2022 Events

• January 15th 2022:The following expanded abstracts have been submitted for presentation at the 2022 annual EAGE meeting (Madrid):

1. Kamal Aghazade, Hossein S Aghamiry, Ali Gholami, and Stephane Operto, Multiparameter wavefield reconstruction inversion in elastic media using augmented Lagrangian.
2. Kamal Aghazade, Hossein S Aghamiry, Ali Gholami, and Stephane Operto, Sketched waveform inversion (SWI): an efficient augmented Lagrangian based full-waveform inversion with randomized source sketching.
3. Guo Gaoshan, Stephane Operto, Ali Gholami and Hossein S Aghamiry, Approximate data-domain Hessian in extended-source time-domain full waveform inversion using matching filter and conjugate gradient method.
4. Arash Rezaei, Hossein S. Aghamiry, Kamal Aghazade, Ali Gholami, and Stephane Operto, Improving block conjugate gradient method by randomized sketching for large-scale frequency-domain WRI.
5. Xiaole Zhou, Serge Sambolian, Haiqiang Lan, Ling Chen, and Stephane Operto, First-arrival traveltime + slope tomography in tilted anisotropic media with undulant topography.
6. Mahdi. Sonbolestan, Hossein. S. Aghamiry, Ali. Gholami, and Stephane Operto, On the role of data-space Hessian in wavefield reconstruction inversion.
7. Hossein. S. Aghamiry, Ali. Gholami, and Stephane Operto, Highly-accurate wavefield reconstruction inversion using convergent Born series.

The PDF of the abstracts are available on the intranet sponsor page.

• January 21st 2022: The paper entitled " Accurate 3D frequency-domain seismic wave modeling with the wavelength-adaptive 27-point finite-difference stencil: a tool for full waveform inversion" by H. Aghamiry, A. Gholami, L. Combe and S. Operto has been accepted for publication in Geophysics.

The final version of the paper is available on the intranet sponsor page.

• February 15th 2022: The paper entitled "Large-scale highly-accurate extended full waveform inversion using convergent Born series" by Hossein S. Aghamiry, Ali Gholami, Kamal Aghazade, Mahdi Sonbolestan, and Stéphane Operto has been submitted to the Journal of Computational Physics. This paper proposes a new algorithm to implement the iteratively-refined wavefield reconstruction inversion, which avoids solving the normal system associated with the augmented wave equation. Moreover, it assesses the convergent Born series method as a forward engine for FWI. The preprint is available on the intranet page for sponsors.
• February 21st 2022: PETROBRAS joins the WIND consortium.
• February 21st 2022: The FDFDMATRIX3D package for 3D finite-difference frequency-domain seismic wave simulation in the VTI visco-acoustic approximation is freely available on the Open Code page. This package includes the recent wavelength-adaptive finite-difference stencil.
• March 27th 2022: The following extended abstracts have been submitted to Annual SEG/Image conferenece:

1. H. S. Aghamiry, A. Gholami, S. Operto and A. Malcolm, Localized Wavefield Inversion (LWI): an Adaptation of Multi-Block ADMM for Localized FWI.
2. A. Gholami, H. S. Aghamiry and S. Operto, On the connection between WRI and FWI: Analysis of the nonlinear term in the Hessian matrix.
3. G. Guo, S. Operto, A. Gholami, and H. S. Aghamiry, A practical implementation of data-space Hessian in the time-domain formulation of source extended full-waveform inversion.
4. A. Jalali, H. S. Aghamiry and A. Gholami, ADMM-based full waveform inversion of ground penetrating radar data.
5. Y. Wu, H. S. Aghamiry, S. Operto, A. Rezaei and J. Ma, Wave simulation in non-smooth media with PINN by quadratic neural network and PML condition.
6. X. Zhou, S. Sambolian, G. Guo, S. Operto, H. Lan and L. Chen, First-arrival traveltime and slope tomography in tilted anisotropic media with undulant topography.

The first abstract deals with target-oriented FWI.Target-oriented imaging is implemented in the framework of ADMM-based frequency-domain WRI. The method is simple to implement and cheap. Moreover, the baseline model outside the targeted area can be updatedduring the first iteration of each frequency inversion. The second abstract reviews some connexions between the non linear term of the FWI Hessian and the extended-space wavefield reconstruction inversion. The third abstract discusses some algorithmic aspects related to the implementation of IR-WRI in the time domain, where one challenge is to take into account accurately the data-domain Hessian associated with the source extension reconstruction problem. The fourth abstract presents a proof-of-concept study of IR-WRI applied on GPR data. The fifth abstract shows how PML boundary conditions improve the scattered wavefield simulation performed with Physics-Informed Neural Networks (PINNs). The sixth abstract presents first-arrival slope tomography with undulant surface. The preprint of these extended abstracts can be downloaded from the intranet page for sponsors.

