Real Seismic Data Research Articles

Seismic stratigraphy interpretation by deep convolutional neural networks: A semisupervised workflow

Depicting geologic sequences from 3D seismic surveying is of significant value to subsurface reservoir exploration, but it is usually time- and labor-intensive for manual interpretation by experienced seismic interpreters. We have developed a semisupervised workflow for efficient seismic stratigraphy interpretation by using the state-of-the-art deep convolutional neural networks (CNNs). Specifically, the workflow consists of two components: (1) seismic feature self-learning (SFSL) and (2) stratigraphy model building (SMB), each of which is formulated as a deep CNN. Whereas the SMB is supervised by knowledge from domain experts and the associated CNN uses a similar network architecture typically used in image segmentation, the SFSL is designed as an unsupervised process and thus can be performed backstage while an expert prepares the training labels for the SMB CNN. Compared with conventional approaches, the our workflow is superior in two aspects. First, the SMB CNN, initialized by the SFSL CNN, successfully inherits the prior knowledge of the seismic features in the target seismic data. Therefore, it becomes feasible for completing the supervised training of the SMB CNN more efficiently using only a small amount of training data, for example, less than 0.1% of the available seismic data as demonstrated in this paper. Second, for the convenience of seismic experts in translating their domain knowledge into training labels, our workflow is designed to be applicable to three scenarios, trace-wise, paintbrushing, and full-sectional annotation. The performance of the new workflow is well-verified through application to three real seismic data sets. We conclude that the new workflow is not only capable of providing robust stratigraphy interpretation for a given seismic volume, but it also holds great potential for other problems in seismic data analysis.

Retrieval of amplitude and attenuation from ambient seismic noise: synthetic data and practical considerations

SUMMARY Ambient noise correlation has been used extensively to retrieve traveltimes of surface waves. However, studies of retrieving amplitude information and attenuation from ambient noise are limited. In this study, we develop methods and strategies to extract Rayleigh wave amplitude and attenuation from ambient noise correlation, based on theoretical derivation, numerical simulation, and practical considerations of real seismic data. The synthetic data included a numerical simulation of a highly anisotropic noise source and Earth-like temporally varying strength. Results from synthetic data validate that amplitudes and attenuations can indeed be extracted from noise correlations for a linear array. A temporal flattening procedure is effective in speeding up convergence while preserving relative amplitudes. The traditional one-bit normalization and other types of temporal normalization that are applied to each individual station separately are problematic in recovering attenuation and should be avoided. In this study, we propose an ‘asynchronous’ temporal flattening procedure for real data that does not require all stations to have data at the same time. Furthermore, we present the detailed procedure for amplitude retrieval from ambient noise. Tests on real data suggest attenuations extracted from our noise-based methods are comparable with those from earthquakes. Our study shows an exciting promise of retrieving amplitude and attenuation information from ambient noise correlations and suggests practical considerations for applications to real data.

Wavelet analysis for spectral inversion of seismic reflection data

In the current work we present a seismic inversion alternative based on the spectral inversion method developed by Puyear and Castagna (2008). We incorporate the wavelet analysis to synthesize singularities of interest and obtain a detailed analysis of seismic windows. The methodology is a suitable alternative to constrained seismic inversion for determination of the reflection coefficients as functions of frequency. Moreover, the forward problem of seismic data as a function of frequency is non-linear; therefore, in order to solve the inverse problem, global optimization techniques are applied: Iterative Solution of Generalized Inversion, Genetic Algorithms and Particle Swarm Optimization. The most reliable results were obtained based on the fact that the wavelet synthesis facilitates getting the first approximation of the events retaining the characteristic amplitude and establishing the initial domain model for the optimization algorithms being only slightly slower than the deterministic deconvolution technique. We show the results of the developed methodology on synthetic and real seismic reflection data.

