Acquiring economical land data with compressive sensing requires data reconstruction. In the presence of complex near-surface weathering layers, which on their own typically pose a challenge to processing densely sampled data, data reconstruction suffers. The conventional approach of near-surface correction followed by interpolation rely on knowledge of the subsurface. However, obtaining a velocity model is difficult from subsampled data influenced by the weathering layers. To avoid that, we propose to reconstruct the data with a model-independent rank-reduction-based near-surface correction followed by interpolation. We showcase the proposed reconstruction on synthetic data. A field data example will also be presented during the meeting to demonstrate the potential of the method.@en

The accuracy of a model obtained by full-waveform inversion can be estimated by analysing the sensitivity of the data to perturbations of the model parameters in selected subsurface points. Each perturbation requires the computation of the seismic response in the form of Born scattering data for a typically very large number of shots, making the method time consuming. The computational cost can be significantly reduced by considering the point where the subsurface parameters are perturbed as a Born scatterer. Instead of modelling each shot separately, reciprocity relations provide the Green functions from the sources to the scatterer in terms of Green’s functions from the scatterer to the sources. In this way, the Born scattering data from a single point in the isotropic elastic case for a marine acquisition with pressure sources and receivers can be expressed in terms of the Green functions for force and moment tensor sources located at the scatterer and only a small number of forward runs are required. A 2-D example illustrates how the result can be used to determine the hessian and local covariance matrix for the model parameters at the scatterer at the cost of 5 forward simulation.@en

This extended abstract discusses a novel approach to identify geological sub-seismic scale features using angle-gather spectrograms. Traditional seismic methods face challenges in identifying fine scale geological structures, especially in sublayers thinner than about 1/10 of the wavelength. This paper introduces the use of spectrogram analysis on angle gathers, to capture local spectrum information which is not visible and distinguishable in the reflectivity image and angle gathers themselves. The authors apply this technique to simulated seismic data for a range of basic geological scenarios, extracting spectrograms for each incident angle within angle gathers. The analysis explores the impact of density, velocity variations, layer number and layer thickness on spectrogram patterns, providing insights into their effectiveness for computer vision-related research. The study aims to rejuvenate the utilization of spectrograms for revealing hidden geological structures beyond standard seismic resolution.@en

Enhanced pre-stack depth migration, characterized by improved resolution and amplitudes, ensures a more accurate representation of the subsurface, proving essential for reducing the likelihood of geological misinterpretations and facilitating informed decision-making in seismic exploration. However, obtaining high-resolution images with preserved amplitudes through standard depth migration could face several hurdles known as migration artifacts. Iterative least-squares migration (LSM) was developed to address these migration artifacts. However, the convergence rate of LSM using a gradient descent approach tends to be slow. Several researchers have attempted to achieve computational efficiency in linearized LSM through gradient preconditioning. In the context of iterative least-squares wave-equation migration, this extended abstract compares three minimization approaches that differ in error functions and gradient preconditioning—including the depth-dependent Hessian approximation inverse—through two numerical examples, one with an inverse-crime scenario and the other with a non-inverse-crime scenario.@en

The phenomenon of wave conversions, where acoustic, pressure (P) waves are converted to elastic, shear (S) waves is commonly disregarded in seismic imaging, which can lead to lower-quality images in regions with strong reflectors. While a number of methods exist which do take wave conversions into account, most deal with P- and S-waves separately, rather than in using a single, unified, framework. Elastic Full-Waveform Inversion (FWI), on the other hand, which does offer a unified framework for all elastic effects, is prohibitively computationally expensive in many cases. We present an alternative approach by extending Full Wavefield Migration (FWM) to account for wave conversions. Full Wavefield Migration describes seismic data in terms of convolutional propagation and reflection operators in the space-frequency domain. By applying these operators recursively, multi-scattering data can be modelled and inverted. Using Shuey’s approximation to constrain the number of parameters necessary to describe the full, elastic, reflection and transmission operators, we present an elastic FWM algorithm which accounts for wave conversions. The resulting algorithm is tested on a synthetic model to give a proof of concept. The results show that the proposed extension can model wave conversions accurately and yields better inversion results than applying conventional, acoustic FWM.@en

The data-driven surface-related multiple elimination (SRME)-type approach requires fully sampled sources and receivers during the multidimensional convolution process. Otherwise, the estimated multiples will be aliased. Compared to expensive reconstruction processes before prediction, dealiasing on the estimated multiples from limited sources might provide a potential easier solution in a 2D scenario, where deep learning (DL)-based methods suit well for this highly non-linear problem. Unfortunately, DL-based multiple dealising will not function well for 3D data due to extremely coarse sampling in either source or receiver side. Thus, data interpolation/reconstruction is the only option, though the performance might not be desired. Generalized surface multiple prediction (GSMP) is the most used on-the-fly interpolation approach in 3D. Still, GSMP accuracy heavily relies on the existing traces. When fed with coarsely sampled recorded data only, GSMP tends to generate multiples with low amplitude and distorted phase, especially for small offsets. We propose a U-Net framework to repair GSMP estimated multiples such that the amplitude loss and distorted phase can be restored. In this way, the strong non-linear mapping power from DL can help repair the GSMP estimated multiples.@en