• • April 5th 2022: The paper entitled " Quantitative Photoacoustic Tomography Using Extended Full-Wave Inversion: A Simulation Study" by Ranjbaran, M., Aghamiry, H. S., Gholami, A., Operto, S. & Avanaki, K has been submitted to the IEEE Transactions on Medical Imaging. The paper, which is a colabration with the OPIRA team, deals with methodological development in photoacoustic tomography (PAT). PAT aims at recovering spatially varying absorption coefficients (AC), which act as an internal passive source induced by electromagnetic radiation. We recast the AC reconstruction as a source location problem that is tackled by ADMM-based FWI. The practical implementation of this procedure raises however two major issues. First, the velocity model cannot be reconstructed from the photoacoustic source alone due to insufficient illumination and hence cannot be processed as an optimization variable. Therefore, we need a good apriori knowledge of velocity to reconstruct AC. Second, the ill-posedness of PAT prevents assigning a degree of freedom to each sample discretizing the medium. To overcome the later issue, we design a parsimonious parametrization of the ACs by finding regions with equal AC from the segmentation of the SOS map. By doing so, we end up with an ADMM loop that embeds two least-squares problems for wavefields and ACs. First, the ACs are kept fixed, and the wavefields are reconstructed with data assimilation. Then, the wavefields are kept fixed, and the ACs are updated when a total-variation regularization is applied to tighten the search space of the problem. One difficulty with the proposed algorithm is the requirement for a good approximation of the velocity model. To address the first issue, we propose to build the velocity and intrinsic attenuation a priori by applying another ADMM-based FWI on an ultrasound dataset obtained by sending and recording acoustic waves from the device surrounding the target. The preprint is available on the intranet page for sponsors.
  • April 14th 2022: The scientific proposal for the second round of the WIND project is available under request. Contact Stéphane Operto (email: operto@geoazur.unice.fr) or Hossein Aghamiry (email: aghamiry@geoazur.unice.fr).
  • April 25th 2022: The FDFDMATRIX3D package was updated. Some bugs in scripts for plotting results were corrected (an undefined environment variable was used). Additional benchmarks are provided to validate implementation of arbitrary source and receiver positioning in the finite-difference methods with Kaiser-windowed sinc (KWS) functions . Accordingly, we also provide in the OpenCode page the KWSINC package for resampling and arbitrary source and receiver positionings in finite-difference methods with Kaiser-windowed sinc (KWS) functions. This package contains a fortran module that implements 1D/2D/3D KWS functions and a simple 2D finite-difference time-domain (FDTD) wave simulation code, which shows how to interface the KWS method in FDTD wave simulation. The accuracy of temporal resampling of seismograms and arbitrary source and receiver positioning for wave simulation is validated with easy-to-run numerical examples. A technical report that reviews in detail the implementation of the KWS method and the numerical benchmarks is provided in the distribution.
  • June 5-10 2022: Participation of the WIND team to the 2022 EAGE meeting. The slides of the following presentation are available on the sponsor page.

WS02:Cycle skipping workshop S. Operto, On the role of source extensions in FWI: Beyond the Born approximation. WS03: Multiparameter FWI K. Aghazade. Visco Elastic Full Waveform Inversion by Alternating Direction Method of Multipliers. Technical sessions H. Aghamiry. Highly-accurate wavefield reconstruction inversion using convergent Born series. Wavefield Reconstruction Inversion session M. Sombolestan. On the role of data-space Hessian in wavefield reconstruction inversion. A. Rezaei, Improving block conjugate gradient method by randomized sketching for large-scale frequency-domain WRI K. Aghazade. Sketched waveform inversion (SWI): an efficient augmented Lagrangian based full-waveform inversion with randomized source sketching. K. Aghazade. Multiparameter wavefield reconstruction inversion in elastic media using augmented Lagrangian G. Guo. Approximate data-domain Hessian in extended-source time-domain full waveform inversion using matching filter and conjugate gradient method. X. Zhou. First-arrival traveltime + slope tomography in tilted anisotropic media with undulant topography Velocity model building session

• • Tuesday June 7th: Meeting with current and potential sponsors for the second round. Slides are available on the sponsor page.
  • June 17th 2022: The paper entitled " Three-dimensional finite-difference & finite-element frequency-domain wave simulation with multi-level optimized additive Schwarz domain-decomposition preconditioner: A tool for FWI of sparse node datasets" by P.-H. Tournier, P. Jolivet, V. Dolean, H. S. Aghamiry, S. Operto, and S. Riffo has been accepted for publication in Geophysics.
  • June 20 th 2022: Online ExxonMobil seminar by S. Operto, WIND (Waveform Inversion of Node Data): Past, present and future. Slides are available on the sponsor page.
  • June 20-24 2022: Participation to the "Solvers for frequency-domain wave problems and applications" workshop in University of Strathclyde, Glasgow.