Amplitude variation with offset inversion using acoustic-elastic local solver

Conventional amplitude variation with offset (AVO) inversion analysis uses the Zoeppritz equations, which are based on a plane-wave approximation. However, because real seismic data are created by point sources, wave reflections are better modeled by spherical waves than by plane waves. Indeed, spherical reflection coefficients deviate from planar reflection coefficients near the critical and postcritical angles, which implies that the Zoeppritz equations are not applicable for angles close to critical reflection in AVO analysis. Elastic finite-difference simulations provide a solution to the limitations of the Zoeppritz approximation because they can handle near- and postcritical reflections. We have used a coupled acoustic-elastic local solver that approximates the wavefield with high accuracy within a locally perturbed elastic subdomain of the acoustic full domain. Using this acoustic-elastic local solver, the local wavefield generation and inversion are much faster than performing a full-domain elastic inversion. We use this technique to model wavefields and to demonstrate that the amplitude from within the local domain can be used as a constraint in the inversion to recover elastic material properties. Then, we focus on understanding how much the amplitude and phase contribute to the reconstruction accuracy of the elastic material parameters ([Formula: see text], [Formula: see text], and [Formula: see text]). Our results suggest that the combination of amplitude and phase in the inversion helps with the convergence. Finally, we analyze elastic parameter trade-offs in AVO inversion, from which we find that to recover accurate P-wave velocities we should invert for [Formula: see text] and [Formula: see text] simultaneously with fixed density.

Background noise suppression using trainable nonlinear reaction diffusion assisted by robust principal component analysis

Due to the severe interference of background noise, the signal-to-noise ratio of desert seismic data is extremely low. In addition, due to low-frequency characteristics of sand in the Tarim desert region, the background noise in desert seismic data is mainly distributed in low-frequency band, so that the frequency spectrum aliasing of effective signals and background noise is more serious than the general land seismic data. Thus, conventional filtering methods cannot effectively suppress background noise in desert seismic data and recover effective signals. In order to overcome the problem that low-frequency background noise in desert seismic data is hard to suppress, a new method called R-TNRD based on robust principal component analysis (RPCA) algorithm and trainable nonlinear reaction diffusion (TNRD) network is proposed in this paper. By using the good sparsity of RPCA, the input noisy desert seismic data are decomposed into a low-rank matrix and a sparse matrix, and these two matrices contain background noise and effective signals. Due to the serious spectrum aliasing of desert seismic data, conventional thresholds have been unable to extract effective signals from the two matrices obtained by RPCA effectively. Therefore, we introduce TNRD network into desert seismic data denoising. By network training with a low-frequency noise set, the optimisation of TNRD network can be achieved, so as to accurately extract the effective signals from the low-rank matrix and the sparse matrix. In the experimental part, we test the performance of R-TNRD on both synthetic and real seismic data. The results demonstrate that the proposed method can suppress background noise more effectively than conventional methods.

Potential Use of Time-Lapse Surface Seismics for Monitoring Thawing of the Terrestrial Arctic

The terrestrial Arctic is warming rapidly, causing changes in the degree of freezing of the upper sediments, which the mechanical properties of unconsolidated sediments strongly depend upon. This study investigates the potential of using time-lapse surface seismics to monitor thawing of currently (partly) frozen ground utilizing synthetic and real seismic data. First, we construct a simple geological model having an initial temperature of −5 °C, and infer constant surface temperatures of −5 °C, +1 °C, +5 °C, and +10 °C for four years to this model. The geological models inferred by the various thermal regimes are converted to seismic models using rock physics modeling and subsequently seismic modeling based on wavenumber integration. Real seismic data reflecting altered surface temperatures were acquired by repeated experiments in the Norwegian Arctic during early autumn to mid-winter. Comparison of the surface wave characteristics of both synthetic and real seismic data reveals time-lapse effects that are related to thawing caused by varying surface temperatures. In particular, the surface wave dispersion is sensitive to the degree of freezing in unconsolidated sediments. This demonstrates the potential of using surface seismics for Arctic climate monitoring, but inversion of dispersion curves and knowledge of the local near-surface geology is important for such studies to be conclusive.

