Simulation study of joint trans-dimenstional bayesian inversion of scattering and reflection data

Gavin Steininger, Jan Dettmer, Charles Holland, Stan E. Dosso

Research output: Contribution to journalArticle

Abstract

A joint trans-dimensional (trans-D) Bayesian inversion approach is developed and is applied to synthetic seabed scattering and reflection data to determine the ability of such data to resolve geoacoustic and scattering parameters. Bayesian inversion requires specifying the posterior probability density (PPD), and a method to sample PPD. To adequately approximate the PPD 600,000 models were sampled from it using the rjMCMC algorithm. These are thinned by one third to reduce sample correlation; only these remaining samples are considered here. The one- and two-dimensional marginal distributions of the scattering parameters for the inversion are found. The parameter distributions are found to be centered near the true values, and the uncertainties indicate a useful level of resolution of the roughness spectrum. Geoacoustic parameters are also well resolved, but are not shown here due to space constraints.

Original languageEnglish (US)
Pages (from-to)82-83
Number of pages2
JournalCanadian Acoustics - Acoustique Canadienne
Volume40
Issue number3
StatePublished - Sep 2012

Fingerprint

inversions
scattering
simulation
roughness

All Science Journal Classification (ASJC) codes

  • Acoustics and Ultrasonics

Cite this

@article{07292e0000b545e7bd761b358b2ac907,
title = "Simulation study of joint trans-dimenstional bayesian inversion of scattering and reflection data",
abstract = "A joint trans-dimensional (trans-D) Bayesian inversion approach is developed and is applied to synthetic seabed scattering and reflection data to determine the ability of such data to resolve geoacoustic and scattering parameters. Bayesian inversion requires specifying the posterior probability density (PPD), and a method to sample PPD. To adequately approximate the PPD 600,000 models were sampled from it using the rjMCMC algorithm. These are thinned by one third to reduce sample correlation; only these remaining samples are considered here. The one- and two-dimensional marginal distributions of the scattering parameters for the inversion are found. The parameter distributions are found to be centered near the true values, and the uncertainties indicate a useful level of resolution of the roughness spectrum. Geoacoustic parameters are also well resolved, but are not shown here due to space constraints.",
author = "Gavin Steininger and Jan Dettmer and Charles Holland and Dosso, {Stan E.}",
year = "2012",
month = "9",
language = "English (US)",
volume = "40",
pages = "82--83",
journal = "Canadian Acoustics - Acoustique Canadienne",
issn = "0711-6659",
publisher = "Canadian Acoustical Association",
number = "3",

}

Simulation study of joint trans-dimenstional bayesian inversion of scattering and reflection data. / Steininger, Gavin; Dettmer, Jan; Holland, Charles; Dosso, Stan E.

In: Canadian Acoustics - Acoustique Canadienne, Vol. 40, No. 3, 09.2012, p. 82-83.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Simulation study of joint trans-dimenstional bayesian inversion of scattering and reflection data

AU - Steininger, Gavin

AU - Dettmer, Jan

AU - Holland, Charles

AU - Dosso, Stan E.

PY - 2012/9

Y1 - 2012/9

N2 - A joint trans-dimensional (trans-D) Bayesian inversion approach is developed and is applied to synthetic seabed scattering and reflection data to determine the ability of such data to resolve geoacoustic and scattering parameters. Bayesian inversion requires specifying the posterior probability density (PPD), and a method to sample PPD. To adequately approximate the PPD 600,000 models were sampled from it using the rjMCMC algorithm. These are thinned by one third to reduce sample correlation; only these remaining samples are considered here. The one- and two-dimensional marginal distributions of the scattering parameters for the inversion are found. The parameter distributions are found to be centered near the true values, and the uncertainties indicate a useful level of resolution of the roughness spectrum. Geoacoustic parameters are also well resolved, but are not shown here due to space constraints.

AB - A joint trans-dimensional (trans-D) Bayesian inversion approach is developed and is applied to synthetic seabed scattering and reflection data to determine the ability of such data to resolve geoacoustic and scattering parameters. Bayesian inversion requires specifying the posterior probability density (PPD), and a method to sample PPD. To adequately approximate the PPD 600,000 models were sampled from it using the rjMCMC algorithm. These are thinned by one third to reduce sample correlation; only these remaining samples are considered here. The one- and two-dimensional marginal distributions of the scattering parameters for the inversion are found. The parameter distributions are found to be centered near the true values, and the uncertainties indicate a useful level of resolution of the roughness spectrum. Geoacoustic parameters are also well resolved, but are not shown here due to space constraints.

UR - http://www.scopus.com/inward/record.url?scp=84866994763&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84866994763&partnerID=8YFLogxK

M3 - Article

AN - SCOPUS:84866994763

VL - 40

SP - 82

EP - 83

JO - Canadian Acoustics - Acoustique Canadienne

JF - Canadian Acoustics - Acoustique Canadienne

SN - 0711-6659

IS - 3

ER -