Accurate ship detection and tracking has become increasingly vital due to the growth of global maritime trade and complex oceanic activities. Passive acoustic or seismic-based tracking methods, though proven effective, require extensive deployment of sensors, posing challenges in terms of range and accuracy. The recently developed Distributed Acoustic Sensing (DAS) technology offers a dense sampling, cost-effective and real-time solution by using optical fiber cables for wide-area vibration monitoring. This study investigates DAS technology for ship tracking by analyzing the Doppler shift characteristics of ship-generated wavefields. First, ships were detected by the seismic energy maps in spatial and spectral domains. Utilizing the Fourier Synchrosqueezing Transform (FSST), the Doppler shift characteristics of ship signals are examined, and ship trajectories are determined. The inverted trajectory of a ship aligns closely with the actual GPS-based trajectory, thereby validating the effectiveness and accuracy of our approach. These results demonstrate the potential of DAS for reliable ship detection and tracking, providing a robust and reliable tool for maritime surveillance and monitoring in future.

Reservoir monitory technology based on Distributed Acoustic Sensing (DAS) technology provides a new means to characterize and monitor the underground structure, so as to achieve rapid and real-time monitoring and identification of underground target structure changes, which has been widely used in hydraulic fracturing micro-seismic monitoring and precision engineering monitoring. When micro-seismic monitoring technology is used for source location imaging, it is necessary to consider the influence of many factors on the location results, among which the velocity model error is one of the most important factors, so it is a key step to correct the initial velocity model to obtain an accurate velocity model. On the basis of receiving the arrival time data of the perforating source by the distributed optical fiber sensing, the underground horizon is divided according to the acoustic logging curve, the micro-seismic velocity model is corrected by using the perforation event, the travel-time relationship formula based on the velocity model and the propagation path is studied, and a more accurate velocity model is further solved and corrected by the particle swarm optimizatio (PSO) in the global optimization method. The accuracy of the method was validated through numerical simulations, which demonstrated small calibration errors that were almost uniformly below 3%. The method was subsequently applied to real DAS perforation data, achieving satisfactory inversion results.

The prediction and subtraction method is a common approach for suppressing multiples, and the key step lies in adaptive subtraction of multiples. The matched filter is addressed using the fast iterative shrinkage thresholding algorithm (FISTA) in the conventional linear regression method (LRM). However, it necessitates manual trial and error for choosing appropriate shrinkage thresholding values and regularization parameters. The U-net method (UM) leverages a non-linear regression framework for adaptive subtraction, which allows for removing more multiples than the LRM. However, it is prone to cause overfitting and primary damage. The FISTA can be unfolded into network layers and FISTA-net is constructed by replacing the shrinkage thresholding operation with U-net. With the unsupervised FISTA-net method (FM) the estimation of primaries is achieved through unsupervised training with initial recorded data, simulated multiples and initial results of primaries as inputs. The regularization parameters and shrinkage thresholding values are estimated adaptively, and the network is understood as iterative steps involved in FISTA. This paper utilizes a limited quantity of labeled data with primaries and a substantial quantity of unlabeled data for semi-supervised training of FISTA-net. The proposed semi-supervised FM can increase the accuracy of adaptive subtraction by using the information of primaries in labels. In the synthetic data we utilize the actual primaries as labels, while in the field data the estimated primaries obtained through the conventional method are employed as pseudo-labels. The semi-supervised FM demonstrates superior accuracy in multiple separation, overfitting avoidance, and primary preservation compared to the traditional LRM, UM, and unsupervised FM.

Adaptively subtracting multiple model from the initial data is an essential assignment for the successful elimination of seismic surface-related multiples. Conventional expanded multichannel linear regression (EMLR) method has been proposed to address this challenge by utilizing multi-pattern multiple model. These patterns include the multiple model itself and its first derivative, its Hilbert transform and its first derivative of the Hilbert transform, which are matched with the initial data in the EMLR method. It may lead to inaccurate primary preservation or give rise to residual multiples by using the LR model. The existing U-Net method effectively mitigates complex disparities between the multiple model and actual multiples through integrating adaptive subtraction into the non-LR architecture. Nevertheless, residual multiples are produced by this method using the multiple model itself, especially in complex media contexts. In order to improve surface-related multiple removal’s accuracy, we propose the expanded multichannel U-Net (EMUN) method with multi-pattern multiple model. In the proposed method, the initial data is matched with the multi-pattern multiple model through U-Net in the way of self-supervised training without true primaries as labels. The proposed method incorporates rich information from expanded multichannel of U-Net, enabling better U-Net training for adaptive subtraction. In contrast to the EMLR method and the existing U-Net method, the proposed EMUN method exhibits exceptional efficacy in protecting primaries and eliminating surface-related multiples, as evidenced by its outstanding performance in both synthetic and field data tests.