Recognizing of Microseismic Source Types Based on Small-Aperture Seismic Array Data

Abstract—In this paper, we consider the problem of recording and processing the microseismic data with their subsequent interpretation and making the decisions concerning the probable threats associated with the violation of integrity of terrestrial environment due to mining activity. These decisions are made based on the detection of microseismic sources and recognition of their types by processing the multichannel seismograms from small-aperture seismic arrays. The procedures of seismogram processing include the algorithms for detecting the signals of microseismic sources in the noisy seismograms from the array sensors and the procedures for estimating various parameters of these sources. In this paper, we propose a new detection algorithm which is based on evaluating multiple coherence of the array seismograms and is capable of identifying the waves radiated from microseismic sources with complex focal mechanisms against the diffuse seismic noise. In contrast to the widely used STA/LTA array detector which is aimed at separating plane seismic waves, the proposed detector is intended for the recognition of waves from local seismic events in the complexly structured subsurface environment. Based on the numerical experiments with model and real seismic array data containing signals from the sources with complex focal mechanisms, it is shown that the types of microseismic sources can be recognized with the use of various algorithms of seismic array data analysis: with new algorithms for estimating the coordinates of microseismic sources and with traditional algorithms of space-time spectral analysis.

Denoising of desert seismic signal based on synchrosqueezing transform and Adaboost algorithm

Seismic data in desert area generally have low signal-to-noise ratio (SNR) due to special surface conditions. Desert noise is characterized as low-frequency, non-Gaussian and non-stationary noise, which makes the noise suppression in desert area more challenging by conventional methods. Conventional methods are effective for the signal with high SNR, but in desert seismic signal, the SNR is low and the signal can easily be obliterated in desert noise. In this paper, we propose an approach that operates in synchrosqueezing transform (SST) domain and use classification techniques obtained from supervised machine learning to identify the coefficients associated with signal and noise. First of all, we transform the real desert seismic data into time–frequency domain by SST. Secondly, we select features by calculating the SST coefficients of signal and noise. And then, we train them in the Adaboost classifier. Finally, when the training is completed, we can obtain the final classifier that can effectively separate the signal from noise. We perform tests on synthetic and field records, and the results show great advantages in suppressing random noise as well as retaining effective signal amplitude.

Horizon extraction using ordered clustering on a directed and colored graph

Extracting horizons from a seismic image has been playing an important role in seismic interpretation. However, how to fully use global-level information contained in the seismic images such as the order of horizon sequences is not well-studied in existing works. To address this issue, we have developed a novel method based on a directed and colored graph, which encodes effective context information for horizon extraction. Following the commonly used framework, which generates horizon patches and then groups them into horizons, we first build a directed and colored graph by representing horizon patches as vertices. In addition, edges in the graph encode the relative spatial positions of horizon patches. This graph explicitly captures the geologic context, which guides the grouping of the horizon patches. Then, we conduct premerging to group horizon patches through matching some predefined subgraph patterns that are designed to capture some special spatial distributions of horizon patches. Finally, we have developed an ordered clustering method to group the rest of the horizon patches into horizons based on the pairwise similarities of horizon patches while preserving geologic reasonability. Experiments on real seismic data indicate that our method can outperform the autotracking approach solely based on the similarity of local waveforms and can correctly pick the horizons even across the fault without any crossing, which demonstrates the effectiveness of exploring the spatial information, i.e., the order of horizon sequences and special spatial distribution of horizon patches.

Multiple attenuation methods in short-offset 2D marine seismic data: a case study in Cendrawasih Bay

Real marine seismic data are typically embedded with free-surface multiples energy, which are troublesome in imaging an accurate seismic cross-section. In addition, more challenging situation is to bring optimum result with a short-offset streamer due to the coherent nature of multiples. In this study, we present a comparison of three methods for attenuating free-surface multiples energy in short-offset 2D seismic data from Cendrawasih Bay. Multiple attenuation methods include F-K filter, Radon transform, and Surface Related Multiple Elimination (SRME) are processed until the final Pre-Stack Time Migration (PSTM) results. Predictive deconvolution is applied in order to suppress short period multiples prior to free-surface multiple attenuation method. Predictive deconvolution successfully identifies and removes the predictable wavelet of short period multiples. Radon transform shows poor result in short-offset seismic data even if it is combined with F-K filter method. Combination of both F-K filter and SRME are successfully attenuate free-surface multiples and should be considered as proper solution to increase signal to noise ratio.

