Near-surface Velocity Model Research Articles

Active source seismic imaging on near-surface granite body: case study of siting a geological disposal repository for high-level radioactive nuclear waste

In order to research whether it is suitable to set a geological disposal repository for high-level radioactive nuclear waste into one target granite body, two active source seismic profiles were arranged near a small town named Tamusu, Western China. The study area is with complex surface conditions, thus the seismic exploration encountered a variettraveltimey of technical difficulties such as crossing obstacles, de-noising harmful scattered waves, and building complex near-surface velocity models. In order to address those problems, techniques including cross-obstacle seismic geometry design, angle-domain harmful scattered noise removal, and an acoustic wave equation-based inversion method jointly utilizing both the and waveform of first arrival waves were adopted. The final seismic images clearly exhibit the target rock’s unconformable contact boundary and its top interface beneath the sedimentary and weathered layers. On this basis, it could be confirmed that the target rock is not thin or has been transported by geological process from somewhere else, but a native and massive rock. There are a few small size fractures whose space distribution could be revealed by seismic images within the rock. The fractures should be kept away. Based on current research, it could be considered that active source seismic exploration is demanded during the sitting process of the geological disposal repository for nuclear waste. The seismic acquisition and processing techniques proposed in the present paper would offer a good reference value for similar researches in the future.

Memory‐efficient frequency‐domain Gauss–Newton method for wave‐equation first‐arrival traveltime inversion

ABSTRACTWave‐equation traveltime inversion (WTI) can be used to automatically obtain a background near‐surface velocity model (NSM), which overcomes the high‐frequency approximation in ray theory. It is generally implemented in the time domain. However, the commonly used gradient‐based optimisation methods (such as the steepest‐descent method) in WTI have a low convergence rate and may yield less accurate results within limited iterations in geologically complex regions. To increase the convergence rate and improve the inversion accuracy, we propose a frequency‐domain truncated Gauss–Newton first‐arrival wave‐equation traveltime inversion (GN‐WTI) method to retrieve the background NSM. As only a few frequencies are used for inversion, the proposed frequency‐domain WTI method significantly reduces the computational memory requirements by more than two orders of magnitude in comparison with the conventional time‐domain WTI method. Therefore, the proposed method is especially advantageous for the building of large three‐dimensional models. In this GN‐WTI method, according to the derived explicit traveltime residual kernel, the gradient and Hessian vector products can be computed efficiently using an elegant and improved scattering integral approach as long as the source‐side wavefields and non‐redundant receiver‐side Green's functions are computed and stored in advance. The conjugate gradient approach is used to solve the Gauss–Newton normal equation to obtain the Gauss–Newton direction in inner loops. Here, the Gauss–Newton Hessians of the ray‐based traveltime inversion and WTI are compared to demonstrate the advantages of WTI. The trial runs with a simple periodic velocity model example showed that the proposed GN‐WTI method outperforms the WTI method when using the steepest‐descent and limited‐memory Broyden–Fletcher–Goldfarb–Shanno approaches in terms of the convergence rate and inversion accuracy. A complex Marmousi model was further used to illustrate the effectiveness of GN‐WTI. The proposed method should be beneficial in near‐surface velocity model building.

Automatic picking of multi-mode surface-wave dispersion curves based on machine learning clustering methods

The surface-wave analysis method is widely adopted to build near-surface shear-wave velocity structure. Automatic picking of surface-wave dispersion curves is a problem that needs to be solved, especially when dealing with a large amount of data. In this paper, we proposed an automatic picking method of multi-mode surface-wave dispersion curves based on unsupervised machine learning methods. Firstly, seismic data is transformed to two-dimensional dispersion image by dispersion imaging method. The discrete dispersion image points are separated into two clusters: dispersion energy and background noise by Gaussian mixture model clustering method. Then the dispersion energy points are projected into frequency-velocity domain, and are distinguished as different surface-wave modes by the density-based spatial clustering of applications with noise algorithm. Finally, multi-mode dispersion curves are obtained by searching for the local peaks in different mode areas of surface waves. Particle filter is used to smooth the picked dispersion curves to eliminate the influence of noise. The synthetic tests show that the automatically picked dispersion curves can match the theoretical dispersion curves well. We also applied the proposed method to the picking of dispersion curves of two field data. The automatically picked dispersion curves are applied to the subsequent inversion. In the second field data, the one-dimensional inversion results are interpolated to obtain the near-surface three-dimensional velocity model of the work area. The inversion results are in good agreement with borehole data, which further proves the accuracy and efficiency of the automatic picking method for multi-mode dispersion curves.

