 m. Furthermore, we identify pronounced lateral heterogeneities and a structural discontinuity interpreted as a fault-controlled talik. This subsurface anomaly spatially coincides with localized subsidence observed via InSAR, highlighting a coupled structural–hydrothermal mechanism for degradation. Our workflow demonstrates the efficacy of AI-enhanced "dark fiber" sensing for identifying concealed cryospheric hazards in remote alpine regions.

Seismic attenuation provides a highly sensitive constraint on fluid-driven processes in the shallow subsurface. These attenuation-derived spatiotemporal insights complement conventional seismic velocity monitoring and can be used for environmental monitoring and engineered subsurface infrastructure management. In this study, we implemented time-lapse Rayleigh-wave attenuation measurements during controlled shallow water injections to quantify the coupled evolution of seismic attenuation and pore-fluid infiltration. The monitoring experiment was conducted over a 14-day period at a localized test site where two vertical wells were hydraulically connected by a permeable pipeline. The frequency-dependent Rayleigh-wave attenuation coefficients are estimated from spectral-ratio slope fitting of multichannel active-source surface-wave records. These measurements are subsequently combined with phase velocities and S- and P-wave velocities to invert for depth-dependent energy dissipation factors and within a layered medium. The resulting attenuation variations are interpreted as proxies for changes in fluid saturation and hydrological properties in the shallow subsurface. The attenuation images clearly delineate the boundary between the pipeline and the surrounding medium and exhibit pronounced temporal variations driven by injection-induced fluid migration. Daily time-lapse variations of attenuation over the 14-day experiment reveal frequency-dependent responses to the intermittent injection schedule, with peak values near the period of maximum injection. These patterns reflect the migration and redistribution of pore fluids within the near-surface formation. The inverted Q images further identify localized low-Q zones around the pipeline and the two wells, indicating enhanced energy dissipation associated with fluid accumulation and increasing saturation. This study establishes a powerful framework for monitoring fluid migration and its physical impacts from time-lapse seismic attenuation. Our results highlight the importance of attenuation-based imaging for advancing high-resolution characterization of near-surface hydrological and engineered subsurface environments.

Monitoring geological disasters and characterizing hydrogeological processes have intensified the urgency of obtaining high-resolution shear(S)-wave velocity structures. Despite waveform inversion of surface waves playing an increasingly important role in obtaining high-precision S-wave velocity structure, conventional waveform inversion faces challenges, including the local-minima problem and a lack of robustness. Deep learning has recently shown great promise in waveform inversion. Generative adversarial network (GAN) is a superior deep learning framework, rooted in game theory, applicable to this problem. GAN focuses on learning the distribution of data with an enhanced data representation ability and robustness. Therefore, we propose integrating GAN into the waveform inversion of Rayleigh waves (GANWIR) to form an unsupervised inversion framework with physical constraints and no labeled training dataset to mitigate conventional waveform inversion challenges. A detailed explanation of our scheme is presented and its accuracy, reliability, and robustness are evaluated using synthetic cases. Furthermore, our inversion framework is applied to field data acquired in Olathe, KS, USA, demonstrating successful reconstruction of a high-precision S-wave velocity model and a model of the wastewater migration path. Our findings delineate the bedrock interface as well as drainage systems and revise the position of a paleochannel infilled with weathered materials or a fracture/fault zone reported in a previous study. Revised interpretations offering a reasonable and reliable subsurface structure that improves the understanding of local hydrogeological characteristics. This method provides a new approach for waveform inversion of surface waves, establishing a robust foundation for further research and development of advanced high-precision seismic imaging and monitoring techniques.

The spatial distribution and characteristics of ambient seismic noise sources critically influence the reliability and accuracy of ambient noise interferometric imaging, a widely used technique for subsurface characterization. We propose a novel localization method based on the similarity of cross-component Rayleigh wave cross-correlation functions. Unlike conventional approaches that rely on the assumption of retrograde particle motion of Rayleigh waves, our method eliminates such prerequisites, enhancing robustness and applicability in complex wavefields with multiple surface wave modes, especially higher-order modes that often exhibit prograde motion. Comprehensive synthetic simulations demonstrate accurate resolution of source backazimuths under various scenarios, including multi-source configurations and disparate source energy ratios, outperforming two existing methods in terms of accuracy and robustness. Field data from a dense seismic array deployed on a university campus confirm the method’s effectiveness in capturing both persistent traffic-related sources and time-varying active sources. By integrating localization results to selectively extract data from the stationary phase zone, we enhance surface wave dispersion imaging quality and suppress artifacts from off-line and incoherent sources. The method also mitigates near-station effects caused by localized, non-coherent energy that often degrades interferometric stability. This method provides a more stable and reliable framework for ambient noise source localization, advancing the practical application of ambient noise interferometry.