Evolution in seismic monitoring system and updating seismic zones of Egypt

ABSTRACT On 3 September 2015, an unexpected earthquake with magnitude MB = 4.5 occurred in a completely silent area, located at 40 km south El Alamein city north western desert, where there are ~ 20 million pieces of unexploded ordnance. This study aims to: i) review and interpret the widespread seismic activity in Egypt during the last three decades, and ii) delineate the active seismic source zones map of Egypt based on realistic assumptions. These assumptions include only the real and accurate seismic events from the whole earthquake catalogue of Egypt. The idea based on the accuracy of the monitoring system, magnitude size of earthquakes, and detailed structure elements map of Egypt. The seismic catalogue of Egypt was updating to the end of 2019 including the historical and instrumental earthquakes since −2200 BC. The seismicity map of Egypt was outlined along five interval periods based on the significant evolution in the seismic monitoring system. The advanced monitoring period in the detection and location of seismic events in entire Egypt was supposed to begin in 2003 where the ENSN almost accomplished. Seismicity of advanced monitoring period 2003–2019 can be used as a tracer of real seismicity boundaries entire Egypt. Active seismic source zones map was delineated based on the advanced monitoring period and the unified Mw ≥ 4.5 instrumental catalogue from 1900 to 2019. To validate the seismic zones model, it compared with the structural elements map and given a rational matching. The new local seismic spots entire Egypt was discussed because of the invention in the seismic monitoring system entire Egypt and attributed to the misdetection and/or dislocation events due to the lack of monitoring system before the accomplished of ENSN. In conclusion, a validated active seismic zones map of Egypt was delineated using real and accurate seismic and geologic data based on powerful assumptions.

A Physics-Based Neural-Network Way to Perform Seismic Full Waveform Inversion

Seismic full waveform inversion is a common technique that is used in the investigation of subsurface geology. Its classic implementation involves forward modeling of seismic wavefield based on a certain type of wave equation, which reflects the physics nature of subsurface seismic wavefield propagation. However, obtaining a good inversion result using traditional seismic waveform inversion methods usually comes with a high computational cost. Recently, with the emerging popularity of deep learning techniques in various computer vision tasks, deep neural network (DNN) has demonstrated an impressive ability in dealing with complex nonlinear problems, including seismic velocity inversion. Now, extensive efforts have been made in developing a DNN architecture to tackle the problem of seismic velocity inversion, and promising results have been achieved. However, due to the dependence of a labeled dataset, i.e., the barely accessible true velocity model corresponding to real seismic data, the current supervised deep learning inversion framework may suffer from limitations on generalization. One possible solution to mitigate this issue is to impose the governing physics into this kind of purely data-driven method. Thus, following the procedures of traditional seismic full waveform inversion, we propose a seismic waveform inversion network, namely SWINet, based on wave-equation-based forward modeling network cells. By treating the single-shot observation data and its corresponding shot position as training data pairs, the inverted velocity model can be obtained as the trainable network parameters. Moreover, since the proposed seismic waveform inversion method is performed in a neural-network way, its implementation and inversion effect could benefit from some built-in tools in Pytorch, such as automatic differentiation, Adam optimizer and mini-batch strategy, etc. Numerical examples indicate that the SWINet method may possess great potential in resulting a good velocity inversion effect with relatively fast convergence and lower computation cost.

