4.7 Article

Scattering Radiation Pattern Atlas: What Anisotropic Elastic Properties Can Body Waves Resolve?

Journal

JOURNAL OF GEOPHYSICAL RESEARCH-SOLID EARTH
Volume 124, Issue 3, Pages 2781-2811

Publisher

AMER GEOPHYSICAL UNION
DOI: 10.1029/2018JB016687

Keywords

anisotropic elasticity; full-waveform inversion; resolution analysis; Fourier analysis; body waves; null spaces

Funding

  1. King Abdullah University of Science and Technology (KAUST)
  2. KAUST

Ask authors/readers for more resources

Full-waveform inversion (FWI) optimizes the subsurface properties of geophysical Earth models in such a way that the modeled data, based on these subsurface properties, match the observed data. The anisotropic properties, whether monoclinic, orthorhombic, triclinic, or vertical transversally isotropic (VTI), of the subsurface, be it a fractured reservoir or the core-mantle boundary, are necessary to describe the observed wave phenomena. There are no principal limitations on the complexity of the anisotropy that can be inverted using FWI. However, the question remainswhat kind of anisotropic descriptions of the elastic properties of the Earth can or cannot be inverted reliably from seismic waveforms? We reveal the resolution that can be achieved through reconstructions of each elastic parameter by building vertical resolution patterns from the scattering radiation patterns of body waves. A visual analysis of these patterns indicates trade-offs, that is, perturbations of parameters that have the same reflection-based scattering patterns as other perturbations. Each trade-off leads to an apparent ambiguity in the inversion, which must be addressed by additional assumptions, constraints, or regularizations. For orthorhombic media, we find the exact trade-offs that exist between parameters. Our parameterization isolates the VTI parameters, and therefore, we also obtain trade-offs for every scattering mode for VTI media. We discover that only monoclinic parameters are recoverable from the first-order scattering of monotypic waves. We summarize the trade-offs in tables for easy reference. This paper is intended to be useful for researchers setting up anisotropic FWI problems, and interpreting or controlling the quality of such inversions.

Authors

I am an author on this paper
Click your name to claim this paper and add it to your profile.

Reviews

Primary Rating

4.7
Not enough ratings

Secondary Ratings

Novelty
-
Significance
-
Scientific rigor
-
Rate this paper

Recommended

Article Geochemistry & Geophysics

Data-Driven Microseismic Event Localization: An Application to the Oklahoma Arkoma Basin Hydraulic Fracturing Data

Hanchen Wang, Tariq Alkhalifah, Umair bin Waheed, Claire Birnie

Summary: The microseismic monitoring technique is widely used in studying hydraulic fracturing. In this study, a deep convolutional neural network is proposed to predict the event location of microseismic data. The results demonstrate that the proposed approach provides accurate and efficient microseismic event localization.

IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING (2022)

Article Geochemistry & Geophysics

Target-oriented high-resolution elastic full-waveform inversion with an elastic redatuming method

Yuanyuan Li, Tariq Alkhalifah

Summary: Characterizing the elastic properties of deep-buried reservoirs beneath complex overburden structures is a challenging task for seismic inversion. Elastic full-waveform inversion (FWI) can quantitatively estimate subsurface elastic properties with high resolution, but using high frequencies is computationally expensive and obtaining high-resolution inversion results for deep targets is difficult due to complex overburden structures. To address these limitations, a target-oriented high-resolution elastic FWI scheme is proposed, using estimated elastic data for a virtual survey deployed above a zone of interest.

GEOPHYSICS (2022)

Article Geochemistry & Geophysics

PINNup: Robust Neural Network Wavefield Solutions Using Frequency Upscaling and Neuron Splitting

Xinquan Huang, Tariq Alkhalifah

Summary: Seismic wave-equation based methods and physics-informed neural network (PINN) have great potential in illuminating the interior of Earth. However, their accuracy and training cost are limited when dealing with high-frequency wavefields. Therefore, a novel approach using frequency upscaling and neuron splitting is proposed to improve the accuracy and convergence speed of wavefield solutions.

JOURNAL OF GEOPHYSICAL RESEARCH-SOLID EARTH (2022)

Article Geosciences, Multidisciplinary

ML-misfit: A neural network formulation of the misfit function for full-waveform inversion

Bingbing Sun, Tariq Alkhalifah

Summary: A robust misfit function is crucial for stable velocity model updates in full-waveform inversion. We propose ML-misfit, a machine learning-based approach, to learn a data-adaptive misfit function. The neural network architecture is designed to allow for global comparison of the predicted and measured data, guaranteeing efficient training. By training the network using a meta-learning framework, the ML-misfit automatically improves and provides robust updating of the velocity model.

FRONTIERS IN EARTH SCIENCE (2022)

Article Geochemistry & Geophysics

Improving the Generalization of Deep Neural Networks in Seismic Resolution Enhancement

Haoran Zhang, Tariq Alkhalifah, Yang Liu, Claire Birnie, Xi Di

Summary: Seismic resolution enhancement is crucial for subsurface structure characterization. We propose a simple domain adaptation procedure called MLReal-Lite, which improves the performance of neural networks by bringing the distributions of real and synthetic data closer to each other through linear operations.

IEEE GEOSCIENCE AND REMOTE SENSING LETTERS (2023)

Article Geosciences, Multidisciplinary

Transfer learning for self-supervised, blind-spot seismic denoising

Claire Birnie, Tariq Alkhalifah

Summary: Noise is a common issue in seismic data, affecting the performance of supervised deep learning denoising. Self-supervised blind-spot networks can train directly on raw noisy data but struggle with correlated noise. We propose initial supervised training on synthetic data followed by self-supervised fine-tuning on field data, resulting in improved denoising performance.