Selected aspects of modern seismic imaging and near-surface velocity model building in the area of Carpathian fold and thrust belt

Despite the increasing technological level of the reflection seismic method, the imaging of fold and thrust belts remains a demanding task, and usually leaves some questions regarding the dips, the shape of the subthrust structures or the most correct approach to velocity model building. There is no straightforward method that can provide structural representation of the near-surface geological boundaries and their velocities. The in-terpretation of refracted waves frequently remains the only available technique that may be used for this purpose, although one must be aware of its limitations which appear in the complex geological settings. In the presented study, the analysis of velocity values obtained in the shallow part of Carpathian orogenic wedge by means of various geophysical methods was carried out. It revealed the lack of consistency between the results of 3D refraction tomography and both the sonic log and uphole velocities. For that reason, instead of the indus-try-standard utilization of tomography, a novel, geologically-consistent method of velocity model building is pro-posed. In the near-surface part, the uphole velocities are assigned to the formations, documented by the surface geologic map. Interpreted time-domain horizons, supplemented by main thrusts, are used to make the velocity field fully-compatible with the litho-stratigraphic units of the Carpathians. The author demonstrates a retrospective overview of seismic data imaging in the area of the Polish Carpathian orogenic wedge and discusses the most recent global innovations in seismic methodology which are the key to successful hydrocarbon exploration in fold and thrust regions.

A robust first-arrival picking workflow using convolutional and recurrent neural networks

A deep-learning-based workflow is proposed in this paper to solve the first-arrival picking problem for near-surface velocity model building. Traditional methods, such as the short-term average/long-term average method, perform poorly when the signal-to-noise ratio is low or near-surface geologic structures are complex. This challenging task is formulated as a segmentation problem accompanied by a novel postprocessing approach to identify pickings along the segmentation boundary. The workflow includes three parts: a deep U-net for segmentation, a recurrent neural network (RNN) for picking, and a weight adaptation approach to be generalized for new data sets. In particular, we have evaluated the importance of selecting a proper loss function for training the network. Instead of taking an end-to-end approach to solve the picking problem, we emphasize the performance gain obtained by using an RNN to optimize the picking. Finally, we adopt a simple transfer learning scheme and test its robustness via a weight adaptation approach to maintain the picking performance on new data sets. Our tests on synthetic data sets reveal the advantage of our workflow compared with existing deep-learning methods that focus only on segmentation performance. Our tests on field data sets illustrate that a good postprocessing picking step is essential for correcting the segmentation errors and that the overall workflow is efficient in minimizing human interventions for the first-arrival picking task.

Novel strategies for complex foothills seismic imaging — Part 1: Mega-near-surface velocity estimation

Seismic imaging in foothills areas is challenging because of the complexity of the near-surface and subsurface structures. Single seismic surveys often are not adequate in a foothill-exploration area, and multiple phases with different acquisition designs within the same block are required over time to get desired sampling in space and azimuths for optimizing noise attenuation, velocity estimation, and migration. This is partly because of economic concerns, and it is partly because technology is progressing over time, creating the need for unified criteria in processing workflows and parameters at different blocks in a study area. Each block is defined as a function of not only location but also the acquisition and processing phase. An innovative idea for complex foothills seismic imaging is presented to solve a matrix of blocks and tasks. For each task, such as near-surface velocity estimation and static corrections, signal processing, prestack time migration, velocity-model building, and prestack depth migration, one or two best service companies are selected to work on all blocks. We have implemented streamlined processing efficiently so that Task-1 to Task-n progressed with good coordination. Application of this innovative approach to a mega-project containing 16 3D surveys covering more than [Formula: see text] in the Kelasu foothills, northwestern China, has demonstrated that this innovative approach is a current best practice in complex foothills imaging. To date, this is the largest foothills imaging project in the world. The case study in Kelasu successfully has delivered near-surface velocity models using first arrivals picked up to 3500 m offset for static corrections and 9000 m offset for prestack depth migration from topography. Most importantly, the present megaproject is a merge of several 3D surveys, with the merge performed in a coordinated, systematic fashion in contrast to most land megaprojects. The benefits of this approach and the strategies used in processing data from the various subsurveys are significant. The main achievement from the case study is that the depth images, after the application of the near-surface velocity model estimated from the megasurveys, are more continuous and geologically plausible, leading to more accurate seismic interpretation.

