2014
DOI: 10.1121/1.4899489
|View full text |Cite
|
Sign up to set email alerts
|

Efficient trans-dimensional Bayesian inversion for geoacoustic profile estimation

Abstract: This paper considers sampling efficiency of trans-dimensional (trans-D) Bayesian inversion based on the reversible-jump Markov-chain Monte Carlo (rjMCMC) algorithm, with application to seabed acoustic reflectivity inversion. Trans-D inversion is applied to sample the posterior probability density over geoacoustic parameters for an unknown number of seabed layers, providing profile estimates with uncertainties that include the uncertainty in the model parameterization. However, the approach is computationally i… Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
1
1
1
1

Citation Types

0
15
0

Year Published

2015
2015
2020
2020

Publication Types

Select...
7

Relationship

1
6

Authors

Journals

citations
Cited by 8 publications
(15 citation statements)
references
References 38 publications
0
15
0
Order By: Relevance
“…Trans-dimensional inversion represents a powerful and general approach that has been applied for several geophysical problems (e.g., Malinverno 2002;Sambridge et al 2006;Bodin and Sambridge 2009;Dettmer et al 2010;Bodin et al 2012;Dosso et al 2014). Ray and Key (2012) implemented a trans-dimensional inversion for frequency-domain CSEM data to estimate the resolution of different field components of the electromagnetic response, and they illustrated the challenges of CSEM inversion when targeting deep-situated, thin hydrocarbon reservoirs in anisotropic environments.…”
Section: Inversionmentioning
confidence: 99%
See 1 more Smart Citation
“…Trans-dimensional inversion represents a powerful and general approach that has been applied for several geophysical problems (e.g., Malinverno 2002;Sambridge et al 2006;Bodin and Sambridge 2009;Dettmer et al 2010;Bodin et al 2012;Dosso et al 2014). Ray and Key (2012) implemented a trans-dimensional inversion for frequency-domain CSEM data to estimate the resolution of different field components of the electromagnetic response, and they illustrated the challenges of CSEM inversion when targeting deep-situated, thin hydrocarbon reservoirs in anisotropic environments.…”
Section: Inversionmentioning
confidence: 99%
“…Probabilistic sampling approaches, unlike linearized inversions, can provide rigorous estimates of model parameter uncertainties (e.g., Malinverno 2002). Trans-dimensional Bayesian inversion has recently been applied to invert various geophysical data sets (e.g., Bodin and Sambridge 2009;Dettmer, Dosso and Holland 2010;Ray et al 2013a;Dosso et al 2014).…”
Section: Introductionmentioning
confidence: 99%
“…For this reason, the algorithm can be improved by sampling from an approximate posterior distribution rather than from the prior model. To this end, many strategies could be adopted, for example, implementing the adaptive Metropolis algorithm (Haario et al 2001), or sampling from an approximated estimate of the posterior covariance matrix computed from a local approximation of the Jacobian matrix (see Dosso et al 2014). In a 2D interval-oriented inversion of field seismic data we performed (not shown here for confidentiality reasons), we also found particularly useful deriving the starting model for the MCMC sampling from the results of a previous Bayesian linear inversion.…”
Section: Discussionmentioning
confidence: 99%
“…While this appears to improve the acceptance of resistivity and move steps, we find that the acceptance rates of transdimensional steps (birth and death) are normally very low (<10%). Hence, a natural extension of this study might include attempting to improve the acceptance rates of transdimensional steps by implementing a transdimensional delayed rejection scheme as developed by Green and Mira (2001), and/or by sampling the log(resistivity) value for newly generated Voronoi cells from the prior rather than using a Gaussian perturbation (equation (B16)) as suggested by Dosso et al (2014).…”
Section: Convergencementioning
confidence: 99%
“…Plots such as those shown in Figure 20 may therefore be used to choose an appropriate value for bd , such that the crossing point of the two curves is greater than 1 or 2 standard deviations from i , ensuring that natural parsimony will be achieved on average. Alternatively, the resistivity of a newly generated cell in a birth step may be sampled from the prior as suggested by Dosso et al (2014), which obviates the need for such tuning when prior ranges are narrow.…”
Section: Similar Tomentioning
confidence: 99%