FRONTIERS IN EARTH SCIENCE (2022)

Article Geosciences, Multidisciplinary

Machine learning in microseismic monitoring

Denis Anikiev, Claire Birnie, Umair bin Waheed, Tariq Alkhalifah, Chen Gu, Dirk J. Verschuur, Leo Eisner

Summary: The combination of enhanced big data handling capabilities, improved instrumentation density and quality, and rapid advances in machine learning algorithms has opened the door for significant progress in Earth Sciences. Machine learning methods are increasingly gaining attention in the seismic community, particularly in microseismic monitoring where they have the potential to revolutionize real-time processing. Recent developments in microseismic monitoring have shown a strong trend towards utilizing machine learning techniques to enhance passive seismic data quality, detect microseismic events, and locate their hypocenters. Additionally, machine learning methods are being adopted for advanced event characterization and seismic velocity inversion, providing valuable by-products such as uncertainty analysis and data statistics. Future trends in machine learning utilization point towards its application on distributed acoustic sensing (DAS) data and real-time monitoring to handle the large amount of data acquired in these cases.

EARTH-SCIENCE REVIEWS (2023)

Article Multidisciplinary Sciences

A neural network based global traveltime function (GlobeNN)

Mohammad H. Taufik, Umair bin Waheed, Tariq A. Alkhalifah

Summary: We developed a neural network-based method for global traveltime computation, which can efficiently handle a large number of receivers provided by DAS arrays. This method allows rapid and accurate estimation of seismic wave propagation time, making it an essential tool for advancing seismological studies.

SCIENTIFIC REPORTS (2023)

Article Chemistry, Analytical

Diffusion Model for DAS-VSP Data Denoising

Donglin Zhu, Lei Fu, Vladimir Kazei, Weichang Li

Summary: This study pioneers the use of diffusion models for denoising DAS-VSP data, demonstrating their effectiveness in removing noise and surpassing conventional methods, and showcasing the potential of diffusion models in DAS processing.

SENSORS (2023)

Article Geochemistry & Geophysics

Elastic Seismic Imaging Enhancement of Sparse 4C Ocean-Bottom Node Data Using Deep Learning

Shijun Cheng, Xingchen Shi, Weijian Mao, Tariq A. Alkhalifah, Tao Yang, Yuzhu Liu, Heping Sun

Summary: The ocean-bottom node (OBN) seismic acquisition system aims to improve imaging quality through a deep learning-based method using a multiscale convolutional neural network (Ms-CNN) for sparse data acquisition. The Ms-CNN is trained to map sparse data images to dense data images, allowing for direct processing and improving event continuity and noise reduction in migration results.

IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING (2023)

Article Geochemistry & Geophysics

An Anisotropic Waveform Inversion Using an Optimal Transport Matching Filter Objective: An Application to an Offshore Field Dataset

Bingbing Sun, Tariq Alkhalifah

Summary: Due to the nature of the Earth's layering and conventional seismic wavelength, seismic waves often experience significant anisotropy in many subsurface areas, particularly vertical transverse isotropy (VTI). Inverting to this type of Earth model using waveforms poses challenges such as nonlinearity and parameter tradeoff. The optimal transport of the matching filter (OTMF) has been introduced as a robust misfit function for full-waveform inversion (FWI). Applying a VTI FWI with OTMF misfit on offshore Australian data yields geologically meaningful models and demonstrates the effectiveness of OTMF in mitigating cycle skipping.

IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING (2023)

Proceedings Paper Computer Science, Artificial Intelligence

DIRECT IMAGING USING PHYSICS INFORMED NEURAL NETWORKS

Tariq Alkhalifah, Xinquan Huang

Summary: Imaging is a crucial task in various fields, and the exploding reflector assumption provides a direct imaging approach for zero-offset data. However, aliasing problems arise when the data are coarsely sampled. By formulating the frequency-domain wavefield as a neural network function, the physics-informed neural network framework enables subsurface imaging.

2022 IEEE INTERNATIONAL CONFERENCE ON IMAGE PROCESSING, ICIP (2022)

Article Computer Science, Artificial Intelligence

Direct domain adaptation through reciprocal linear transformations

Tariq Alkhalifah, Oleg Ovcharenko

Summary: In this paper, a direct domain adaptation (DDA) approach is proposed to enhance the training of supervised neural networks on synthetic data by incorporating features from real-world data. The experimental results show that this method can improve the accuracy of the network and the input features from the source and target domains share similarities along the principal components after specific transformations.

FRONTIERS IN ARTIFICIAL INTELLIGENCE (2022)

Article Geochemistry & Geophysics

Upwind, No More: Flexible Traveltime Solutions Using Physics-Informed Neural Networks

Mohammad Hasyim Taufik, Umair bin Waheed, Tariq A. A. Alkhalifah

Summary: The eikonal equation is important in various scientific and engineering fields, particularly in seismic wave propagation and travel time modeling. Traditional finite-difference methods suffer from numerical inaccuracies and high computational costs, while physics-informed neural networks offer higher accuracy and scalability. This article demonstrates the flexibility and interpolation-extrapolation capability of PINN solutions, particularly in handling gaps in velocity models.

IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING (2022)

Article Geochemistry & Geophysics

StorSeismic: A New Paradigm in Deep Learning for Seismic Processing

Randy Harsuko, Tariq A. Alkhalifah

Summary: This paper presents a dataset-centric framework called StorSeismic for seismic data processing, which utilizes a neural network (NN) for preprocessing to extract and store geometrical features of the data. By pretraining and fine-tuning, this framework can be used for various seismic processing tasks such as denoising and velocity estimation.

IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING (2022)

No Data Available