Seismic Random Noise Attenuation Based on PCC Classification in Transform Domain

Random noise attenuation of seismic data is an essential step in the processing of seismic signals. However, as the exploration environment is becoming more and more complicated, the energy of valid signals is weaker and the signal to noise (SNR) is much lower, which brings great difficulty to seismic data processing and interpretation. To this end, we propose an unconventional and effective seismic random noise attenuation method based on proximal classifier with consistency (PCC) in transform domain. Firstly, we analyze various transforms for seismic data from traditional wavelet transform and curvelet transform to emerging non-subsampled shearlet transform (NSST) and non-subsampled contourlet transform (NSCT). And, we select the excellent NSST to decompose the noisy seismic data into different sub-bands of frequency and orientation responses. Secondly, unlike traditional sparse transform based seismic denoising methods that often directly use a thresholding operator and corresponding inverse transform to denoise seismic data, our proposed method employs a superior performance PCC to classify the NSST coefficients of seismic data before thresholding operator. The added step can effectively divide the NSST coefficients into reflected useful signal coefficients and noise-related coefficients, which can preserve the edge of reflected signals and keep the information of events intact as much as possible. In addition, we also introduce an adaptive threshold computing method and a soft-thresholding method to achieve seismic data denoising better. Finally, the experimental results on the typical synthetic example and real seismic data show the superior performance of the proposed method.

A Spatially Adaptive Edge-Preserving Denoising Method Based on Fractional-Order Variational PDEs

Image denoising is a basic problem in image processing. An important task of image denoising is to preserve the significant geometric features such as edges and textures while filtering out noise. So far, this is still a problem to be further studied. In this paper, we firstly introduce an edge detection function based on the Gaussian filtering operator and then analyze the filtering characteristic of the fractional derivative operator. On the basis, we establish the spatially adaptive fractional edge-preserving denoising model in the variational framework, discuss the existence and uniqueness of our proposed model solution and derive the nonlinear fractional Euler-Lagrange equation for solving our proposed model. This forms a fractional order extension of the first and second order variational approaches. Finally, we apply the proposed method to the synthetic images and real seismic data denoising to verify the effectiveness of our method and compare the experimental results of our method with the related state-of-the-art methods. Experimental results illustrate that our proposed method can not only improve the signal to noise ratio (SNR) but also adaptively preserve the structural information of an image compared with other contrastive methods. Our proposed method can also be applied to remote sensing imaging, medical imaging and so on.

Supervised Dimension Reduction by Local Neighborhood Optimization for Image Processing

Background: Subspace learning-based dimensionality reduction algorithms are important and have been popularly applied in data mining, pattern recognition and computer vision applications. They show the successful dimension reduction when data points are evenly distributed in the high-dimensional space. However, some may distort the local geometric structure of the original dataset and result in a poor low-dimensional embedding while data samples show an uneven distribution in the original space. Methods: In this paper, we propose a supervised dimension reduction method by local neighborhood optimization to disposal the uneven distribution of high-dimensional data. It extends the widely used Locally Linear Embedding (LLE) framework, namely LNOLLE. The method considers the class label of the data to optimize local neighborhood, which achieves better separability inter-class distance of the data in the low-dimensional space with the aim to abstain holding together the data samples of different classes while mapping an uneven distributed data. This effectively preserves the geometric topological structure of the original data points. Results: We use the presented LNOLLE method to the image classification and face recognition, which achieves a good classification result and higher face recognition accuracy compared with existing manifold learning methods including popular supervised algorithms. In addition, we consider the reconstruction of the method to solve noise suppression for seismic image. To the best of our knowledge, this is the first manifold learning approach to solve high-dimensional nonlinear seismic data for noise suppression. Conclusion: The experimental results on forward model and real seismic data show that LNOLLE improves signal to noise ratio of seismic image compared with the widely used Singular Value Decomposition (SVD) filtering method.

A high‐precision time–frequency analysis for thin hydrocarbon reservoir identification based on synchroextracting generalized S‐transform

ABSTRACTImproving the seismic time–frequency resolution is a crucial step for identifying thin reservoirs. In this paper, we propose a new high‐precision time–frequency analysis algorithm, synchroextracting generalized S‐transform, which exhibits superior performance at characterizing reservoirs and detecting hydrocarbons. This method first calculates time–frequency spectra using generalized S‐transform; then, it squeezes all but the most smeared time–frequency coefficients into the instantaneous frequency trajectory and finally obtains highly accurate and energy‐concentrated time–frequency spectra. We precisely deduce the mathematical formula of the synchroextracting generalized S‐transform. Synthetic signal examples testify that this method can correctly decompose a signal and provide a better time–frequency representation. The results of a synthetic seismic signal and real seismic data demonstrate that this method can identify some reservoirs with thincknesses smaller than a quarter wavelength and can be successfully applied for hydrocarbon detection. In addition, examples of synthetic signals with different levels of Gaussian white noise show that this method can achieve better results under noisy conditions. Hence, the synchroextracting generalized S‐transform has great application prospects and merits in seismic signal processing and interpretation.