Three-dimensional data-domain full traveltime inversion using a practical workflow of early arrival selection

A new method of data-domain full traveltime inversion (FTI) is proposed to estimate the near-surface velocity model using early arrivals in seismic shot gathers. Data-domain FTI is capable of generating a background velocity model from which the predicted early arrivals can kinematically match the observed ones. Such a match is measured and quantified in terms of the crosscorrelation function between the computed and observed traces. Our method aims to find an optimal estimated velocity model that minimizes the crosscorrelation computed from the selected early arrivals. The early arrivals are isolated via a sequence of operations, including the [Formula: see text]-[Formula: see text] scan, autopicking, multidomain quality control, and guide interpolation. Because windows, rather than exact arrival times, are constructed, the difficulties encountered while picking precise arrivals are reduced. In addition, the gradient of data-domain FTI is derived based on an amplitude-constrained optimization problem, which makes the gradient essentially different from that derived with the Born approximation in which no constraint is used. The constraint requires the inversion to honor traveltime information only, and it thus ignores any amplitude changes caused by velocity variations. This method is validated using 3D synthetic as well as field data sets. The results show that data-domain FTI, combined with the early arrival selection workflow, is able to generate reasonable background velocities that kinematically match the predicted early arrivals with the observed ones, and the associated depth-domain images are clearly improved.

Random objective waveform inversion of surface waves

Full-waveform inversion (FWI) of surface waves is becoming increasingly popular among shallow-seismic methods. Due to a huge amount of data and the high nonlinearity of the objective function, FWI usually requires heavy computational costs and may converge toward a local minimum. To mitigate these problems, we have reformulated FWI under a multiobjective framework and adopted a random objective waveform inversion (ROWI) method for surface-wave characterization. Three different measure functions were used, whereas the combination of one measure function with one shot independently provided one of the [Formula: see text] objective functions ([Formula: see text] is the total number of shots). We have randomly chose and optimized one objective function at each iteration. We performed a synthetic test to compare the performance of the ROWI and conventional FWI approaches, which showed that the convergence of ROWI is faster and more robust compared with conventional FWI approaches. We also applied ROWI to a field data set acquired in Rheinstetten, Germany. ROWI successfully reconstructed the main geologic feature, a refilled trench, in the final result. The comparison between the ROWI result and a migrated ground-penetrating radar profile further proved the effectiveness of ROWI in reconstructing the near-surface S-wave velocity model. We also ran the same field example by using a poor initial model. In this case, conventional FWI failed whereas ROWI still reconstructed the subsurface model to a fairly good level, which highlighted the relatively low dependency of ROWI on the initial model.

Introduction to this special section: Near-surface imaging and modeling

This special section on near-surface imaging and modeling was intended originally to focus on improving deeper imaging for exploration purposes through more accurate representations of the near surface, the highly variable zone that energy must traverse through on the way down and back up again to be recorded at the surface. However, as proposed manuscript topics started coming in, it became clear that this section would cover a wider range, from kilometers down to meters. Papers in this section highlight a range of near-surface-related work that includes applying full-waveform inversion (FWI) to improve near-surface velocity models, identifying potential sinkhole hazards before they collapse, the potential of smartphones as geophysical sensors, and new open-source software for ground-penetrating radar data.

A Layer-cell Tomography Method for Near-surface Velocity Model Building Using First Arrivals

Variations in the topography and thickness of the near-surface low-velocity weathering zone negatively impact the quality of onshore seismic data and images. Such impact can be mitigated using accurate near-surface velocity models. We propose to estimate near-surface velocities using a layer-cell tomography method which sequentially combines the model parameterizations of multiscale deformable-layer tomography (DLT) and cell tomography. We first estimate the long-wavelength velocity variations using multiscale DLT, which maps the undulating geometry of the weathering base where the angular coverage of first-arrival raypaths is limited. Then we convert the solution model of multiscale DLT into a cell model, and further determine the short-wavelength velocity variations using cell tomography. We test the proposed layer-cell tomography method using synthetic datasets and a field dataset. The results indicate that the new method improves the resolution of the velocity model from the multiscale DLT, and reduces the artifacts in the velocity solution compared to those observed in the cell tomography method. Reverse time migration of the synthetic data shows that using the velocity model from the layer-cell tomography method gives more accurate images of reflectors. For the field dataset test, the layer-cell tomography method yields a high-resolution near-surface velocity model that could be validated by the improvements in the tomostatic correction results.

