Time-to-depth conversion and velocity estimation by image-wavefront propagation

Geophysics ◽  
2017 ◽  
Vol 82 (6) ◽  
pp. U75-U85 ◽  
Author(s):  
Leandro da S. Sadala Valente ◽  
Henrique B. Santos ◽  
Jessé C. Costa ◽  
Jörg Schleicher
Keyword(s):  
Ray Tracing ◽  
Model Building ◽  
Velocity Model ◽  
Velocity Estimation ◽  
Time Step ◽  
Velocity Anomaly ◽  
Interval Velocity ◽  
Velocity Models ◽  
Wavefront Propagation ◽  
Velocity Model Building

A new strategy for time-to-depth conversion and interval-velocity estimation is based entirely on image-wavefront propagation without the need to follow individual image rays. The procedure has three main features: (1) It computes the velocity field and the traveltime directly, allowing us to dispense with dynamic ray tracing; (2) it requires only the knowledge of the image wavefront at the previous time step; and (3) it inherently smooths the image wavefront, inhibiting the formation of caustics. As a consequence, the method tends to be faster than the usual techniques and does not carry the constraints and limitations inherent to common ray-tracing strategies. Synthetic tests using a Gaussian velocity anomaly as well as the Marmousi velocity model, and two smoothed versions of it show the feasibility of the method. A field-data example demonstrates the use of different numerical procedures. Our results indicate that the present strategy can be used to construct reasonable depth-velocity models that can be used as reliable starting models for velocity-model building in depth migration or for tomographic methods.

scholarly journals Seismic velocity estimation: A deep recurrent neural-network approach

Geophysics ◽  
2019 ◽  
Vol 85 (1) ◽  
pp. U21-U29
Author(s):  
Gabriel Fabien-Ouellet ◽  
Rahul Sarkar
Keyword(s):  
Neural Network ◽  
Neural Networks ◽  
Recurrent Neural Networks ◽  
Model Building ◽  
Seismic Velocity ◽  
Velocity Model ◽  
Velocity Estimation ◽  
Interval Velocity ◽  
Velocity Model Building ◽  
3D Velocity Model

Applying deep learning to 3D velocity model building remains a challenge due to the sheer volume of data required to train large-scale artificial neural networks. Moreover, little is known about what types of network architectures are appropriate for such a complex task. To ease the development of a deep-learning approach for seismic velocity estimation, we have evaluated a simplified surrogate problem — the estimation of the root-mean-square (rms) and interval velocity in time from common-midpoint gathers — for 1D layered velocity models. We have developed a deep neural network, whose design was inspired by the information flow found in semblance analysis. The network replaces semblance estimation by a representation built with a deep convolutional neural network, and then it performs velocity estimation automatically with recurrent neural networks. The network is trained with synthetic data to identify primary reflection events, rms velocity, and interval velocity. For a synthetic test set containing 1D layered models, we find that rms and interval velocity are accurately estimated, with an error of less than [Formula: see text] for the rms velocity. We apply the neural network to a real 2D marine survey and obtain accurate rms velocity predictions leading to a coherent stacked section, in addition to an estimation of the interval velocity that reproduces the main structures in the stacked section. Our results provide strong evidence that neural networks can estimate velocity from seismic data and that good performance can be achieved on real data even if the training is based on synthetics. The findings for the 1D problem suggest that deep convolutional encoders and recurrent neural networks are promising components of more complex networks that can perform 2D and 3D velocity model building.

Practical approaches for subsalt velocity model building

Geophysics ◽  
2008 ◽  
Vol 73 (5) ◽  
pp. VE183-VE194 ◽  
Author(s):  
Junru Jiao ◽  
David R. Lowrey ◽  
John F. Willis ◽  
Ruben D. Martínez
Keyword(s):  
Software Package ◽  
Model Building ◽  
Velocity Model ◽  
Fine Tuning ◽  
Inherent Difficulty ◽  
Velocity Models ◽  
First Approximation ◽  
Velocity Model Building ◽  
Trade Offs ◽  
Efficient Performance

Imaging sediments below salt bodies is challenging because of the inherent difficulty of estimating accurate velocity models. These models can be estimated in a variety of ways with varying degrees of expense and effectiveness. Two methods are commercially viable trade-offs. In the first method, residual-moveout analysis is performed in a layer-stripping mode. The models produced with this method can be used as a first approximation of the subsalt velocity field. A wave-equation migration scanning technique is more suitable for fine-tuning the velocity model below the salt. Both methods can be run as part of a sophisticated interactive velocity interpretation software package that makes velocity interpretation efficient. Performance of these methods has been tested on synthetic and field data examples.

