Search

This study aims to detect the bioacoustics signal in the underwater soundscape, specifically those produced by snapping shrimp, using adaptive iterative transfer learning. The proposed network is initially trained with pre-classified snapping shrimp sounds and Gaussian noise, then applied to classify and remove snapping-free noise from field data. This separated ambient noise is subsequently used for transfer learning. This process was iterated to distinguish more effectively between ambient noise and snapping shrimp sounds characteristics, resulting in improved classification. Through iterative transfer learning, significant improvements in precision and recall were observed. The application to field data confirmed that the trained network could detect signals that were difficult to identify using existing threshold classification methods. Furthermore, it was found that the rate of false detection decreased, and detection probability improved with each stage. This research demonstrates that incorporating the noise characteristics of field data into the trained network via iterative transfer learning can generate more realistic training data. The proposed network can successfully detect signals that are challenging to identify using existing threshold classification methods. [ABSTRACT FROM AUTHOR]

Least‐squares reverse time migration, a linearized inversion problem, can provide high‐quality migration image by minimizing the misfit function, which is defined by predicted and observed data. According to the theory of data‐domain least‐squares reverse time migration, a forward source wavefield that is simulated with a fixed background velocity does not change during iterations. However, storing the forward source wavefield directly into computer memory involves substantial memory consumption. Although a source wavefield reconstruction technique can be applied during least‐squares reverse time migration iterations, this approach can increase the computational cost because of the need for additional wavefield simulations. To alleviate this computational issue in storing the forward source wavefield, we propose an efficient least‐squares reverse time migration scheme based on an excitation amplitude method. Unlike conventional excitation amplitude imaging conditions, the proposed least‐squares reverse time migration scheme enables one to reconstruct the forward source wavefield by convolving a source wavelet with the excitation amplitude of Green's function at the excitation time. With this excitation amplitude method, the forward source wavefield can be efficiently stored in the computer memory because the sizes of the excitation amplitude and excitation time maps are equal to the size of one snapshot. To validate the feasibility of our least‐squares reverse time migration scheme, we perform dot‐product tests and compare forward source wavefields, demigrated data and gradient vectors obtained by conventional least‐squares reverse time migration and our proposed least‐squares reverse time migration. Using numerical tests on synthetic data, we confirm that our least‐squares reverse time migration can produce high‐quality migration results with a significant improvement in computational efficiency with respect to performance time and memory consumption. [ABSTRACT FROM AUTHOR]

Least-squares reverse-time migration (LSRTM) is linearized inversion based on the Born approximation, which commonly seeks to high-quality migration result by the least-squares sense. Contrary to time-domain LSRTM which performs the wavefield simulation as much as several times of the number of shots, frequency-domain LSRTM (F-LSRTM) has the advantage to efficiently deal with multiple shot records. Furthermore, if Green’s function can be saved on memory storage, the full wavefield simulation is implemented only once at the first iteration during entire LSRTM iterations. However, huge memory storage may be required to save the Green’s function for large dataset and model size. To alleviate this computational issue, we propose an efficient F-LSRTM scheme using singular value decomposition (SVD). In our scheme, Green’s function can be saved efficiently as two unitary matrices and one singular value vector with a few number of dominant singular values. Because the number of dominant singular values, called the optimal rank, is much smaller than the minimum value in each dimension size of Green’s function, the proposed method can make it possible to save the Green’s function into the computing memory with keeping the accuracy. After demonstrating the feasibility of reducing the rank of Green’s function, we examine our proposed F-LSRTM scheme and comparative F-LSRTM schemes (F-LSRTM using an adjoint-state method and saved full Green’s function, respectively) using the simple layered and modified marmousi-2 model. Numerical tests indicate that our proposed F-LSRTM scheme can generate migration results as accurate as comparative F-LSRTM schemes with less memory usage and computational cost. [ABSTRACT FROM AUTHOR]