Nearly half of the Netherlands’ natural gas consump tion is allocated to heating, with direct -use geothermal heating being one of the available low-carbon energy solutions. A geothermal well doublet, designed with the two primary aims of research and commercial heat supply, is currently being installed on the campus of Delft University of Technology. The project is a key national research infrastructure and is being incorporated into the European sustainable and distributed infrastructure (EPOS: European Plate Observing System, https://www.epos-eu.org/), such that accessibility and data availability will be as wide as possible. All observations will be included in a digital-twin framework, which will allow us to make better decisions in future geothermal projects. The project includes a comprehens ive research program, involving the installation of a wide range of instruments alongside an extensive logging and coring program and monitoring network. The doublet has been cored, with substantial continuous samples from the heterogeneous reservoir, alongside a large suite of well logs in both the reservoir and overlying geological units. Such investigation is rarely undertaken in geothermal projects. A fiber-optic cable will monitor the producer well all the way down to the reservoir section, at approximately 2300m depth, in the Lower Cretaceous Delft Sandstone that is used as a geothermal reservoir in a series of existing and planned doublets in the West Netherlands Basin. A local seismic monitoring network has been installed in the surrounding area with the aim of monitoring very low-magnitude natural or induced seismicity. A vertical observation well with electromagnetic sensors will be drilled in the near future between the injector and producer to monitor cold-front propagation. This paper presents the initial modeling for the project and steps towards the production of a digital twin. Two modeling examples in the paper will emp hasize current operational challenges relevant to the project.@en

Since the appearance of wave-equation migration, many have tried to improve the resolution and effectiveness of this technology. Least-squares wave-equation migration is one of those attempts that tries to fill the gap between the migration assumptions and reality in an iterative manner. However, these iterations do not come cheap. A proven solution to limit the number of least-squares iterations is to correct the gradient direction within each iteration via the action of a preconditioner that approximates the Hessian inverse. However, the Hessian computation, or even the Hessian approximation computation, in large-scale seismic imaging problems involves an expensive computational bottleneck, making it unfeasible. Therefore, we propose an efficient computation of the Hessian approximation operator, in the context of one-way wave-equation migration (WEM) in the space-frequency domain. We build the Hessian approximation operator depth by depth, considerably reducing the operator size each time it is calculated. We prove the validity of our proposed method with two numerical examples. We then extend our proposal to the framework of full-wavefield migration, which is based on WEM principles but includes interbed multiples. Finally, this efficient preconditioned least-squares full-wavefield migration is successfully applied to a dataset with strong interbed multiple scattering.

@en

We apply supervirtual interferometry to boost the surface-wave content of two different seismic surveys. The method uses seismic interferometric principles to exploit data redundancy in multi-fold surveys. The effect on the first survey is generally positive, where the signal-to-noise ratio is improved and the relative amplitude of other events, like direct waves or reflections, is decreased. The second survey shows that the effects are not always positive. For some shots, the quality of the dispersion curve decreases and for some a higher mode becomes more dominant. This can be caused when assumptions made for seismic interferometry by corrrelation are not complied with, primarily heterogeneities in the medium and attenuation. As such, the effect of applying supervirtual interferometry could be used as an indication for local heterogeneities.@en

Since seismic imaging creates an image of the subsurface structure based on information received from the measured wavefield, it is essential to fully utilize the reflected waves. Full Wavefield Modeling (FWMod) was developed with recursive and iterative up-and-down wavefield propagation, using one-way wave propagation, to model both primary and multiple reflections. Using FWMod as the modeling engine, Full Wavefield Migration (FWM) has been introduced to directly image data including internal multiples, where internal multiple crosstalk is suppressed automatically via an inversion-based data-fitting process. This avoids the need for applying internal multiple removal, which is often challenging. Conventional one-way wave propagators calculated in the wavenumber domain, like the phase shift (PS) operator, have limitations when applied to strongly inhomogeneous media. Even when computing a new operator at each lateral grid point, they still suffer difficulties because the medium is assumed to be locally homogeneous. In the past, matrix eigendecomposition has been proposed as a way to create accurate, local velocity-based one-way propagation operators. In this article, an accurate propagator based on eigendecomposition is incorporated into FWMod and FWM. In the numerical examples, four models with strong lateral velocity variations were used to test the propagator. With a comparison of the conventional FWM based on the PS operator with input data including FWMod and a finite-difference (FD) approach, the numerical examples demonstrated that the proposed method has the potential to significantly enhance image reflectivity, suppress internal multiples, and maintain convergence speed during the least-squares inversion.