A deep learning network, GPR1DNet-depth, consisting of self-attention mechanism layers and convolutional layers, has been innovatively proposed to obtain 1-D depth domain permittivity from 1-D time domain ground penetrating radar (GPR) data. Convolutional methods can successfully obtain the time domain permittivity model from time domain GPR data. However, it is not easy to obtain the depth permittivity model directly from time domain data due to the spatial misalignment between the two domains. The proposed network adopts an encoder-decoder structure overall. The proposed GPR1DNet-depth network can accurately extract global features using self-attention mechanism layers and obtain local features using convolutional layers. The architecture designed is helpful in representing the local spatial misalignment between time domain data and depth domain model. The synthetic and field data experiments verified that GPR1DNet-depth is more accurate in inverting the depth information of subsurface permittivity. The proposed network has great potential for development in the inversion of permittivity from GPR data and solving the problem of transformations between time and depth domains.

Effectively suppressing seismic multiples relies heavily on the crucial task of adaptively subtracting the simulated multiples from the initial recorded data. By executing adaptive subtraction within the non-linear regression (non-LR) framework the U-net method has shown superior capability in mitigating the intricate disparities between the simulated and actual multiples when compared to the LR method. The low, medium and high frequency-bands of simulated multiples have been employed to effectively address frequency-dependent inconsistencies in the LR method. To further improve multiple suppression accuracy three frequency-bands of simulated multiples are employed as three channels of the U-net input, which are matched with the initial recorded data during self-supervised training in this letter. Compared to the LR method inputting simulated multiples alone, the LR method inputting three frequency-bands of simulated multiples and the U-net method inputting simulated multiples alone, the proposed U-net method inputting three frequency-bands of simulated multiples improves the signal-to-noise ratio (SNR) by 4.41, 2.07 and 1.99 in the synthetic data example, and demonstrates superior improvement in preserving primaries and eliminating residual multiples in the field data example.

Nowadays, one of the predominant deep learning approaches to improve the quality of DAS VSP seismic data is executed through supervised learning, which requires paired training set including data simulation with relevant parameters and solutions of elastic wave equations. However, differences between simulated data and field data in terms of signal regulations and noise distributions often leads to poor results. An alternative approach is self-supervised learning, such as the representative framework--Blind Spot Network (BSN), but unfortunately, the effective information in blind spots cannot be fully utilized. To solve this problem, this paper considers BSN as a basis and establishes a novel self-supervised network--blind spot visualization (BSV) to suppress random noise and improve the quality of DAS VSP data. In BSV, one branch is dedicated to first produce more denoised data with blind spots and then recover the valid information covered by the blind spots, assuming that the signal is partially data-dependent and the DAS noise is conditionally data-independent. The other branch is designed to generate a target for training without blind spots, so that the dual-branch network can accomplish a self-supervised task in the way of supervised learning. More than that, unlike BSN, we utilize a tailor-made blind spot mapper (BSM) to recover effective information in the blind spots. Results of field data testing prove BSV’s advantages in suppressing random noise and improving the quality of DAS VSP data, although test on synthetic data is nearly identical to supervised learning.

Acoustic waves propagating to the surface interact with it, producing seismic waves through coupling. The type of seismic waves generated by acoustic wave coupling at the surface depends on the angle of incidence of the acoustic wave and the physical parameters of the solid medium. We numerically simulate the stationary source acoustic–seismic coupled wavefields of various half-space media. We synthesize the moving source acoustic–seismic coupled wavefields using the interference method. We use the interference method to synthesize the Doppler curves of various seismic phases. The Doppler curves in the acoustic–seismic coupled signals are highly consistent with the theoretically computed Doppler curves, confirming the accuracy of our results. The Doppler equations for seismic waves propagating along the air–solid interface were theoretically derived. This study introduces a novel approach to examining shallow surface properties. It offers theoretical guidance for seismologists to utilize Doppler curves in acoustic–seismic coupled signals for monitoring air traffic events.

