Prestack Data Enhancement Research Articles

Deep-learning-based local wavefront attributes and their application to 3D prestack data enhancement

The work presents a novel workflow to accelerate the estimation of local wavefront attributes (LWAs) from massive 3D prestack seismic data using deep learning (DL) focusing on data enhancement. A standard estimation method based on a semblance-based brute-force optimization provides good results but is time consuming. A modification of the U-net convolutional neural network, commonly used in image processing applications, is proposed to link the seismic data with the wavefront attributes. Color pixel image input for the neural network is generated through a straightforward seismic data regularization based on supergrouping followed by red, green, and blue encoding. The proposed workflow can be adapted to any 3D prestack seismic volume. Conventional semblance-based attributes estimation is required for the training step but only for approximately 1% of the total data. The prediction step is very efficient and reduces the overall run time significantly. The verification of the proposed approach is performed on challenging real land and marine data sets. As a result, DL-based estimation of LWAs accelerates computation up to 200 times compared to the standard method. The attributes from the proposed DL-based approach indicate an acceptable match compared with the brute-force semblance-based optimization results. Conventional and proposed estimation methods result in comparable prestack data enhancement results for more reliable seismic processing in challenging areas.

Evaluating strategies for estimation of local kinematic parameters in noisy land data: quality versus performance trade-offs

Abstract Modern land seismic data are typically acquired using high spatial trace density with small source and receiver arrays or point sources and sensors. These datasets are challenging to process due to their massive size and relatively low signal-to-noise ratio caused by scattered near-surface noise. Therefore, prestack data enhancement becomes a critical step in the processing flow. Nonlinear beamforming had proved very powerful for 3D land data. However, it requires computationally intensive estimations of local coherency on dense spatial/temporal grids in 3D prestack data cubes. We present an analysis of various estimation methods focusing on a trade-off between computational efficiency and enhanced data quality. We demonstrate that the popular sequential «2 + 2 + 1» scheme is highly efficient but may lead to unreliable estimation and poor enhancement for data with a low signal-to-noise ratio. We propose an alternative algorithm called «dip + curvatures» that remains stable for such challenging data. We supplement the new strategy with an additional interpolation procedure in spatial and time dimensions to reduce the computational cost. We demonstrate that the «dip + curvatures» strategy coupled with an interpolation scheme approaches the «2 + 2 + 1» method's efficiency while it significantly outperforms it in enhanced data quality. We conclude that the new algorithm strikes a practical trade-off between the performance of the algorithm and the quality of the enhanced data. These conclusions are supported by synthetic and real 3D land seismic data from challenging desert environments with complex near surface.

Prestack seismic data interpolation and enhancement with common‐reflection‐surface–based migration and demigration

ABSTRACTThe standard common‐reflection‐surface stacking method simulates high‐quality zero‐offset stacked data from multi‐coverage prestack data and, as by‐products, provides three kinematic wavefield attributes in two dimensions that can be applied to solve reflection seismic problems. One of the most significant applications of those attributes is for interpolation and enhancement of prestack data through the partial common‐reflection‐surface stack approach. Because this interpolation method is based on the stacking process, it usually introduces spurious noises that can create artefacts in the migration results. In order to overcome these limitations, a new prestack data interpolation and enhancement method is presented by applying the attributes and the common‐reflection‐surface operator based on two fundamental seismic imaging operations: the Kirchhoff‐type time migration and demigration. The diffraction curves required to apply these two fundamental operations are particular cases of the common‐reflection‐surface stack operator, which better fits the diffraction events and, consequently, is the best approximation for the Kirchhoff migration operator. From synthetic and real data, impulse responses of the migration and demigration operations are shown to demonstrate how this new approach regularizes and interpolates prestack data. A simple synthetic example shows good accuracy for reconstructed reflection events and well‐resolved conflicting dip events. The application example in real data reveals that the proposed approach provides cleaner regularized sections with better reconstructed events than the results of the well‐known partial common‐reflection‐surface stack approach. It has been successfully shown that the proposed approach is a well‐founded alternative for interpolation and denoising of seismic data and produces better preconditioned data for prestack migration.

Prestack data enhancement with phase corrections in time‐frequency domain guided by local multidimensional stacking

ABSTRACTWe present a new approach to enhancing weak prestack reflection signals without sacrificing higher frequencies. As a first step, we employ known multidimensional local stacking to obtain an approximate ‘model of the signal’. Guided by phase spectra from this model, we can detect very weak signals and make them visible and coherent by ‘repairing’ corrupted phase of original data. Both presented approaches – phase substitution and phase sign corrections – show good performance on complex synthetic and field data suffering from severe near‐surface scattering where conventional processing methods are rendered ineffective. The methods are mathematically formulated as a special case of time‐frequency masking (common in speech processing) combined with the signal model from local stacking. This powerful combination opens the avenue for a completely new family of approaches for multi‐channel seismic processing that can address seismic processing of land data with nodes and single sensors in the desert environment.