Extreme climate events and increasing geohazard risks require high-resolution near-surface seismic imaging to better characterize subsurface structures. Ambient seismic noise provides a cost-effective alternative to active-source surveys and has been widely used for S-wave velocity imaging through dispersion-based ambient-noise tomography. However, these approaches rely on accurate Green’s function retrieval, which assumes isotropic and uncorrelated noise sources—conditions rarely satisfied in real field environments. As a result, waveform distortions and resolution loss are common, limiting the quantitative interpretability of conventional ambient-noise imaging.Ambient-noise full waveform inversion (FWI) offers a promising pathway to overcome these limitations by directly fitting cross-correlation waveforms and fully exploiting waveform information. Nevertheless, its application remains challenging due to strong trade-offs between subsurface structure and unknown noise source characteristics, severe nonlinearity and cycle-skipping, and the lack of reliable constraints on noise source distributions. These issues have so far hindered the practical implementation of ambient-noise FWI in complex near-surface settings.To address these challenges, we develop a physics-informed generative adversarial network (PIGAN) framework for ambient-noise waveform inversion to accurately estimate physically consistent velocity models in a distributional sense. The wave-equation-based cross-correlation operator is embedded into the generator to ensure physical consistency, while a neural-network discriminator evaluates the mismatch between observed and simulated data. A one-dimensional Wasserstein distance is adopted to enhance robustness to noise and phase uncertainties. The proposed method organically integrates wave-equation constraints, deep learning, optimal transport metrics, and a minimax game formulation, combining the strengths of physics-informed modeling and data-driven representation. This framework enables joint inversion for subsurface velocity structure and ambient noise source characteristics, effectively mitigating source–structure trade-offs. Moreover, it does not require labeled datasets or network pretraining; therefore, the framework is flexible and enables inversion with minimal user interaction. Synthetic tests and field applications in the Qinghai–Tibet Engineering Corridor demonstrate improved resolution and deeper illumination, providing new constraints on fault zone structures and implications for geohazard assessment.

The integrity of dam structures is critical for ensuring the safety and functionality of water resource management systems worldwide. Traditional monitoring methods often fail to capture the dynamic and complex behaviors of dam structures under various environmental conditions. This study introduces a novel approach using high-resolution Fiber-optic Distributed Acoustic Sensing (DAS) with non-destructive ambient noise imaging to enhance the monitoring capabilities for dam health assessment. We deployed DAS technology at Chenjia Reservoir, capturing continuous seismic data that facilitated the detection of subtle structural changes and potential weaknesses within the dam. The methodology leverages ambient noise interferometry and advanced pseudo three-dimensional (3D) inversion techniques to extract and analyze surface wave signals, thereby constructing detailed subsurface shear wave velocity profiles. Comparative analysis with nodal observations validated the superiority of DAS in providing denser spatial sampling and clearer signal resolution. This led to the identification of two low-velocity anomalies aligned with observed seepage outfalls and the revelation of a potential seepage channel at a depth of 10-15 m through 3D visualization. The study underscores the robust capabilities of DAS in providing high-resolution data and real-time monitoring for enhanced dam safety protocols, with important implications for engineering geology in non-invasive subsurface characterization and geohazard detection.

Abstract Train traffic represents a powerful but underutilized seismic source for near-surface geophysics. We propose a method that exploits this source to extract guided P waves, which complement surface wave normal modes, for constraining P wave velocity structure. Using seismic interferometry by cross-coherence on stationary-phase train seismic data segments, we retrieve multimodal Rayleigh and guided P waves. Their dispersion curves are then jointly inverted for P and S wave velocity profiles via a determinant misfit function. The effectiveness of this approach is validated by synthetic data tests. From field train seismic data, we extract strong guided P waves in a frequency band of 18 to 30 Hz. P and S wave velocity structures down to a depth of 50 m are simultaneously determined from guided P wave and multimodal Rayleigh wave dispersion data. The results demonstrate that guided P waves can be extracted from train-generated vibrations and inverted for determining shallow P-wave velocities, which bring a novel approach for near-surface P wave velocity estimation using a train traffic seismic source.