Velocity estimation by beam stack

Geophysics ◽  
1992 ◽  
Vol 57 (8) ◽  
pp. 1034-1047 ◽  
Author(s):  
Biondo Biondi
Keyword(s):  
Estimation Method ◽  
Velocity Model ◽  
Velocity Estimation ◽  
Gradient Algorithm ◽  
Stratified Medium ◽  
Detailed Knowledge ◽  
Low Velocity ◽  
Velocity Anomaly ◽  
Interval Velocity ◽  
Lateral Variations

Imaging seismic data requires detailed knowledge of the propagation velocity of compressional waves in the subsurface. In conventional seismic processing, the interval velocity model is usually derived from stacking velocities. Stacking velocities are determined by measuring the coherency of the reflections along hyperbolic moveout trajectories in offset. This conventional method becomes inaccurate in geologically complex areas because the conversion of stacking velocities to interval velocities assumes a horizontally stratified medium and mild lateral variations in velocity. The tomographic velocity estimation proposed in this paper can be applied when there are dipping reflectors and strong lateral variations. The method is based on the measurements of moveouts by beam stacks. A beam stack measures local coherency of reflections along hyperbolic trajectories. Because it is a local operator, the beam stack can provide information on nonhyperbolic moveouts in the data. This information is more reliable than traveltimes of reflections picked directly from the data because many seismic traces are used for computing beam stacks. To estimate interval velocity, I iteratively search for the velocity model that best predicts the events in beam‐stacked data. My estimation method does not require a preliminary picking of the data because it directly maximizes the beam‐stack’s energy at the traveltimes and surface locations predicted by ray tracing. The advantage of this formulation is that detection of the events in the beam‐stacked data can be guided by the imposition of smoothness constraints on the velocity model. The optimization problem of maximizing beam‐stack energy is solved by a gradient algorithm. To compute the derivatives of the objective function with respect to the velocity model, I derive a linear operator that relates perturbations in velocity to the observed changes in the beam‐stack kinematics. The method has been successfully applied to a marine survey for estimating a low‐velocity anomaly. The estimated velocity function correctly predicts the nonhyperbolic moveouts in the data caused by the velocity anomaly.

Seismic Velocity Model Building Using Neural Networks: Training Data Design and Learning Generalization

Geophysics ◽  
2021 ◽  
pp. 1-73
Author(s):  
Hani Alzahrani ◽  
Jeffrey Shragge
Keyword(s):  
Seismic Data ◽  
Model Building ◽  
Seismic Velocity ◽  
Velocity Model ◽  
Training Data ◽  
Training Process ◽  
Training Models ◽  
Velocity Models ◽  
Velocity Model Building ◽  
Layer Velocity

Data-driven artificial neural networks (ANNs) offer a number of advantages over conventional deterministic methods in a wide range of geophysical problems. For seismic velocity model building, judiciously trained ANNs offer the possibility of estimating high-resolution subsurface velocity models. However, a significant challenge of ANNs is training generalization, which is the ability of an ANN to apply the learning from the training process to test data not previously encountered. In the context of velocity model building, this means learning the relationship between velocity models and the corresponding seismic data from a set of training data, and then using acquired seismic data to accurately estimate unknown velocity models. We ask the following question: what type of velocity model structures need be included in the training process so that the trained ANN can invert seismic data from a different (hypothetical) geological setting? To address this question, we create four sets of training models: geologically inspired and purely geometrical, both with and without background velocity gradients. We find that using geologically inspired training data produce models with well-delineated layer interfaces and fewer intra-layer velocity variations. The absence of a certain geological structure in training models, though, hinders the ANN's ability to recover it in the testing data. We use purely geometric training models consisting of square blocks of varying size to demonstrate the ability of ANNs to recover reasonable approximations of flat, dipping, and curved interfaces. However, the predicted models suffer from intra-layer velocity variations and non-physical artifacts. Overall, the results successfully demonstrate the use of ANNs in recovering accurate velocity model estimates, and highlight the possibility of using such an approach for the generalized seismic velocity inversion problem.

Illumination compensation for image-domain wavefield tomography

Geophysics ◽  
2013 ◽  
Vol 78 (5) ◽  
pp. U65-U76 ◽  
Author(s):  
Tongning Yang ◽  
Jeffrey Shragge ◽  
Paul Sava
Keyword(s):  
Model Building ◽  
Velocity Model ◽  
Model Errors ◽  
Image Domain ◽  
Illumination Compensation ◽  
Illumination Problem ◽  
Velocity Models ◽  
Velocity Model Building ◽  
Penalty Operator ◽  
Seismic Images

Image-domain wavefield tomography is a velocity model building technique using seismic images as the input and seismic wavefields as the information carrier. However, the method suffers from the uneven illumination problem when it applies a penalty operator to highlighting image inaccuracies due to the velocity model error. The uneven illumination caused by complex geology such as salt or by incomplete data creates defocusing in common-image gathers even when the migration velocity model is correct. This additional defocusing violates the wavefield tomography assumption stating that the migrated images are perfectly focused in the case of the correct model. Therefore, defocusing rising from illumination mixes with defocusing rising from the model errors and degrades the model reconstruction. We addressed this problem by incorporating the illumination effects into the penalty operator such that only the defocusing by model errors was used for model construction. This was done by first characterizing the illumination defocusing in gathers by illumination analysis. Then an illumination-based penalty was constructed that does not penalize the illumination defocusing. This method improved the robustness and effectiveness of image-domain wavefield tomography applied in areas characterized by poor illumination. Our tests on synthetic examples demonstrated that velocity models were more accurately reconstructed by our method using the illumination compensation, leading to a more accurate model and better subsurface images than those in the conventional approach without illumination compensation.