Near-real-time near-surface 3D seismic velocity and uncertainty models by wavefield gradiometry and neural network inversion of ambient seismic noise

With the advent of large and dense seismic arrays, novel, cheap, and fast imaging and inversion methods are needed to exploit the information captured by stations in close proximity to each other and produce results in near real time. We have developed a sequence of fast seismic acquisition for dispersion curve extraction and inversion for 3D seismic models, based on wavefield gradiometry, wave equation inversion, and machine-learning technology. The seismic array method that we use is Helmholtz wave equation inversion using measured wavefield gradients, and the dispersion curve inversions are based on a mixture of density neural networks (NNs). For our approach, we assume that a single surface wave mode dominates the data. We derive a nonlinear relationship among the unknown true seismic wave velocities, the measured seismic wave velocities, the interstation spacing, and the noise level in the signal. First with synthetic and then with the field data, we find that this relationship can be solved for unknown true seismic wave velocities using fixed point iterations. To estimate the noise level in the data, we need to assume that the effect of noise varies weakly with the frequency and we need to be able to calibrate the retrieved average dispersion curves with an alternate method (e.g., frequency wavenumber analysis). The method is otherwise self-contained and produces phase velocity estimates with tens of minutes of noise recordings. We use NNs, specifically a mixture density network, to approximate the nonlinear mapping between dispersion curves and their underlying 1D velocity profiles. The networks turn the retrieved dispersion model into a 3D seismic velocity model in a matter of seconds. This opens the prospect of near-real-time near-surface seismic velocity estimation using dense (and potentially rolling) arrays and only ambient seismic energy.

Ambient Noise Tomography With Common Receiver Clusters in Distributed Sensor Networks

Near-surface imaging with distributed sensor networks (DSN) is promising for planet exploration, which affordably generates a near-surface velocity model. Recently, an Eikonal tomography-based ambient noise seismic imaging (ANSI) algorithm was implemented in a DSN to realize real-time and in-situ near-surface imaging. However, only using data from neighbors to generate a velocity map cannot have enough stacking samples to generate high-quality results. Also, the neighbor range increase will result in the exponential rise of communication costs. To overcome this problem, we propose a new decentralized Eikonal tomography algorithm in the DSN. The main idea is to change the source-based algorithm to a receiver-based one, which we call common receiver decentralized Eikonal tomography (CR-TomoEK). With CR-TomoEK, nodes fully utilize signals from neighbors to generate partial velocity maps, when combined, lead to the final output. When compared with the original Eikonal algorithm, the stacking number is significantly increased, output quality is higher than before, and there is a significant reduction in communication cost. We performed experiments on both synthetic data and real data from the USArray Transportable Array. Both imaging quality and communication cost are considered in the algorithm validation. The result shows that our algorithm significantly increases the output quality while keeping the communication cost safe to generate a real-time result.

FWI as an effective solution for land near-surface model building into the area with complex geological settings: Eastern Siberia case study

Seismic imaging in the Eastern Siberia region is challenging due to the presence of complex overburden anomalies such as magmatic intrusions and thrusts in the suprasalt complex leading to significant vertical and lateral property variations of high-velocity halite-carbonate geological sections with multiple velocity inversions, complex shallow stratigraphy and substantial topography. Such geological settings create amplitude dimming and wipeouts in the images and cause structural distortions at the reservoir level. We applied full-waveform inversion (FWI) technology to overcome limitations of the conventional workflow and to support the exploration programme. The primary output of FWI is a high-resolution subsurface velocity model, which can be used as an attribute to improve interpretation of the overburden sediments, to identify shallow hazards, to derisk exploration prospects and to optimize well placement. We tested FWI on a conventional 3D land data set to create a near-surface velocity model. Detailed subsurface models generated with FWI provided an accurate description of the overburden and helped us to investigate shallow geological environments. Predrill predictions using the FWI velocity volume showed excellent agreement with well measurements and made it possible to recommend the technology for inclusion in the processing flow.