In near-surface investigations, the advent of massive seismic data has ushered in the application of deep learning (DL) techniques for surface wave inversion to attain the shear-wave velocity (Vs). While the efficiency of DL inversion surpasses that of classic physics-driven methods, its broader attributes remain underexplored. Our study delves into a comparative analysis of DL inversion versus physics-driven inversion, focusing on three key aspects: anti-noise ability, stability, and generalization.In numerical experiments, we employ the neighborhood algorithm (NA) (Wathelet, 2008) as a representative of physics-driven inversion, and a convolutional neural network (CNN) constructed for near-surface investigations (Chen et al., 2022) as a representative of data-driven inversion. In addition to comparing the two methods, we also explore the characteristics of joint inversion using Rayleigh-wave dispersion curves (DCs) and Love-wave DCs in the three above aspects. To quantitatively evaluate inversion results, we calculate the root mean square error and relative error of both DCs and Vs. The assessment of anti-noise performance involves applying NA and CNN to DCs with varying noise levels. To gauge stability, we introduce errors in compressional-wave velocity (Vp) and density, examining their effects on inversion precision. Lastly, to assess generalization, we use NA and CNN to invert DCs whose Vs exceeds the range of the training dataset by different percentages.Our findings reveal that DL inversion has a higher anti-noise ability compared with NA. Both methods demonstrate high stability, with errors in Vp and density exerting a slight impact on inversion results, aligning with surface wave inversion characteristics. Compared with physics-driven inversion, generalization is a unique feature of data-driven inversion. The experimental results indicate that the CNN can predict Vs models that are not included in the training dataset although this ability is somewhat limited. Furthermore, like physics-driven inversion, joint inversion enhances all three examined aspects for data-driven inversion. This analysis of characteristics can guide the selection of inversion methods for surface wave applications in near-surface investigations. References:Wathelet M., "An improved neighborhood algorithm: parameter conditions and dynamic scaling," Geophysical Research Letters, vol. 35, no. 9, pp. 2008, doi: 10.1029/2008GL033256. Chen X., Xia J., Pang J., Zhou C., and Mi B., "Deep learning inversion of Rayleigh-wave dispersion curves with geological constraints for near-surface investigations," Geophysical Journal International, vol. 231, no. 1, pp. 1-14, 2022, doi: 10.1093/gji/ggac171.

Retrieving surface waves using linear arrays is becoming gradually popular in urban areas with abundant anthropogenic noise [Mi et al., 2020]. Dispersion measurements can be problematic due to the presence of off-line noise sources. We use the polarization analysis of three-component noise recordings to estimate the back-azimuth and intensity of sources for linear arrays. The noise segment where the source locates in the stationary-phase zones (SPZs) is retained and the noise cross-correlation function (NCF) is weighted according to the source intensity. In this way, we obtain accurate virtual shot gathers and dispersion images.A single three-component seismic station can simultaneously record vertical, north, and east displacements. The back-azimuth and intensity of a noise source within a time segment can be estimated by the relationship between the vertical-horizontal cross-spectra [Takagi et al., 2018]. In practical applications, we remove the mean and trend of the raw three-component noise recordings and divide them into multiple segments. We use the polarization analysis for each segment to locate the orientations and intensities of the noise sources. We then average the results obtained at multiple stations in a linear array to obtain more robust results. We retain the noise segments where the noise sources are distributed in the SPZs and perform weighted stacking of their NCFs according to the intensities of the noise sources in these segments to obtain the final NCF and perform the subsequent dispersion measurement. We use a synthetic experiment and two field examples to demonstrate the superiority of our proposed method. After using the proposed method, the NCFs become more accurate with a higher signal-to-noise ratio, and the trend of the dispersion energy is more continuous. Takagi, R., Nishida, K., Maeda, T. & Obara, K., 2018. Ambient seismic noise wavefield in Japan characterized by polarization analysis of Hi-net records. Geophysical Journal International, 215, 1682–1699. doi:10.1093/gji/ggy334Mi, B., Xia, J., Bradford, J.H. & Shen, C., 2020. Estimating near-surface shear-wave-velocity structures via multichannel analysis of Rayleigh and Love waves: an experiment at the Boise hydrogeophysical research site. Surveys in Geophysics, 41, 323–341. doi:10.1007/s10712-019-09582-4