Reverse time migration (RTM) is used to obtain complex structural images of subsurface media. The RTM can be expressed as a zero-lag cross-correlation between the source and receiver wave fields. As imaging conditions can be calculated based on the pre-stored source wavefield and the received wavefield generated during backward modelling, only the source wavefield must be stored in the memory (or disc). Therefore, reducing source-wavefield storage requirements can improve memory efficiency. High-performance computing based on graphic processing units (GPUs) is being developed to reduce the computational time in wave-propagation modelling. Accordingly, GPU-based RTM technology has the potential to improve the computational efficiency of RTM. Storage of the source wavefield wholly in GPU video random-access memory (VRAM) may further improve computational efficiency. In this paper, we present a new algorithm for a three-dimensional (3D) GPU-based RTM that can enable efficient storage of the source wavefield in VRAM using both Nyquist sampling and discrete cosine transform (DCT) compression. A numerical example employing a modified SEG/EAGE 3D overthrust model presented in this study verifies that the proposed algorithm requires only 2% of the memory usage of conventional RTM while producing similar results. [ABSTRACT FROM AUTHOR]

The data-domain least-squares reverse time migration (LSRTM), a promising imaging method for obtaining a high-resolution reflectivity image, can be implemented by matching the de-migrated data to the observed one. However, LSRTM requires expensive computational costs in its implementation due to tremendous memory usage (direct storage on disk/memory) for saving source wavefields and repeatable forward/backward wavefield simulations with iterations compared to conventional RTM. Although the source wavefield reconstruction technique can be used to reduce the memory usage, additional computational cost is inevitable because it is implemented during backward wavefield simulation. To alleviate the computational burden, we have developed an efficient LSRTM scheme with local cross-correlation imaging condition (LSRTM-LC), which can use pre-saved source wavefields according to the window size. Because the window size is much shorter than the total recording time for wavefield extrapolation, storing source wavefields into computer memory can be feasible. In addition, the procedure for storing source wavefields is implemented only at the first iteration because the background velocity is fixed during iterations of LSRTM. To validate the feasibility of the LSRTM-LC scheme, we have carried out several numerical tests with synthetic datasets. Numerical tests demonstrated that LSRTM-LC can provide us with equivalent de-migrated data from saved source wavefields and the gradient vector obtained from LSRTM-LC compared to those obtained from the conventional LSRTM. As a result, LSRTM-LC can generate the same quality of reflectivity models as conventional LSRTM, however, it requires less computational costs compared to conventional LSRTM. [ABSTRACT FROM AUTHOR]

The reverse time migration (RTM) aims to image subsurface reflectors by incorporating with the two-way wave equation. RTM has advantages that it does not have dip limitations and can generate subsurface images clearly even in the case of complex geological structures in the near surface. We have proposed a reproduction wavefield reverse time migration (RWRTM) algorithm implemented with backward reproduced wavefield instead of the back-propagation of the observed data. Reproduced wavefield can be achieved by using the same modelling algorithm used in the source propagation. In general acquisition systems, e.g. streamer in marine or surface geophone in land, conventional RTM back-propagates data acquired on the top grid points in the numerical domain while RWRTM back-propagates reproduced data located on all grid points in the numerical domain, which lead to the generation of migration images with less migration noises and well-balanced migrated events. To validate RWRTM, numerical tests are conducted. For our examples, RWRTM images have better resolution, improved continuity of structure, reduced migration noises, and balanced amplitude compared to the conventional RTM. [ABSTRACT FROM AUTHOR]

Full waveform inversion (FWI) in the time domain has limitations due to large computing time and memory requirements. Some studies have addressed this problem by using machine learning techniques. Most FWI studies using machine learning directly estimate the subsurface velocity structure by training the seismic data generated through various synthetic models to obtain the subsurface velocity structure. In this study, we propose a method to convert the common midpoint (CMP) gather to zero-offset data at a CMP location using a convolutional neural network (CNN) to increase the computing efficiency for the FWI. As the training data, we use synthetic data generated by the seismic exploration geometry and the source signature of field data. Since the proposed method performs FWI using the converted zero-offset data, it can be performed more efficiently than FWI using the existing multichannel data. However, it is difficult to apply a seismic exploration geometry and a source signature that has not been used for training. To verify the proposed method, it is applied to a synthetic model not used for learning as well as focused field data. It is confirmed that a proper subsurface velocity structure was obtained. [ABSTRACT FROM AUTHOR]