@en

The Marchenko method is capable of estimating Green’s functions between the surface of the Earth and arbitrary locations in the subsurface. These Green’s functions are used to redatum wavefields to a deeper level in the subsurface. The Marchenko method enables the isolation of the response of a specific layer or package of layers, free from the influence of the overburden and underburden. In this study, we apply the Marchenko-based isolation technique to land S-wave seismic data acquired in the Groningen province, the Netherlands. We apply the technique for combined elimination of the overburden and underburden. Our results indicate that this approach enhances the resolution of reflection data. These enhanced reflections can be utilised for imaging and monitoring applications.@en

Seismic waves traveling through the subsurface experience several forms of attenuation, including geometric spreading, reflection, transmission, and earth attenuation (via the so-called quality factor Q). To achieve high-resolution sub-surface details, it is essential to tackle all types of attenuation resulting from the overburden. Attenuation is associated with dispersion, causing gradual changes in signal shape and strength, leading to a shift of energy towards lower frequencies and potential signal distortion over time. Precisely measuring Q in seismic data is challenging but crucial for accurate subsurface imaging. The primary objective of this study is to explore the role of internal multiples (reflections within subsurface layers) for more accurate Q estimation, although traditional methods remove multiples in advance. Therefore, we utilize the Full Wavefield Migration method, which make use of the Full Wavefield Modeling (FWMod) scheme as its forward engine. This approach encompasses not only geometric spreading, reflectivity and transmission effects but also includes multiple scattering. The FWMod process is structured in a modular manner, where wavefield operators describe propagation and reflection/transmission. Consequently, including Q is relatively straightforward by redefining the propagator. Based on a synthetic data example it is demonstrated that multiples, when integrated into the inversion process, enhance the Q estimation.@en

Ground penetrating radar (GPR) is a commonly used technology for identifying and examining ice. The low electrical conductivity and the uniformity of ice covers provide GPR with exceptional signal penetration and, thus, the ability to reveal the internal layers of glaciers. To extract the necessary information, wavefield separation and imaging processing is required. This abstract presents a simultaneous diffraction and reflection imaging (SDRI) framework for ice detection using GPR data. The framework can extract hidden information in the recorded data by using wavefield separation and enhancement, for instance, the internal small-scale diffracted objects and the internal reflection layer. The traditional methods of processing and imaging data from GPR cannot provide a comprehensive understanding of the subsurface, particularly in Antarctica, due to the mutual interference between diffraction and reflection energy. This leads to the valuable geological information being concealed. The SDRI framework allows for information from both diffraction and reflection to be obtained without any interference. The diffraction method will focus on small-scale geological features while reflection will highlight large-scale structural information. The proposed SDRI framework has been applied to a field ice GPR data set from Antarctica, demonstrating its effectiveness in uncovering the hidden geology buried under the ice.@en

In the process of seismic subsurface imaging, there is no acceptable forward model reflecting the AVO response in a laterally inhomogeneous medium for reservoir characterization. This means that even when inversion is performed in full waveform, local heterogeneity is typically not fully incorporated while emplying a local 1.5D assumption. Thus, it is impossible to image and classify the subsurface features with these local heterogeneities. Still, the angle-dependent response encodes heterogeneity information that assists overcoming this issue if used properly. To exploit its capabilities, we present a way for identifying reservoir characteristics in the presence of local heterogeneity by linking encoded angle-dependent responses created using angle-dependent Full Wavefield Migration with their originating source - the relevant geological context. To accomplish this purpose, a pipeline technique that integrates the produced angle-dependent responses with a pattern categorization deep-learning tool is proposed. For a basic test on synthetic data, the method successfully identified the produced different stratigraphic architectures and classified them in the training stage. The method is then validated on angle gathers generated from different models with comparable geological circumstances.@en

Seismic data interpolation is a topic well suited for deep learning (DL) applications. Scaling operation-based DL neural networks, e.g., U-Net, have been popular since its booming development in the field of seismic data processing. Although many successful studies using U-Net on seismic data, scientists start to realize the downside of its implementation, i.e., large trainable parameters (normally larger than 1 million), the potential risks of over-fitting, and tedious hyper-parameter selection. Therefore, in this abstract, we introduce a mixed-scale dense convolutional neural network (MS-DCNN) for seismic data interpolation with relatively few trainable parameters to reduce the risk of over-fitting. This MS-DCNN was originally developed for biomedical image processing. In addition, this neural network can be trained with relatively small training set. Via a field data case study, the different behavior of U-Net and MS-DCNN is analyzed and compared for a specific interpolation problem, where 9 consecutive shot records were missing from a 2D line of marine seismic data.@en