Reliability of data-driven wavefront attributes in laterally heterogeneous media

3D wavefront attributes play a major role in many processing steps, such as prestack data enhancement, diffraction separation, and wavefront tomography. For the determination of the 3D wavefront attributes, various stacking operators can be used by adopting semblance optimization. These operators are derived for laterally homogeneous media. In praxis, however, they are applied in real geologic environments with even strong lateral velocity variations such as salt structures. This leads to the question of the quality of the 3D wavefront attributes using these operators when determined in the presence of strong lateral velocity changes. We compared the 3D wavefront attributes determined by 3D common-reflection-surface (CRS) operator (called data-driven wavefront attributes) with the 3D wavefront attributes computed by 3D kinematic and dynamic ray tracing (called model-driven wavefront attributes). For the determination of the 3D CRS wavefront attributes, we have developed a global optimization scheme based on differential evolution. Reflection seismic data of the laterally heterogeneous 3D SEG C3WA salt model are considered, and the model-driven wavefront attributes are computed for a smoothed version of the 3D SEG salt model. The comparison reveals that the wavefront attributes for the normal-incidence-point ray indicate a very good match not only in areas of mild lateral velocity variation but even in regions with strong lateral velocity variations. Approximately 80%–90% of the total picks indicate the good match with a relative error of less than 10% when a semblance threshold of 0.1 is considered in the automatic picking process. This confirms the validity of the determined wavefront attributes even in the presence of strong lateral velocity changes. Using a higher semblance threshold in the automatic picking leads to fewer picks but with an even better match between model- and data-driven wavefront attributes.

Determination of wavefront attributes by differential evolution in the presence of conflicting dips

The common-reflection surface (CRS) method represents a multidimensional stacking approach; i.e., the stacking surface is determined in the midpoint and offset directions. In the 2D case, three attributes span the stacking surface, thus requiring a three-parameter search contrary to a one-parameter search in the classic common-midpoint stack. However, CRS wavefront attributes use data redundancy in the midpoint direction as well, which makes them very useful in several seismic applications, e.g., data preconditioning, velocity model building, and migration. Contrary to previous works, we simultaneously estimate CRS attributes using differential evolution in subcubes of the 3D search space. Differential evolution is a global optimization technique that performs particularly well when the objective function is unknown. Because we apply DE for each subcube, we could find local maxima, additionally to the global maximum. Therefore, conflicting dips are recognized and can be used for the stack and subsequent CRS attribute-based processing, which has been an issue in the past. Our land data results from the Donbas Foldbelt in southeast Ukraine demonstrate that our method reduces coherent steep dipping noise and reveal more subsurface structures. Application of the CRS attributes for prestack data enhancement shows that velocity analysis can be carried out more reliably.

Enhancement of prestack diffraction data and attributes using a traveltime decomposition approach

Diffractions not only carry important information about small-scale subsurface structures, they also possess unique properties, which make them a powerful tool for seismic processing and imaging. Since a point diffractor scatters an incoming wave to all directions, a diffraction event implies better illumination than a reflection, because the rays travel through larger parts of the subsurface. Furthermore, unlike the reflection case, in which the emergence location of the reflected wave depends on the source position, in the case of non-Snell scattering, up-going and down-going raypaths are decoupled. Based on this decoupling, we introduce a diffraction traveltime decomposition principle, which establishes a direct connection between zero-offset and finite-offset diffraction wavefield attributes. By making use of this approach, we are able to enhance diffractions and obtain high-quality diffraction wavefield attributes at arbitrary offsets in the prestack domain solely based on zero-offset processing without any further optimization of attributes. We show the accuracy of the method by fitting diffraction traveltimes, and on simple waveform data. Application to complex synthetic data shows the ability of the proposed approach to enhance diffractions and provide high-quality wavefield attributes even in sparsely illuminated regions such as subsalt areas. The promising results reveal a high potential for improved prestack data enhancement and further applications such as efficient diffraction-based finite-offset tomography.

From time to depth with CRS attributes

A processing workflow was introduced for reflection seismic data that is based entirely on common-reflection-surface (CRS) stacking attributes. This workflow comprises the CRS stack, multiple attenuation, velocity model building, prestack data enhancement, trace interpolation, and data regularization. Like other methods, its limitation is the underlying hyperbolic assumption. The CRS workflow provides an alternative processing path in case conventional common midpoint (CMP) processing is unsatisfactory. Particularly for data with poor signal-to-noise ratio and low-fold acquisition, the CRS workflow is advantageous. The data regularization feature and the ability of prestack data enhancement provide quality control in velocity model building and improve prestack depth-migrated images.