Interval velocity estimation via NMO-based differential semblance

Geophysics ◽  
2007 ◽  
Vol 72 (6) ◽  
pp. U75-U88 ◽  
Author(s):  
Jintan Li ◽  
William W. Symes
Keyword(s):  
Velocity Model ◽  
Velocity Estimation ◽  
Imaging Method ◽  
Velocity Analysis ◽  
Multiple Reflections ◽  
Coherent Noise ◽  
Interval Velocity ◽  
Velocity Models ◽  
Mean Square Difference ◽  
Rich Data

The differential semblance method of velocity analysis flattens image gathers automatically by updating interval velocity to minimize the mean square difference of neighboring traces. We detail an implementation using hyperbolic normal moveout correction as the imaging method. The algorithm is fully automatic, accommodates arbitrary acquisition geometry, and outputs 1D, 2D, or 3D interval velocity models. This variant of differential semblance velocity analysis is effective within the limits of its imaging methodology: mild lateral heterogeneity and data dominated by primary events. Coherent noise events such as multiple reflections tend to degrade the quality of the velocity model estimated by differential semblance. We show how to combine differential semblance velocity analysis with dip filtering to suppress multiple reflections and thus improve considerably the accuracy of the velocity estimate. We illustrate this possibility using multiple-rich data from a 2D marine survey.

scholarly journals Deep-learning inversion: A next-generation seismic velocity model building method

Geophysics ◽  
2019 ◽  
Vol 84 (4) ◽  
pp. R583-R599 ◽  
Author(s):  
Fangshu Yang ◽  
Jianwei Ma
Keyword(s):  
Deep Learning ◽  
Seismic Data ◽  
Model Building ◽  
Seismic Velocity ◽  
Velocity Model ◽  
Computational Time ◽  
Learning Methods ◽  
Velocity Models ◽  
Velocity Model Building ◽  
Training Stage

Seismic velocity is one of the most important parameters used in seismic exploration. Accurate velocity models are the key prerequisites for reverse time migration and other high-resolution seismic imaging techniques. Such velocity information has traditionally been derived by tomography or full-waveform inversion (FWI), which are time consuming and computationally expensive, and they rely heavily on human interaction and quality control. We have investigated a novel method based on the supervised deep fully convolutional neural network for velocity-model building directly from raw seismograms. Unlike the conventional inversion method based on physical models, supervised deep-learning methods are based on big-data training rather than prior-knowledge assumptions. During the training stage, the network establishes a nonlinear projection from the multishot seismic data to the corresponding velocity models. During the prediction stage, the trained network can be used to estimate the velocity models from the new input seismic data. One key characteristic of the deep-learning method is that it can automatically extract multilayer useful features without the need for human-curated activities and an initial velocity setup. The data-driven method usually requires more time during the training stage, and actual predictions take less time, with only seconds needed. Therefore, the computational time of geophysical inversions, including real-time inversions, can be dramatically reduced once a good generalized network is built. By using numerical experiments on synthetic models, the promising performance of our proposed method is shown in comparison with conventional FWI even when the input data are in more realistic scenarios. We have also evaluated deep-learning methods, the training data set, the lack of low frequencies, and the advantages and disadvantages of our method.

Velocity model building by semblance maximization of modulated-shot gathers

Geophysics ◽  
2007 ◽  
Vol 72 (5) ◽  
pp. U67-U73 ◽  
Author(s):  
Robert Soubaras ◽  
Bruno Gratacos
Keyword(s):  
Wave Equation ◽  
Model Building ◽  
Velocity Model ◽  
Multiscale Approach ◽  
Nonlinear Inversion ◽  
Velocity Models ◽  
Velocity Model Building ◽  
Frequency Maximum ◽  
The Cost ◽  
Starting Model

In recent years, wave-equation migration has greatly enhanced imaging in complex velocity models. However, velocity model building is still dependent on ray-theory approximations. We propose a full wave-equation methodology for velocity model building based on the nonlinear inversion of a semblance criterion with respect to the velocity field. A newly described type of migration, called the modulated-shot migration, is used to obtain the necessary gathers, which are indexed in surface angle. The semblance of these gathers, after spatial averaging, is used as the cost function. This methodology is shown to successfully image the Marmousi model and the subsalt part of the Sigsbee model, especially in terms of focusing, which is as good as with the true model, but also in terms of depthing which is enhanced compared with the initial model. Realistic constraints are used in terms of minimum frequency, maximum offset, and crudeness of the starting model. A key point in the success of this methodology is the multiscale approach wherein the iterations are started on a coarse scale, and ended at a finer scale.

Sign in / Sign up

Export Citation Format

Share Document