Two-grid full-waveform Rayleigh-wave inversion via a genetic algorithm — Part 1: Method and synthetic examples

When reliable a priori information is not available, it is difficult to correctly predict near-surface S-wave velocity models from Rayleigh waves through existing techniques, especially in the case of complex geology. To tackle this issue, we have developed a new method: two-grid genetic-algorithm Rayleigh-wave full-waveform inversion (FWI). Adopting a two-grid parameterization of the model, the genetic algorithm inverts for unknown velocities and densities at the nodes of a coarse grid, whereas the forward modeling is performed on a fine grid to avoid numerical dispersion. A bilinear interpolation brings the coarse-grid results into the fine-grid models. The coarse inversion grid allows for a significant reduction in the computing time required by the genetic algorithm to converge. With a coarser grid, there are fewer unknowns and less required computing time, at the expense of the model resolution. To further increase efficiency, our inversion code can perform the optimization using an offset-marching strategy and/or a frequency-marching strategy that can make use of different kinds of objective functions and allows for parallel computing. We illustrate the effect of our inversion method using three synthetic examples with rather complex near-surface models. Although no a priori information was used in all three tests, the long-wavelength structures of the reference models were fairly predicted, and satisfactory matches between “observed” and predicted data were achieved. The fair predictions of the reference models suggest that the final models estimated by our genetic-algorithm FWI, which we call macromodels, would be suitable inputs to gradient-based Rayleigh-wave FWI for further refinement. We also explored other issues related to the practical use of the method in different work and explored applications of the method to field data.

Land FWI in the Delaware Basin, west Texas: A case study

As part of a wide-reaching full-waveform inversion (FWI) research program, FWI is applied to an onshore seismic data set collected in the Delaware Basin, west Texas. FWI is routinely applied on typical marine data sets with high signal-to-noise ratio (S/N), relatively good low-frequency content, and reasonably long offsets. Land seismic data sets, in comparison, present significant challenges for FWI due to low S/N, a dearth of low frequencies, and limited offsets. Recent advancements in FWI overcome limitations due to poor S/N and low frequencies making land FWI feasible to use to update the shallow velocities. The chosen area has contrasting and variable near-surface conditions providing an excellent test data set on which to demonstrate the workflow and its challenges. An acoustic FWI workflow is used to update the near-surface velocity model in order to improve the deeper image and simultaneously help highlight potential shallow drilling hazards.

Local site effects estimation at Amatrice (Central Italy) through seismological methods

We present the results of seismological and geophysical investigations performed by the “Istituto Nazionale di Geofisica e Vulcanologia” team operating in Amatrice village (Central Italy), in the emergency phases following the Mw 6.0 event of August 24th 2016, that caused severe damage in downtown and surrounding areas. Data from seven seismic stations equipped with both weak and strong motion sensors are analyzed in terms of standard spectral ratio to empirically define amplification function using a bedrock reference site. Ambient vibration spectral ratios between horizontal and vertical component of motion are also evaluated in a large number of sites, spread out in the investigated area, to recover the resonance frequency of the soft soil outcropping layers and to generalize the results obtained by earthquake data. Ambient noise vibration are also used for applying a 2D array approach based on surface waves techniques in order to define the near-surface velocity model and to verify its lateral variation. The results allows to better understand the amplification factors in the investigated area, showing spatial variation of site effects despite of the homogeneous shallow geological condition indicated by the microzonation studies available at moment of the described field campaign. The analysis reveals a diffuse amplification effect which reaches its maximum values in downtown area with a resonant frequency of about 2 Hz. The obtained results were used to integrate the microzonation studies and they can be used for urban planning and reconstruction activities.

Full Waveform Inversion in generalized coordinates for zones of curved topography

Full waveform inversion (FWI) has been recently used to estimate subsurface parameters, such as velocity models. This method, however, has a number of drawbacks when applied to zones with rugged topography due to the forced application of a Cartesian mesh on a curved surface. In this work, we present a simple coordinate transformation that enables the construction of a curved mesh. The proposed transformation is more suitable for rugged surfaces and it allows mapping a physical curved domain into a uniform rectangular grid, where acoustic FWI can be applied in the traditional way by introducing a modified Laplacian. We prove that the proposed approximation can have a wide range of applications, producing precise near-surface velocity models without increasing the computing time of the FWI.