Wavefield reconstruction inversion (WRI) mitigates cycle skipping by using an inaccurate initial velocity. This attractive technique is usually implemented with shot records. However, if large numbers of shot records are used, WRI can become computationally burdensome due to the many over-determined linear systems that need to be solved. To alleviate this computational issue, we propose an efficient WRI scheme involving plane-wave encoding (WRI-PW) in the frequency domain. Plane-wave encoding can dramatically reduce the number of relevant datasets by transforming shot records into common ray-parameter gathers with time shifting. Therefore, plane-wave encoding is widely used in many aspects of seismic data processing (e.g. waveform inversion, reverse time migration, etc.). Initially, we performed a simple numerical experiment using a velocity model with a box-shaped anomaly. WRI-PW also could generate scattering wavefields in a homogeneous model. Next, computational efficiency was checked with a modified Marmousi-2 model. The results show that the usage of a sufficient plane-wave angle can achieve satisfactory inversion results. It indicates that WRI-PW requires small datasets compared to WRI. Thus, the computational costs for solving the augmented system can be reduced. Further experiments were conducted to evaluate the robustness of WRI-PW to random noise and to compare WRI-PW and conventional full waveform inversion (FWI) with a modified SEG/EAGE salt velocity model. We verify that WRI-PW is more robust to random noise than WRI, it exhibited less dependency on the accuracy of the initial velocity model than conventional FWI and it is computationally efficient. [ABSTRACT FROM AUTHOR]

This paper examines anisotropic prestack Kirchhoff migration. We used pseudo-acoustic wave equations in the complex frequency domain to describe the wave propagation in a vertical transversely isotropic (VTI) medium. Both amplitudes and traveltimes were calculated efficiently using the suppressed wave equation estimation of traveltime (SWEET) algorithm. The accuracy of the traveltimes obtained with the SWEET algorithm was verified by comparing the traveltime contours simulated with the anisotropic elastic wave equation using a staggered-grid method. Finally, we tested our migration algorithm using the two-dimensional HESS VTI model. We correctly imaged the shape of both the salt and background layer structures. We also reduced the numerical artefacts compared to the isotropic technique. [ABSTRACT FROM AUTHOR]

SUMMARY The Laplace-domain full waveform inversion method can build a macroscale subsurface velocity model that can be used as an accurate initial model for a conventional full waveform inversion. The acoustic Laplace-domain inversion produced is promising for marine field data examples. Although applying an acoustic inversion method to the field data generally requires several pre-processing steps, pre-processing for the Laplace-domain inversion has not been explained in detail. We provide a detailed explanation of how to apply the Laplace-domain waveform inversion to field data through numerical tests with Gulf of Mexico data sets. The pre-processing includes bandpass filtering, muting of the noise before the first arrival, and extraction of the water depth. We choose the range and the interval between the Laplace damping constants empirically by applying a threshold value to the damped time traces and the Laplace-domain wavefields. The observed data are transformed to the Laplace domain using the selected damping; this method yielded a long-wavelength inversion result. The damping constant and the maximum offset affect the penetration depth of the inversion result. The maximum recording time is important for a stable Laplace-transformation and affects the inversion result; however, the latter effect is not significant. [ABSTRACT FROM AUTHOR]

SUMMARY Geophysicists and applied mathematicians have proposed a rich suite of long-wavelength velocity estimation algorithms to construct starting velocity models for subsequent pre-stack depth migration and inversion. Refraction-traveltime tomography derives subsurface velocity models from picked first-arrival traveltimes. In contrast, Laplace-domain waveform inversion recovers long-wavelength velocity structure using the weighted amplitudes of first and later arrivals. There are several implementations of first-arrival traveltime inversion, with most based on ray tracing, and some based on the damped monochromatic wave equation, which accurately represent simple and finite-frequency first arrivals. Computationally, Laplace-domain wavefield inversion is quite similar to refraction-traveltime tomography using damped monochromatic wavefield, but the objective functions used in inversion are radically different. As in classical ray trace-based traveltime inversion, the objective of refraction-traveltime tomography using damped monochromatic wavefield is to match the phase (traveltime) of the first arrival of each measured seismic trace. In contrast, the objective of Laplace-domain wavefield inversion is to match the weighted amplitudes of both first and later arrivals to the weighted amplitudes of the measured seismic trace. Principles of refraction-traveltime tomography were used to generate velocity models of the earth one century ago. Laplace-domain waveform inversion is a more recently introduced algorithm and has been less rigorously studied by the seismic research community, with many workers believing it be equivalent to finite-frequency first-arrival traveltime tomography. We show that Laplace-domain waveform inversion is both theoretically and empirically different from finite-frequency first-arrival traveltime tomography. Specifically, we examine the Jacobian (sensitivity) kernels used in the two inversion schemes to quantify the different sensitivities (and hence the inversion results) of the two methods. Analysing both surface responses and sensitivity results, we show that the Laplace-domain waveform inversion's sensitivity to later arrivals provides significantly improved resolution of deeper velocity structure than the first-arrival traveltime tomography. We demonstrate this capability using numerical inversion examples using a synthetic five-layer model and the synthetic BP benchmark model. Because of the similar algorithmic structure, Laplace-domain waveform inversion fits neatly as a starting velocity model pre-processing component of a larger (multi) frequency-domain wave equation inversion solution package. [ABSTRACT FROM AUTHOR]

