Seismic imaging provides much of our information about the Earth's crustal structure. The principal source errors derives from simplicistically modeled predictions complex, scattered wavefields that interact with each subsurface point to be imaged. A new method wavefield extrapolation based on inverse scattering theory produces accurate estimates these wavefields, while still using relatively little properties. We use it for first time create real target-oriented seismic images a North Sea...

Linear operators and optimization are at the core of many algorithms used in signal image processing, remote sensing, inverse problems. For small to medium-scale problems, existing software packages (e.g., MATLAB, Python NumPy SciPy) allow explicitly build dense or sparse matrices perform algebraic operations with syntax that closely represents their equivalent mathematical notation. However, real-application, large-scale do not lend themselves explicit matrix representations, usually...

Noise suppression is an essential step in many seismic processing workflows. A portion of this noise, particularly land datasets, presents itself as random noise. In recent years, neural networks have been successfully used to denoise data a supervised fashion. However, learning always comes with the often unachievable requirement having noisy-clean pairs for training. Using blind-spot networks, we redefine denoising task self-supervised procedure where network uses surrounding noisy samples...

Abstract The presence of coherent noise in seismic data leads to errors and uncertainties, as such it is paramount suppress early efficiently possible. Self‐supervised denoising circumvents the common requirement deep learning procedures having noisy‐clean training pairs. However, self‐supervised suppression methods require extensive knowledge statistics. We propose use explainable artificial intelligence approaches ‘see inside black box’ that network gained replace need for any prior...

The solution of the inverse scattering problem for one-dimensional Schrödinger equation is given by Marchenko equation. Recently, a Marchenko-type has been derived three-dimensional (3D) acoustic wave fields, whose shown to recover Green's functions from points within medium its exterior, using only single-sided scattered data. Here we extend this approach 3D vectorial fields that satisfy elastodynamic and interior an elastic, solid-state purely external one-sided measurements. method...

Standard seismic processing steps such as velocity analysis and reverse time migration (imaging) usually assume that all reflections are primaries: Multiples represent a source of coherent noise must be suppressed to avoid imaging artifacts. Many suppression methods relatively ineffective for internal multiples. We show how predict remove multiples using Marchenko autofocusing interferometry. first can theoretically reconstructed in convolutional interferometry by combining purely reflected,...

Marchenko redatuming is a revolutionary technique to estimate Green’s functions from virtual sources in the subsurface using only data measured at earth’s surface, without having place either or receivers subsurface. This goal achieved by crafting special wavefields (so-called focusing functions) that can focus energy chosen point Despite its great potential, strict requirements on reflection response such as knowledge and accurate deconvolution of source wavelet (including absolute scaling...

Interpolation of aliased seismic data constitutes a key step in processing workflow to obtain high-quality velocity models and images. Building on the idea describing wavefields as superposition local plane waves, we propose interpolate by using physics informed neural network (PINN). In proposed framework, two feed-forward networks are jointly trained wave differential equation well available terms objective function: primary assisted positional encoding is tasked with reconstructing data,...

A central component of imaging methods is receiver-side wavefield backpropagation or extrapolation in which the from a physical source scattered at any point subsurface estimated data recorded by receivers located near Earth’s surface. Elastic reverse-time migration usually accomplishes simultaneous reversed-time ‘injection’ particle displacements (or velocities) each receiver location into modeling code. Here, we formulate an exact integral expression based on reciprocity theory that uses...

Abstract The inversion of petrophysical parameters from seismic data represents a fundamental step in the process characterizing subsurface, with applications ranging subsurface resource exploration to geothermal, carbon capture and storage, hydrogen storage. We propose novel, data‐driven approach, named Seis2Rock, that utilizes optimal basis functions learned well‐log information create direct link between pre‐stack so‐called band‐limited reflectivities. Seis2Rock is composed two stages:...

Methods for wavefield injection are commonly used to extrapolate seismic data in reverse time migration (RTM). Injecting a single component of the acoustic field, example, pressure, leads ambiguity direction propagation. Each recorded wavefront is propagated both upward and downward, spurious (or ghost) reflectors created alongside real subsurface image. Thus, separation based on combination pressure particle velocity generally performed prior imaging extract only upgoing field from...

