Time-lapse full-waveform inversion using Hamiltonian Monte Carlo: A proof of concept
Name
segam2020-3422774.1.pdf
Description
Published version
Size
737.38 KB
Format
Adobe PDF
Checksum (MD5)
b7385a9cd2ed11f4e9757626a90f9d57
Author(s)
Kotsi, Maria
Malcolm, Alison
Ely, Gregory
Date Issued
September 25, 2020
Journal
SEG Technical Program Expanded Abstracts 2020
Publisher
Society of Exploration Geophysicists
Citation
Maria Kotsi, Alison Malcolm, Gregory Ely; September 1, 2020. "Time-lapse full-waveform inversion using Hamiltonian Monte Carlo: A proof of concept." Proceedings of the SEG Technical Program Expanded Abstracts 2020. SEG Technical Program Expanded Abstracts 2020. (pp. pp. 845-849). ASME.
Version
Final published version
Abstract
Uncertainty quantification is an important aspect of time–lapse imaging and is typically done using Bayesian inference. Traditional random–walk sampling methods are slow to converge and they fail to efficiently explore the high dimensional space that must be characterized in time-lapse imaging. We propose the use of a local acoustic Helmholtz solver for an efficient time–lapse Hamiltonian Monte Carlo (HMC) inversion. Using a local acoustic solver offers the advantage of quick and local gradient computations. Our numerical models demonstrate the robustness of the method over the Metropolis–Hastings algorithm, and set up the path towards non–linear uncertainty quantification of high dimensional velocity models. To our knowledge this is the first direct comparison of Metropolis– Hastings and HMC on a seismic example.
Terms of Use
Article is made available in accordance with the publisher's policy and may be subject to US copyright law. Please refer to the publisher's site for terms of use.
Persistent DSpace Link
DOI of Published Version
https://doi.org/10.1190/segam2020-3422774.1