Acoustic-seismic coupling is a prominent phenomenon in seismology. Seismologists can track airborne traffic events by analyzing the Doppler curves in acoustic-seismic coupled signals. We conducted a theoretical analysis to explain the phenomenon of multiple-mode Doppler curves seen in the actual data and point out that the multiple-mode Doppler curves in seismic signals originate from the periodic signals generated during the flight of flying objects. Numerical simulations further confirm this theoretical viewpoint: when the signals generated during aircraft flight can be decomposed into multiple single-frequency periodic signals, we can observe a corresponding number of Doppler curves in seismic signals. Conversely, if the airplane signals cannot be decomposed into periodic single-frequency signals, we cannot observe the presence of Doppler curves in the seismic record. If the frequency of the signals generated by the aircraft varies linearly with time, i.e., exhibits the characteristics of linear frequency modulation, a regular frequency shift curve may still appear in the time–frequency spectrum of the seismic record. In such cases, the Doppler curves need to be corrected to be used for accurate tracking of the aircraft. These research findings provide a solid theoretical foundation for seismologists to utilize Doppler curves in seismic data to track airborne events.

When the acoustic waves interact with the ground, seismic waves can be generated. These seismic waves are influenced by the incident angles of acoustic waves and the near-surface seismic-wave velocities. To understand the acoustic-seismic coupling signals, we designed a hard-stratum model (where the transverse wave of the subsurface velocity is higher than acoustic wave velocity) and a soft-stratum model (where the transverse wave of the subsurface velocity is lower than acoustic wave velocity) for acoustic-seismic signals simulation. We employed the hybrid Galerkin method to simulate the acoustic-seismic signals generated by moving acoustic sources. We then validated the existence of Rayleigh and evanescent waves induced by the acoustic-seismic coupling process. Even in high-attenuation subsurfaces, relatively high-amplitude seismic waves can still be received underground at a depth of 200 m. When the transverse wave of the subsurface velocity is greater than the acoustic wave velocity, the surface geophone detects four kinds of seismic waves: longitudinal, transverse, Rayleigh, and evanescent waves. We could successfully capture the Doppler frequency shift phenomenon in both the generated seismic waves and acoustic waves. Moreover, the velocities of both these waves can influence the shape of the Doppler curves.

Distributed acoustic sensing (DAS) is a novel and revolutionary technology that is widely used in the field of seismic exploration. However, the problem of low signal-to-noise ratio (SNR) has always been a serious challenge affecting its processing and interpretation. At present, deep learning-based algorithms show remarkable potential in DAS vertical seismic profile (VSP) data denoising, but the recovery of deep-layer weak signals still needs to be further improved. To solve the above problems, we propose a dual attention denoising network (DADN) combining spatial attention (SA) and channel attention (CA) double attention block (DAB) to improve the recovery effect of deep-layer weak signals. The DADN consists of multiple DABs and utilizes encoder–decoder structure to extract signal features at multiple scales. At each scale, DADN uses a DAB to reassign the weights of feature maps on each channel and region to focus on useful information, which is beneficial for accurately extracting signal features and recovering deep-layer weak signals. The information flow undergoes downsampling and upsampling operations to finally achieve an accurate estimation of the DAS VSP signals. In addition, gradient-weighted class activation mapping (Grad-CAM) is introduced to visually interpret the signal features learned by the network. The visualization results show that the network does accurately distinguish between signals and noise. After training on the constructed semisynthetic DAS VSP dataset, DADN demonstrated competitive performance in both noise suppression and weak signals retention.

Applying the staggered grid finite difference method (SGFDM) for simulating acoustic responses in large-scale, complex, three-dimensional models poses substantial challenges in geophysics, especially in high-resolution stratigraphic model, due to the high computational burden. To address this problem, we have proposed a low overhead parallel optimization method (LOPOM) suitable for heterogeneous architectures, using high-resolution borehole models as examples. The LOPOM enhances computational intensity and optimizes memory bandwidth utilization. This is achieved through the technique of data reuse along the discontinuities of the model and by minimizing such discontinuities within the halo region. LOPOM was implemented and tested on the CUDA platform and the Sunway supercomputer. In each instance, LOPOM demonstrated an optimal acceleration ratio, thus proving its robust performance. Furthermore, the method’s effectiveness was confirmed through its application to fracture-vuggy formation models and digital core models. Finally, the numerical simulation results and the actual logging data are combined to illustrate the application value of LOPOM.

