Stochastic full-waveform inversion (SWI) of seismic data for FWI starting models

Jinsong Chen,Uwe K. Albertin,Kurt Nihei,Sam Kaplan,Michael Hoversten,Lin Zhang,Dimitri Bevc

DOI: https://doi.org/10.1190/SEGAM2020-3418905.1

2020-01-01

SEG Technical Program Expanded Abstracts

Abstract:PreviousNext No AccessSEG Technical Program Expanded Abstracts 2020Stochastic full-waveform inversion (SWI) of seismic data for FWI starting modelsAuthors: Jinsong ChenUwe K. AlbertinKurt NiheiSam KaplanMichael HoverstenLin ZhangDimitri BevcJinsong ChenChevron Energy Technology CompanySearch for more papers by this author, Uwe K. AlbertinChevron Energy Technology CompanySearch for more papers by this author, Kurt NiheiChevron Energy Technology CompanySearch for more papers by this author, Sam KaplanChevron Energy Technology CompanySearch for more papers by this author, Michael HoverstenChevron Energy Technology CompanySearch for more papers by this author, Lin ZhangChevron Energy Technology CompanySearch for more papers by this author, and Dimitri BevcChevron Energy Technology CompanySearch for more papers by this authorhttps://doi.org/10.1190/segam2020-3418905.1 SectionsAboutPDF/ePub ToolsAdd to favoritesDownload CitationsTrack CitationsPermissions ShareFacebookTwitterLinked InRedditEmail AbstractCurrent approaches for FWI based on a waveform misfit objective function have been demonstrated to be effective for high-fidelity imaging of acoustic and elastic properties, when starting from velocity models that are within the basin of attraction of the true velocity model. Such initial models need to include the long-wavelength structure sufficiently and accurately so that the synthetic and observed data misfit is less than half of a wavelength. If this condition is violated, gradient-based optimization methods can converge to a local rather than the desired global solution. Over the past several years, a clear focus of FWI research has been centered on finding new cycle-skip immune approaches for FWI by using a range of data, image and modeling domain objective functions. In this study, we develop a stochastic full waveform inversion (SWI) approach within a Bayesian framework to estimate the long wavelength starting model for FWI. We use Markov Chain Monte Carlo (MCMC) methods to draw many samples from model space described by sparse and smooth Gaussian radial basis functions (RBF). We demonstrate the effectiveness of SWI using a 2D marine synthetic salt model. Starting from a water bottom over half space model, SWI recovers a long wavelength starting model that allows standard least-squares FWI to correctly recover the true structure of the salt.Presentation Date: Tuesday, October 13, 2020Session Start Time: 1:50 PMPresentation Time: 3:30 PMLocation: Poster Station 3Presentation Type: PosterKeywords: inversion, full-waveform inversion, statistics, seismic attributes, saltPermalink: https://doi.org/10.1190/segam2020-3418905.1FiguresReferencesRelatedDetails SEG Technical Program Expanded Abstracts 2020ISSN (print):1052-3812 ISSN (online):1949-4645Copyright: 2020 Pages: 3887 publication data© 2020 Published in electronic format with permission by the Society of Exploration GeophysicistsPublisher:Society of Exploration Geophysicists HistoryPublished Online: 30 Sep 2020 CITATION INFORMATION Jinsong Chen, Uwe K. Albertin, Kurt Nihei, Sam Kaplan, Michael Hoversten, Lin Zhang, and Dimitri Bevc, (2020), "Stochastic full-waveform inversion (SWI) of seismic data for FWI starting models," SEG Technical Program Expanded Abstracts : 840-844. https://doi.org/10.1190/segam2020-3418905.1 Plain-Language Summary Keywordsinversionfull-waveform inversionstatisticsseismic attributessaltPDF DownloadLoading ...