Conventional seismic imaging methods rely on the single-scattering Born approximation, requiring removal of multiply scattered events from reflection data prior to imaging. Additionally, many use an acoustic representing solid earth as (fluid) medium. We have developed for (solid) elastic media that primaries and internal multiples, including their PS SP conversions, thus obviating need multiple improving handling conversions. Our autofocusing method, which estimates full multicomponent...

Numerical integral operators of the convolution type form basis most wave-equation-based methods for processing and imaging seismic data. Because several these require solution an inverse problem, multiple forward adjoint passes modeling operator are generally required to converge a satisfactory solution. We have highlighted memory requirements computational challenges that arise when implementing such on 3D data sets their use solving large systems equations. A Python framework is presented...

Deep-image prior (DIP) is a novel approach to solving ill-posed inverse problems whose solution parameterized with an untrained deep neural network and cascaded the forward modeling operator. A key component success of such method represented by choice architecture, which must act as natural problem at hand provide strong inductive bias toward desired solution. Inspired close link between networks iterative algorithms in classical optimization, we apply unrolled version gradient descent (GD)...

Solving inverse problems with the reverse process of a diffusion model represents an appealing avenue to produce highly realistic, yet diverse solutions from incomplete and possibly noisy measurements, ultimately enabling uncertainty quantification at scale. However, because intractable nature score function likelihood term (i.e., $\nabla_{\mathbf{x}_t} p(\mathbf{y} | \mathbf{x}_t)$), various samplers have been proposed in literature that use different (more or less accurate) approximations...

Summary Elastic full-waveform inversion has recently been utilized to estimate the physical properties of upper tens meters subsurface, leveraging its capability exploit complete information contained in recorded seismograms. However, due non-linear and ill-posed nature problem, standard approaches typically require an optimal starting model avoid producing non-physical solutions. Additionally, conventional optimization methods lack a robust uncertainty quantification, which is essential for...

Summary In recent years, Full-Waveform Inversion (FWI) has been extensively used to derive high-resolution subsurface velocity models from seismic data. However, due the nonlinearity and ill-posed nature of problem, FWI requires a good starting model avoid producing non-physical solutions (i.e., being trapped in local minima). Moreover, traditional optimization methods often struggle effectively quantify uncertainty associated with recovered solution, which is critical for decision-making...

Passive seismic monitoring is an effective tool to assess activity and analyze subsurface structures in around well sites because it does not require active source and, thus, cheap, non-intrusive, environmentally friendly. The on-campus shallow at King Abdullah University of Science Technology (KAUST) was drilled from February April 2024, reaching a target depth ~392 meters, subsequently cased cemented along its entire length total depth. main challenge for the site unconsolidated topmost...

Abstract The knowledge of the local slope field prestack seismic data is essential in several signal processing tasks. Building on our previous slope‐assisted, physics‐informed interpolation framework, dubbed PINNslope, we introduce a series enhancements that elevate framework's versatility. This ultimately enables its application to different separation problems, with specific focus ground roll removal. To begin with, estimated using neural networks framework compared against analytical and...

Summary Ocean-bottom seismic acquisitions are gaining widespread popularity across a variety of subsurface applications. However, the high cost these systems often necessitates receiver geometries with large intervals between ocean-bottom cables or nodes. The upside-down Rayleigh-Marchenko (UD-RM) method has been recently proposed as an effective solution for accurate redatuming and imaging sparse seabed data. In this paper, we present first successful application UD-RM to field We...

To limit the time, cost, and environmental impact associated with acquisition of seismic data in recent decades, considerable effort has been put into so-called simultaneous shooting acquisitions, where sources are fired at short time intervals between each other. As a consequence, waves originating from consecutive shots entangled within recordings, yielding blended data. For processing imaging purposes, generated by individual shot must be retrieved. This process, called deblending, is...

Abstract Full waveform inversion stands at the forefront of seismic imaging technologies, pivotal in retrieving high‐resolution subsurface velocity models. Its application is especially profound when complex geologies such as salt bodies, which are regions notoriously challenging, yet essential given their hydrocarbon potential. However, with power full comes intrinsic challenge estimating associated uncertainties. Such uncertainties crucial understanding reliability models, particularly...

Seismic datasets contain valuable information that originate from areas of interest in the subsurface; such seismic reflections are however inevitably contaminated by other events created waves reverberating overburden. Multi-Dimensional Deconvolution (MDD) is a powerful technique used at various stages processing sequence to create ideal deprived overburden effects. Whilst underlying forward problem well defined for single source, successful inversion MDD equations requires availability...