Bayesian closed-skew Gaussian inversion is defined as a generalization of traditional Bayesian Gaussian inversion, which is used frequently in seismic amplitude-versus-offset (AVO) inversion. The new model captures skewness in the variables of interest; hence, the posterior model for log-transformed elastic material properties given seismic AVO data might be a skew probability density function. The model is analytically tractable, and this makes it applicable in high-dimensional 3D inversion problems. Assessment of the posterior models in high dimensions requires numerical approximations, however. The Bayesian closed-skew Gaussian inversion approach has been applied on real elastic material properties from a well in the Sleipner field in the North Sea. A comparison with results from traditional Bayesian Gaussian inversion shows that the mean square error of predictions of P-wave and S-wave velocities are reduced by a factor of two, although somewhat less for density predictions.
[1]
Niels Bohr,et al.
Monte Carlo sampling of solutions to inverse problems
,
2004
.
[2]
M. Mohammadzadeh,et al.
Bayesian Spatial Prediction for Discrete Closed Skew Gaussian Random Field
,
2011
.
[3]
Arjun K. Gupta,et al.
The Closed Skew-Normal Distribution
,
2004
.
[4]
A. Buland,et al.
Bayesian linearized AVO inversion
,
2003
.
[5]
Arild Buland,et al.
Rapid spatially coupled AVO inversion in the Fourier domain
,
2003
.
[6]
P. Embrechts.
Skew-Elliptical Distributions and Their Applications: A Journey Beyond Normality
,
2005
.
[7]
J. Scales,et al.
Bayesian seismic waveform inversion: Parameter estimation and uncertainty analysis
,
1998
.
[8]
Marc G. Genton,et al.
Skew-elliptical distributions and their applications : a journey beyond normality
,
2004
.