Seismology has advanced significantly with ambient noise interferometry, enhancing the extraction of surface waves for detailed Vs imaging. However, direct Vp measurements remain critical for geological and engineering applications. Guided P waves offer potential for Vp insights, yet their detection from ambient noise is challenging, especially in urban environments. This study explores the feasibility of using underground urban seismic noise, specifically tunnel traffic, for guided P-wave extraction and subsurface imaging. We introduce an adaptive workflow for pre-processing and enhancing guided P-wave signals, applied to data from the Zijingang Tunnel beneath the campus of the Zhejiang University. Our refined data processing workflow significantly improves the quality of retrieved empirical Green's functions of guided P waves. The joint inversion of surface waves and guided P waves provides high-resolution Vs and Vp profiles, revealing detailed subsurface structures. We also examine the impacts of source depth and type on guided P-wave retrieval. We observed distinct differences in urban ambient noise characteristics between daytime and nighttime, suggesting that daytime traffic noise, with higher frequency content, is more suitable for high-frequency guided P-wave retrieval. Our results demonstrate the effectiveness of using underground urban seismic noise for guided P-wave extraction, highlighting the potential applications of urban ambient noise imaging in geotechnical engineering, urban planning and seismic hazard assessment.

Seismic ambient noise, once regarded merely as interference, has emerged as a crucial geophysical technique, attracting substantial interest from researchers. Despite its advancements, there is a perceptible deceleration in the field, sparking debates over its prospective trajectory. This review employs the bibliometric analysis approach to deeply explore the evolution and trends of the seismic ambient noise method. We outline publication trends and identify three distinct developmental phases that have markedly shaped the seismic ambient noise research. Through structured network analysis, we provide a detailed examination of the thematic trends and the evolving landscapes of the field. We also present a critical discussion on present challenges and opportunities, aiming to highlight areas ripe for innovation. The findings are intended to not only illuminate the potential for further advancements but also to stimulate continued research that could extend the boundaries of the discipline.

High-frequency surface-wave analysis methods have been effectively and widely used to determine near-surface shear(S)wave velocities from single-component seismic data.Multicomponent surface waves can be analyzed with three-component seismic data.Rayleigh-wave spectra extracted from vertical and radial components contain different dispersion energy.By joint analysis of vertical and radial components,the extracted Rayleigh-wave dispersion curves will be expanded in a wider frequency range,which can provide more information for the inversions.The polarization of Rayleigh waves can be used to capture the ambient noise source directivity and provide constraints on the estimation of S-wave velocity.Love waves extracted from the horizontal-component data can be jointly analyzed with Rayleigh waves to obtain both S-wave velocities and the radial anisotropy.This paper gives an overview of the recent developments on using multicomponent active and passive surface waves for near-surface characterization with a high accuracy.

Roadside traffic noise,as a readily available source of seismic waves,holds potential for extracting body wave signals in passive seismic imaging.Based on stationary-phase theory,we propose a method to extract high-quality body wave signals from traffic noise,addressing the challenge of body wave extraction in conventional ambient noise imaging.By applying the multichannel coherency-weighted stack method with a targeted velocity window,we effectively enhanced the contribution of body wave sources while suppressing surface wave signals.In the cross-correlation virtual shot gathers,we observed clear P-wave direct arrivals and used their slopes to estimate the P-wave velocity of the shallow medium.Autocorrelation profile analysis further revealed P-wave reflections from the gravel layer and bedrock interfaces,successfully characterizing the depth of these interfaces and their lateral variations,particularly the notable differences in gravel layer thickness at both ends of the profile.This method demonstrates the application potential of body wave signal extraction under roadside traffic ambient noise conditions,providing a new technical approach for high-resolution subsurface imaging.