Mars Rover Penetrating Radar (RoPeR) equipped on China’s Zhurong rover has been used for investigating Martian geology characteristics. The migration algorithm is a common tool to map subsurface structures. However, RoPeR uses a monopole antenna with a tilted angle of 16°. Migration methods depending on omnidirectional radiation antennas can lead to inadequate illuminations for subsurface-inclined geological structures. To overcome this limitation, this article proposes a radiation pattern compensation reverse time migration (RPC-RTM) method to RoPeR data, which can achieve RPC by an opposite-placed tilted antenna. This study first examined the radiation patterns of horizontal- and tilted-placed monopole antennas, analyzing the response characteristics of antennas to scattering points and inclined interfaces. Then, we illustrated an RPC-RTM algorithm, which used the opposite-tilted antenna to propagate backward wavefields for RPC. Finally, numerical simulations were implemented to explore how different antenna placements influence the illumination of RTM images. Laboratory data were used to validate the RPC-RTM method and demonstrate its effectiveness. The proposed RPC-RTM applied RoPeR data to image the Martian subsurface structure. The results show that the proposed method produces high-quality imaging results in insufficient illumination areas and does not require an RPC function. This study confirms the effects of the proposed RPC-RTM method for penetrating radar data acquired through nonstandard antenna deployment.

Adaptive subtraction plays a crucial role in the multiple removal method that involves modeling and subtraction steps. The linear regression (LR) based method utilizes the fast iterative shrinkage thresholding algorithm (FISTA) to solve the optimization problem that contains L1 norm minimization constraint of primaries. It selects the regularization factor and shrinkage thresholding value through trial and error. Under the non-LR framework the U-net is used for adaptive subtraction of modeled multiples from the original recorded data. Since U-net has large network capacity, it is prone to overfit to the original recorded data and lead to primary damage. In this paper, we unfold the iterative steps of FISTA to construct FISTA-Net, which takes the original recorded data and modeled multiples as input data and outputs the estimated primaries. The FISTA-Net based method does not require true primaries as labels and uses L1 norm minimization constraint of primaries for unsupervised training. It can adaptively estimate the regularization factor and shrinkage thresholding value, which is replaced by U-net. FISTA-Net introduces the nonlinear mapping ability of U-net into its structure, which can be interpreted as the iterative steps of FISTA. As a result, the proposed FISTA-Net based method can better attenuate residual multiples, avoid overfitting, and preserve primaries compared to the LR-based and U-net based methods.

The Doppler shift of sound signals has been widely studied. However, monitoring and analyzing the Doppler shift characteristics of aircraft-generated seismic signals is still a relatively new field that requires further exploration. We studied the air-to-ground coupled seismic waves generated by moving aircraft, which were measured by 12 short-period seismometers installed near the Beijing Capital International Airport. The coupled seismic signals generated by 127 aircrafts flying over the observation system were effectively recorded, which confirms the feasibility of using seismic methods to monitor air traffic. We clearly observed the Doppler shift of the coupled signals, which is most noticeable in the frequency range above 500 Hz and serves as important input for analyzing aerial trajectories. Based on the theoretical formula of the Doppler shift, we analyzed the influence of various parameters on the curve shape. Then we proposed a new algorithm for tracking aircraft trajectories using Simulated Annealing inversion method. Finally, using the data collected during the experiment and the proposed trajectory inversion method, we successfully calculated the aircraft trajectory. The implications of our research are significant in integrating seismic technology and data analysis for detecting and monitoring aircraft signals in the field of air traffic.

Distributed acoustic sensing (DAS) has been gradually applied to vertical seismic profiling (VSP), where the generated DAS VSP seismic data contains types of complex noise. Therefore, data denoising plays an important role in collecting high-quality geological information. Generative adversarial network(GAN) has been widely used in seismic exploration data denoising these years, but problems such as insufficient optimization objectives, poor signal retention continuity, and insufficient accuracy still remains when processing DAS VSP data. To address these problems, this paper proposes DuGAN, a deep learning network for multi-scale feature extraction and global information discrimination, to better meet the requirements of high-precision in DAS VSP data denoising. Our method takes GAN as the basic architecture and chooses the multi-scale codec network U-net to explore the potential correlation of DAS data at different scales and a more robust feature representation of DAS signals. In addition, DuGAN is more inclined to emphasize the global role of discriminator so that the entire network ensures the integrity of effective signal structure from a global perspective. Also, for more accurate recovery of the DAS reflected signal, we adjust the loss function in adversarial training and tilt the target optimized space towards the discriminator. Experiments on synthetic and field DAS seismic data show that DuGAN has better denoising performance-not only the noise-covered signal can be recovered, but also the overall effective events are better preserved.