Robust refraction statics solution and near-surface velocity model building using feedback from reflection data

An accurate near-surface velocity model is critical for weathering statics correction and initial model building for depth migration and full-waveform inversion. However, near-surface models from refraction inversion often suffer from errors in refraction data, insufficient sampling, and over-simplified assumptions used in refraction algorithms. Errors in refraction data can be caused by picking errors resulting from surface noise, attenuation, and dispersion of the first-arrival energy with offset. These errors are partially compensated later in the data flow by reflection residual statics. Therefore, surface-consistent residual statics contain information that can be used to improve the near-surface velocity model. We have developed a new dataflow to automatically include median and long-wavelength components of surface-consistent reflection residual statics. This technique can work with any model-based refraction solution, including grid-based tomography methods and layer-based methods. We modify the cost function of the refraction inversion by adding model and data weights computed from the smoothed surface-consistent residual statics. By using an iterative inversion, these weights allow us to update the near-surface velocity model and to reject first-arrival picks that do not fit the updated model. In this nonlinear optimization workflow, the refraction model is derived from maximizing the coherence of the reflection energy and minimizing the misfit between model arrival times and the recorded first-arrival times. This approach can alleviate inherent limitations in shallow refraction data by using coherent reflection data.

Estimating S-wave velocities from 3D 9-component shallow seismic data using local Rayleigh-wave dispersion curves – A field study

Surface waves are widely used to delineate near-surface structures. Most of the surface-wave studies focus on 2D and single-component data, how to use the surface-wave method on 3D multi-component data has not been studied. In this paper, we perform a 3D 9-component (9C) shallow-seismic survey in Rheinstetten, Germany. We transform the 9C data from the cartesian coordinates to the ray-based coordinates and separate Love and Rayleigh waves in the transformed data. The data is windowed into groups constituted by 16 receiver points, and we perform dispersion analysis in each windowed data individually. Rayleigh-wave energy acquired by four different components in each windowed data is stacked in the frequency-velocity domain to improve the accuracy of the dispersion image, and the Rayleigh-wave dispersion curve is picked along the energy peak in the stacked dispersion image. We estimate a total of 242 Rayleigh-wave dispersion curves from the 3D 9C data, which are then inverted to build a 3D S-wave velocity model. The 3D S-wave velocity model shows the existence as well as the spatial variation of the main geological feature, namely a refilled trench from the 18th century. We compare 2D sections of the 3D model to the GPR profiles acquired in the same survey area, which proves the relatively high accuracy of the structures in the estimated model. We also compare the observed seismic waveforms to the synthetic waveforms simulated by using the 3D model, which also proves that the estimated model is fairly reliable. This study provides a cost-efficient way to use surface waves in 3D multi-component data to image the near-surface S-wave velocity model.

Near-surface velocity estimation using source-domain full traveltime inversion and early arrival waveform inversion

Accurate estimation of near-surface velocity is a key step for imaging deeper targets. We have developed a new workflow to invert complex early arrivals in land seismic data for near-surface velocities. This workflow is composed of two methods: source-domain full traveltime inversion (FTI) and early arrival waveform inversion (EWI). Source-domain FTI automatically generates the background velocity that kinematically matches the reconstructed plane-wave sources from early arrivals with true plane-wave sources. This method does not require picking first arrivals for inversion, which is one of the most challenging and labor-intensive steps in ray-based first-arrival traveltime tomography, especially when the subsurface medium contains low-velocity zones that cause shingled multivalue arrivals. Moreover, unlike the conventional Born-based method, source-domain FTI can determine if the initial velocity is slower or faster than the true one according to the gradient sign. In addition, the computational cost is reduced considerably by using the one-way wave equation to extrapolate the plane-wave Green’s function. The source-domain FTI tomogram is then used as the starting model for EWI to obtain the short-wavelength component associated with the velocity model. We tested the workflow on two synthetic and one onshore filed data sets. The results demonstrate that source-domain FTI generates reasonable background velocities for EWI even though the first arrivals are shingled, and that this workflow can produce a high-resolution near-surface velocity model.