Nonstationary Least-Squares Decomposition With Structural Constraint for Denoising Multi-Channel Seismic Data

The seismic data usually contains strong random noise, which impedes the effective usage of the seismic signals for imaging and inversion. We propose an effective seismic denoising method based on a least-squares decomposition model. We assume that each trace in the multi-channel seismic data can be decomposed into several smoothly variable components. Since the decomposition is basically an inverse problem, we apply the temporal smoothness to constrain the inversion and control the stability. Considering the spatial coherency in a multi-channel seismic data, we also apply the spatial smoothness constraint to the decomposition. The space constraint is applied along the structural direction of the seismic events to preserve the dipping energy. The structural constraint is equivalent to applying a structure-oriented smoothing that requires the estimation of the local slope from the input noisy data. We validate the effectiveness of the proposed algorithm via several synthetic and real seismic data. The proposed method outperforms the state-of-the-art single-channel and multi-channel algorithms even in the case of strong random noise.

Seismic Noise Attenuation Using Unsupervised Sparse Feature Learning

Noise attenuation plays an important role in seismic data processing. We propose a novel denoising method for seismic data based on unsupervised sparse feature learning. Our goal is to obtain the identifiable feature of the noisy seismic data and then to represent the effective signals. By preprocessing the raw data and training the autoencoder neural network with sparse constraint, the sparse feature of the seismic data can be learned and stored in the neural network. We use the adaptive moment estimation as a backpropagation algorithm to minimize the cost function with a sparse penalty term and combine the dropout technique in the training process to improve the feature extraction and generalization capability of the neural network. Then, the test data set can be reconstructed by the most important sparse features. The final denoising result can be obtained by rearranging the output test data set. Compared with three commonly used state-of-the-art denoising methods, the proposed method performs well in applications to denoising for synthetic and real seismic data.

Channel boundary detection using a 3D morphological filter and adaptive ellipsoidal structuring element

ABSTRACTChannels are one exploratory object in the oil and gas industry, and detailed channel studies can help identify the sedimentation processes in an area. In addition, channels are a drilling hazard in deep-water sedimentary basins. However, manual interpretation of channels is time-consuming and needs skill and experience. Therefore, in this paper, using morphological filters, we present an algorithm for automatically extracting the channel edges from seismic data. We used a morphological gradient algorithm with an ellipsoidal adaptive structuring element to detect edges. Morphological gradient operators highlight the image edges within the neighbourhood of the structuring element. The adaptive structuring element at each point of the data align with the edges and contours at that point, and so can be more effective in edge detection. Eigenvalue decomposition of the gradient structure tensor gives an estimate of the orientation and anisotropy rate of objects within the data. Hence, parameters of the adaptive structuring element are obtained from the eigenvalues of a 3D gradient structure tensor. Finally, the proposed algorithm is applied to synthetic and real seismic data containing a channel, and the results are compared with other well-known edge detection methods such as Canny, Sobel, semblance and negative curvature. The proposed algorithm was able to correctly identify edge locations, and also outperforms other methods in the presence of noise.

Stacking Seismic Data Based on Ramanujan Sums

Real seismic data consist of multiple traces captured by an array of receivers. These multiple traces are sorted by the common midpoint between the source and the receiver, and the time-lag between different traces is corrected by a normal move-out correction process. After these preprocessing steps, the sorted traces contain the same information about the earth sublayers and are corrupted by noise. The next step, termed stacking in seismic signal processing, involves the construction of an optimum trace with an improved signal-to-noise ratio (SNR) from these sorted traces. In this letter, we present an improved method for weighted stacking, where each trace is weighed in accordance with the noise variance. Using the first-order derivative property of Ramanujan sums, we perform the estimation of noise variance in each trace. Numerical results demonstrate that the method presented in this letter has better SNR for the trace obtained after stacking in comparison with existing methods.