Dip estimation of geological structures plays an important role in geophysical applications. Principal component analysis (PCA) is a common approach to estimating local dips by decomposing the local gradients of a seismic migration image and obtaining its principal eigenvector. However, PCA is difficult to obtain robust and high-resolution dip estimations for low signal-to-noise ratio (SNR) migration images, while multiscale schemes in digital image processing can achieve a better compromise between noise robustness and dip resolution. Therefore, we propose to adopt a multiscale PCA (MPCA) method coupled with a propagation-weight-based fusion mechanism for seismic dip estimation of low SNR migration image. MPCA consists of three steps: 1) constructing an image pyramid by repeating the low-pass filter from fine to coarse scales; 2) estimating the dip using the PCA method at each scale of the image pyramid; and 3) fusing the multiscale dip estimations using propagation weights from coarse to fine scales. We test the MPCA method on an omnidirectional dip pattern and three seismic migration images and compare with the conventional PCA and multiscale methods. The results demonstrate that MPCA yields robust and high-resolution dip estimations for low SNR seismic migration images.

In volcanic monitoring, accurate source locations of volcanic tremors are helpful to understand the volcanic activity and further predict an eruption. The waveform-based interferometric source location imaging method is a conventional and effective tool for locating tremors, which usually have complex signals that lack clear phases and onsets. But the imaged source locations are degraded by the blurring effect caused mainly by the limited source-receiver geometry, showing as low-resolution and high-uncertainty. We introduce Hessian matrix, which governs the blurring effect, into the interferometric source location imaging method. The proposed Hessian-based least-squares interferometric source location imaging method consists of three steps: 1) construct Hessian matrix of source location; 2) obtain an initial source location image by summating back-projections of the correlograms; and 3) update the source location image using a Hessian-based least-squares inversion. We assess the robustness of the proposed method using synthetic data with the consideration of different source locations and velocity errors. Compared with the conventional method, the proposed method can reduce the blurring effect of Hessian matrix and provide high-resolution and low-uncertainty source locations, no matter whether the source is inside or outside the seismic network. We apply the proposed method to a real tremor occurred at Katla volcano. The method recovers the tremor location within 1 km from its most likely source zone.

Time-delay estimation is a critical step in many geophysical applications. Conventional approaches are mainly based on the cross correlation of waveforms in time domain but show strong distortions with multiple oscillations and side lobes. We propose a band-limited phase-only correlation (POC) algorithm for time-delay estimation. The algorithm involves the following key steps: 1) transforming 1-D waveform into 2-D time–frequency spectra using S-transform; 2) calculating the band-limited POC function of the transformed spectra; and 3) measuring time delays by analyzing POC coefficient. We demonstrate the effectiveness of the proposed algorithm using synthetic and real microseismic fiber-optic distributed acoustic sensing (DAS) datasets. Results show that POC can effectively and accurately estimate the time delays of waveforms. Compared with the performance of the conventional cross correlation method in time domain, the proposed method has three main advantages: 1) better identification of event and noise waveforms; 2) lower uncertainty of narrow correlation peaks; and 3) weaker distortions with small oscillations and side lobes.

A high-speed train can be regarded as a moving seismic source when it travels along a railway. Seismic waves from such sources have strong energy and can be used for near-surface characterization, safety monitoring of high-speed railways, and detection of urban underground spaces. Distributed acoustic sensing (DAS) is a newly developed seismic acquisition technology. It has attracted widespread attention due to its advantages of low cost, high sensitivity, high efficiency, and dense sampling. This study investigated near-surface characterization using high-speed train seismic data recorded by DAS. The data were processed to obtain surface waves by seismic interferometry. Thereafter, the extracted surface waves were inverted to obtain the near-surface shear-wave velocity model using a multichannel analysis method. The inverted model is consistent with the subsurface geology of the study area. Our results demonstrate the effectiveness and reliability of DAS-based acquisition and data analysis in near-surface characterization using the high-speed train type of moving sources.