With the rapid advancement of artificial intelligence, deep-learning-based inversion frameworks are increasingly being adopted to tackle the challenges associated with surface-wave dispersion curve (DC) inversion. Compared with classical model-driven methods, the deep-learning-based inversion is known for its higher efficiency and independence from the initial model. Existing researches, however, have focused on algorithm design and case applications. The reforms that deep learning techniques can bring to inversion need further exploration. Therefore, we explored the anti-noise ability, stability, performance in joint inversion scenarios, and generalization ability of deep-learning-based inversions. For the first three characteristics, we select a published neural network and the neighborhood algorithm as representatives of deep-learning-based and model-driven inversions, respectively, to compare the corresponding performance of these two methods. The comparative tests and statistical analyses reveal that deep-learning-based inversion exhibits superior anti-noise ability and stability, but shows limited improvement in joint inversion performance. And the statistical results from tests for generalization ability show that the trained neural network can predict the shear-wave velocity (Vs) model whose Vs oversteps the model space of training dataset within 20

Vehicle traffic generates vibrations propagating in the subsurface. Identification and clustering of these seismic sources are crucial for traffic monitoring and subsurface imaging. We propose a novel method which uses a single seismic station and deep clustering to categorize the traffic signals. We utilize a deep embedded clustering (DEC) to extract features from frequency-time spectrograms of the recorded seismic signals. The similar traffic signals are grouped according to their key features and further used to infer the type of the vehicles. This deep clustering framework is unsupervised without manual labeling. Synthetic tests achieve a clustering accuracy of more than 99 %. We apply the method to field seismic recordings at three sites nearby the roadside with traffic videos for label validation. Results show an average accuracy of approximately 83 % and 91 % for vehicle type classifications at the intersection sites (Sites 1 and 2), respectively, where there are speed bumps in the roads. The vehicles moving in the near and opposite lanes are also distinguished from each other, with an accuracy of 73.3 % and 90.2 % at Site 1, and 88.4 % and 86.3 % accuracy at Site 2, respectively. At Site 3 along a straight road, the deep clustering model maintains 82 % accuracy for identifying heavy vehicles (buses and trucks), although the classification of small vehicles (cars and bikes) is limited to 58 % due to the relatively weak seismic signals generated by the light vehicles. The results confirm the framework's ability to cluster traffic seismic signals. By addressing the lack of single-station methods for traffic signal classification with unsupervised deep clustering, the proposed method offers a low-cost and scalable alternative to traditional camera-based traffic sensing systems, providing an effective tool for traffic seismic monitoring at the city scale.

Extreme climate events and geological disasters have intensified the urgency for advancing seismic imaging and monitoring. Despite developments in seismic instrumentation, particularly with seismometers and Distributed Acoustic Sensing (DAS), fine-scale observations remain challenging due to their inherent limitations and deployment configurations. This study introduces a novel hybrid sensing interferometry method that enhances multi-component signal extraction, especially poor horizontal components, through a two-step cross-correlation of DAS and seismometers. A field application near the Qiantang River in Hangzhou illustrates how our proposed framework retrieves high-quality multi-component empirical Green's functions and advances ultra-short duration ambient noise seismic imaging techniques, including surface wave dispersion measurements and horizontal-to-vertical spectral ratio assessments. Our approach also facilitates monitoring of near-surface seismic velocity changes, dv/v $dv/v$, with an unprecedented 10-min resolution, shedding light on shallow dynamic hydraulic responses. This innovative hybrid sensing framework offers new perspectives and methodologies for transforming future research in seismological observation, imaging, and monitoring.

With the advancements of three-component seismic instruments, much valuable information about source distributions and subsurface structures can be utilized to improve passive surface-wave imaging with anthropogenic seismic noise in urban environments. Current passive surface-wave methods, however, are mainly concerned with the Rayleigh waves in the vertical (Z) component, often neglecting the useful dispersion information in the radial (R) and transverse (T) components, particularly in higher modes of surface waves. Therefore, we introduced the common-midpoint two-station (CMP-TS) analysis to extract the multimode dispersion curves of Rayleigh and Love waves from three-component noise recordings using multicomponent ambient seismic noise cross-correlations (ZZ, RR, and TT components). Results from synthetic datasets from given models show that the CMP-TS method is able to retrieve higher mode Rayleigh waves from the RR component and improve the multimode dispersion measurements of Rayleigh waves by the summation of the ZZ and RR spectrograms. Besides, this method can extract multimode dispersion curves of Love waves with high resolution from the TT component. We applied the CMP-TS method to process three-component field data and retrieved the dispersion curves of Rayleigh waves with the fundamental and the first higher modes, as well as Love waves with the fundamental, the first, and second higher modes. The S-wave velocity model is constructed by inverting the fundamental and higher mode data and validated through borehole S-wave velocity measurements.