Seismic imaging is crucial for subsurface exploration and monitoring, with a focus on deep and complex structures. Seismic wave migration solves the wave equation, and an accurate propagator is essential. Full Wavefield Modeling (FWMod) was developed based on recursive and iterative up/down wavefield propagation, modeling both primaries and multiples. Embedded within Full Wavefield Migration (FWM) it can be used to image data including multiples, resulting in better illumination in case primary illumination is not sufficient. FWM can be efficient and effective, but conventional one-way wave operators, such as Phase Shift Plus Interpolation Migration, have limitations in strongly inhomogeneous media. Local velocity-based one-way operator based on eigen decomposition was proposed and integrated within FWMod and FWM in this study, improving image amplitudes and fidelity and improving converage speed in the least-squares inversion process.@en

In contrast to conventional acquisition practices, simultaneous source acquisition allows for overlapping wavefields to be recorded. Relaxing the shot schedule in this manner has certain advantages, such as allowing for faster acquisition and/or denser shot sampling. This flexibility usually comes at the cost of an extra step in the processing workflow, where the wavefields are deblended, that is, separated. An inversion-type algorithm for deblending, based on the focal transform, is investigated. The focal transform uses an approximate velocity model to focus seismic data. The combination of focusing with sparsity constraints is used to suppress blending noise in the deblended wavefield. The focal transform can be defined in different ways to better match the spatial sampling of different types of marine surveys. To avoid solving a large inverse problem, involving a large part of the survey simultaneously, the input data can be split into sub-sets that are processed independently. We discuss the formation of such sub-sets for ocean bottom node and streamer-type acquisitions. Two deblending experiments are then carried out. The first is on numerically blended ocean bottom node field data. The second is on field-blended towed streamer data with a challenging signal overlap. The latter experiment is repeated using curvelet-based deblending for comparison purposes, showing the virtues of the focal deblending process. Several challenges of basing deblending around the focal transform are discussed as well as some suggestions for improved implementations.

@en

Poor knowledge of source and receiver positions in ultra-high-resolution marine seismic data is the cause of severe damage which requires novel processing techniques to mitigate. This type of seismic data is highly relevant for ultra-shallow subsurface imaging in geo-engineering projects both offshore and in harbours. Current positioning technologies are limited partly by their accuracy but also the fact that they are only placed on head and tail buoys of the towed arrays. This leaves receiver locations on the length of the streamer cable to be interpolated. Rather than developing additional processing methods, we propose to improve the quality of the data by introducing a complimentary receiver positioning system to reconstruct the shape of the streamer cable in 4D using Fiber Optic Shape Sensing (FOSS) technology. In this abstract, we outline the key features of FOSS technology and provide a conceptual overview of our efforts to bring this technology to the field.@en

To avoid multiple iterations of normal moveout (NMO) velocity estimation followed by short-wavelength statics estimation usually performed on land data, and to also improve the accuracy and computational efficiency of the latter, a low-rank-based residuals statics (LR-ReS) estimation and correction framework has been recently proposed. The method iteratively promotes the low-rank structure in the midpoint-offset-frequency domain of 2D data as statics-free data can be approximated by low-rank matrices, while data influenced by the weathering layers exhibits slow singular values decay. For 3D data, there exist different options to organize it into 2D matrices to be able to compute the singular value decomposition (SVD) required for low-rank approximation. It is also essential to find an organization that reveals the rank structure. We examine the different organization options. Based on finding a suitable sorting domain, we extend the LR-ReS estimation and correction to 3D data. We demonstrate the performance of the method on simulated data and will show field data results during the presentation.@en

The confluence of our ability to handle big data, significant increases in instrumentation density and quality, and rapid advances in machine learning (ML) algorithms have placed Earth Sciences at the threshold of dramatic progress. ML techniques have been attracting increased attention within the seismic community, and, in particular, in microseismic monitoring where they are now being considered a game-changer due to their real-time processing potential. In our review of the recent developments in microseismic monitoring and characterisation, we find a strong trend in utilising ML methods for enhancing the passive seismic data quality, detecting microseismic events, and locating their hypocenters. Moreover, they are being adopted for advanced event characterisation of induced seismicity, such as source mechanism determination, cluster analysis and forecasting, as well as seismic velocity inversion. These advancements, based on ML, include by-products often ignored in classical methods, like uncertainty analysis and data statistics. In our assessment of future trends in ML utilisation, we also see a strong push toward its application on distributed acoustic sensing (DAS) data and real-time monitoring to handle the large amount of data acquired in these cases.

@en