Probabilistic two-dimensional water-column and seabed inversion with self-adapting parameterizations

Abstract

This paper develops a probabilistic two-dimensional (2D) inversion for geoacoustic seabed and water-column parameters in a strongly range-dependent environment. Range-dependent environments in shelf and shelf-break regions are of increasing importance to the acoustical-oceanography community, and recent advances in nonlinear inverse theory and sampling methods are applied here for efficient probabilistic range-dependent inversion. The 2D seabed and water column are parameterized using highly efficient, self-adapting irregular grids which intrinsically match the local resolving power of the data and provide parsimonious solutions requiring few parameters to capture complex environments. The self-adapting parameterization is achieved by implementing the irregular grid as a trans-dimensional hierarchical Bayesian model with an unknown number of nodes which is sampled with the Metropolis-Hastings-Green algorithm. To improve sampling, population Monte Carlo is applied with a large number of interacting parallel Markov chains with adaptive proposal distributions. The inversion is applied to simulated data for a vertical-line array and several source locations to several kilometers range. Complex acoustic-pressure fields are computed using a parabolic equation model and results are considered in terms of 2D ensemble parameter estimates and credibility intervals.

Bayesian Seismic Refraction Inversion for Critical Zone Science and Near‐Surface Applications

“…Transdimensional approaches to traveltime tomography in 2D (Bodin & Sambridge, 2009) were extended to explicitly estimate data error through hierarchical parameters (Bodin et al., 2012) and to 3D tomography on local (Piana Agostinetti et al., 2015) and continental (Burdick & Lekic, 2017) scales. THB approaches have been applied to the inversion of controlled‐source data in 2D dimensions: geoacoustic imaging (Dettmer & Dosso, 2013), full waveform reflection inversion (Ray et al., 2018; Visser et al., 2019), marine electro‐magnetic sounding (Ray & Myer, 2019; Ray et al., 2014), electrical resistivity imaging (Galetti & Curtis, 2018), and refraction traveltime tomography (Ryberg & Haberland, 2018).…”

A trans-dimensional polynomial-spline parameterization for gradient-based geoacoustic inversion

“…[2][3][4][5][6][7][8] In particular, trans-D Bayesian inversion provides an effective, automated approach to model selection (e.g., determining the number of sediment layers consistent with the resolving power of the data), such that the parameter estimate uncertainties account for model selection uncertainties. [9][10][11][12][13][14][15][16] As the concept of model parsimony is linked to the choice of model parameterization type, the posterior uncertainty of parameter estimates are still affected by a priori parameterization selection decisions. This work highlights the importance of model parameterization in Bayesian inversion a) Author to whom correspondence should be addressed.…”

Change-point detection for recursive Bayesian geoacoustic inversions

“…III, the method is demonstrated in simulation with parameters corresponding to the low SNR, 100-900 Hz linear frequency modulation (LFM) pulses observed in the Shallow Water 2006 experiment. The change-point approach also could be applied to local matched-field geoacoustic inversions using towed arrays, [12][13][14][15][16] but this scenario is not considered here.…”