SUMMARY Considerable effort has been devoted to the migration of multicomponent data in elastic media with wavefield separation techniques being the most successful. Most of this work has been carried out in the time domain. In this paper, we formulate a multicomponent migration technique in the frequency domain. Reverse-time migration can be viewed as the zero-lag cross-correlation between virtual source and back-propagated wavefields. Cross-correlating the Helmholtz decomposed wavefields rather than directly correlating the vector displacement fields results in sharper, more interpretable images, contaminated by fewer crosstalk artefacts. The end products are separate P and S wave (and if desired, PS and SP) migration images. We test our migration algorithm on synthetic seismic data generated using the SEG/EAGE salt-dome, Overthrust and Marmousi-2 models. We correctly image the location and shape of the target zone for oil exploration using these data sets. Furthermore, we demonstrate that our new migration technique provides good images even when the initial velocity model is only approximate. [ABSTRACT FROM AUTHOR]

SUMMARY Since the recent introduction of the Laplace-domain full waveform inversion, an efficient and accurate modelling technique for the 3-D Laplace-domain wave equation has been sought. The efficiency and accuracy of the 3-D acoustic wave equation modelling in the Laplace domain strongly depends on how to accurately account for free surface conditions and the actual source and receiver locations. In terms of efficiency, fortunately, the Laplace-domain wave equation can be solved on a coarse grid because the field is not propagating as if it were a potential field. However, it is not possible to accurately compute the Laplace-domain response by assuming that the source and the receivers are located at the grid nodes when we use a coarse grid. To resolve this problem, we propose an equivalent source distribution algorithm that allows us to simulate the free surface condition accurately using a coarse-grid finite-element or finite-difference method. It is shown that the equivalent source vector obtained from a homogeneous half-space model can be used for arbitrarily complex models. The extension of the equivalent source to complex heterogeneous media is explained by the approximation of the Dirac delta function. Numerical tests show that our algorithm is better than the Kaiser windowed sinc function method in the Laplace domain. Our technique for solving the 3-D Laplace-domain wave equation can significantly reduce the computational time required for the 3-D Laplace-domain acoustic full waveform inversion because we can use the coarse grid to accurately simulate conventional marine seismic exploration. [ABSTRACT FROM AUTHOR]

Although many full waveform inversion methods have been developed to construct velocity models of subsurface, various approaches have been presented to obtain an inversion result with long-wavelength features even though seismic data lacking low-frequency components were used. In this study, a new full waveform inversion algorithm was proposed to recover a long-wavelength velocity model that reflects the inherent characteristics of each frequency component of seismic data using a single-frequency component decomposed from the spectrogram. We utilized the wavelet transform method to obtain the spectrogram, and the decomposed signal from the spectrogram was used as transformed data. The Gauss–Newton method with the diagonal elements of an approximate Hessian matrix was used to update the model parameters at each iteration. Based on the results of time–frequency analysis in the spectrogram, numerical tests with some decomposed frequency components were performed using a modified SEG/EAGE salt dome (A–A′) line to demonstrate the feasibility of the proposed inversion algorithm. This demonstrated that a reasonable inverted velocity model with long-wavelength structures can be obtained using a single frequency component. It was also confirmed that when strong noise occurs in part of the frequency band, it is feasible to obtain a long-wavelength velocity model from the noise data with a frequency component that is less affected by the noise. Finally, it was confirmed that the results obtained from the spectrogram inversion can be used as an initial velocity model in conventional inversion methods. [ABSTRACT FROM AUTHOR]