Pierre Jolivet, Spectral coarse grids for overlapping Schwarz methods in PETSc.
Pierre-Henri Tournier, Optimized Schwarz preconditioners for finite difference and finite element acoustic simulations.
Hossein S. Aghamiry, LocalizedWavefield Inversion (LWI): an Adaptation of Multi-Block ADMM for Localized FWI.
Victorita Dolean, Two-level DDM preconditioners for Low Frequency Maxwell equations. The details are available on the webpage of the workshop ( ICMS@Strathclyde: Solvers for frequency-domain wave problems and applications | ICMS - International Centre for Mathematical Sciences)

• • July 20th 2022: The paper entitled "Wave simulation in non-smooth media by PINN with quadratic neural network and PML condition" by Yanqi Wu, Hossein S. Aghamiry, Stephane Operto, and Jianwei Ma has been submitted to the Geophysics journal. This paper proposes to solve the acoustic and visco-acoustic scattered-field wave equation in the frequency domain with PINN. In this paper, the perfectly matched layer (PML) conditions are considered in the loss function of PINN. Also, a quadratic neural network is designed to overcome the detrimental effects of non-smooth models in PINN. You can get the submitted version using this link: https://arxiv.org/abs/2208.08276.
  • August 2022: We perform our first application of 3D frequency-domain FWI in the Gorgon OBN data in the frequency band [1.7Hz;8.55Hz]. The forward problem in the frequency domain is performed with our wavelength-adaptive finite-difference stencil (Aghamiry et al., Geophysics, 2022) and the multifrontal sparse direct solver MUMPS. The physics is VTI visco-acoustic. Only the vertical wavespeed is updated. The size of the problem at 8.55Hz involves 68 million unknowns. 650 nodes and 400,000 shots are involved in the inversion at each iteration.

Gorgon FWI model at 8.55Hz.

The depth slice is at reservoir level.

Examples of depth slices.

An illustration of data fit in the frequency domain (Frequency 5 Hz)

Illustration of data fit in the time domain.

Work in progress.

• September 18th: The WIND technical report about the first FWI results on the 3D OBN data from the Gorgon field can be dowlloaded on the sponsor page. Ref: S. Operto and the WIND team, 3D visco-acoustic VTI frequency-domain FWI of OBN Gorgon data - Preprocessing, baseline FWI results and their quality control, WIND project, Technical Report N 12, 316 pages, 2022.
• October 6th: The E-lecture of H. Aghamiry entitled "On The Robustness of Sparsity-Promoting Regularized Wavefield Inversion with Phase Retrieval against Sparse Long-Offset Acquisitions" was uploaded on the EAGE platform ( link). This presentation of Hossein was selected as one of the best-rated talks at the EAGE Annual Conference in Amsterdam and recorded as E-lecture during the EAGE Annual conference in Madrid.
• October 30th: The manuscript entitled "Does 3D frequency-domain FWI of full-azimuth/long offset OBN data feasible? The Gorgon case study" by Operto et al. was submitted to the special section of The Leading Edge journal to appear in March 2023. This manuscript presents our first FWI results on the OBN data from the 2015/16 Gorgon survey, offshore Australia. To the best of our knowledge this is the first OBN frequency-domain FWI case study based on direct solver. We use the sparse multifrontal MUMPS solver to perform this application. The draft of the manuscript is available on the intranet page for sponsors.
• November 3rd: The 2022 activity report of the WIND project as well as the agenda of the Annual Meeting (16-18 November) are available on the intranet page for sponsors.
• November 16-18: Third annual WIND meeting fot sponsors. The program of the meeting, the PDF/PPT of the presentations as well as the recording of the presentation are available on the sponsor page.
• December 19: A tutorial review on extended-space (ES) FWI has been submitted to Geophysics. This paper entitled "Full Waveform Inversion beyond the Born approximation: a tutorial review" aims at reviewing the physical principles underlying FWI when the search space is extended through a relaxation of the wave equation. We show that the relaxed wave equation is in fact the Lippmann-Schwinger equation where a wavefield is computed as a sum of a background wavefield and a scattered wavefield by a scattering source, the latter being formed by the interaction of the total field with a scattering object (or model perturbation). This equation is used to estimate wavefields that best match the true wavefields by estimating approximately the scattering source formed by the unknown incident true wavefields to the unknown sougth model perturbation. These scattering sources are estimated by fitting in the least-squares sense the data residuals of FWI at the current iteration, namely the restriction at receivers of the true scattered wavefields (a strongly underdetermined problem). These scattering sources are computed by propagating backward in time the weighted data residuals by the data-domain Hessian. The impact of this approximation in ES-FWI is discussed in the paper, the role of the data-domain Hessian of the scattering source estimation problem is explained, the accuracy of the reconstructed wavefields is discussed, and the similarities and differences between classical and ES-FWI are clearly highlighted.

A preprint is available on the intranet page for sponsors.