Jan Dettmer

Trans-dimensional geoacoustic inversion

This paper develops a general trans-dimensional Bayesian methodology for geoacoustic inversion. Trans-dimensional inverse problems are a generalization of fixed-dimensional inversion that includes the number and type of model parameters as unknowns in the problem. By extending the inversion state space to multiple subspaces of different dimensions, the posterior probability density quantifies the state of knowledge regarding inversion parameters, including effects due to limited knowledge about appropriate parametrization of the environment and error processes. The inversion is implemented here using a reversible-jump Markov chain Monte Carlo algorithm and the seabed is parametrized using a partition model. Unknown data errors are addressed by including a data-error model. Jumps between dimensions are implemented with a birth-death methodology that allows transitions between dimensions by adding or removing interfaces while maintaining detailed balance in the Markov chain. Trans-dimensional inversion results in an inherently parsimonious solution while partition modeling provides a naturally self-regularizing algorithm based on data information content, not on subjective regularization functions. Together, this results in environmental estimates that quantify appropriate seabed structure as supported by the data, allowing sharp discontinuities while approximating smooth transitions where needed. This approach applies generally to geoacoustic inversion and is illustrated here with seabed reflection-coefficient data.

Model selection and Bayesian inference for high-resolution seabed reflection inversion

This paper applies Bayesian inference, including model selection and posterior parameter inference, to inversion of seabed reflection data to resolve sediment structure at a spatial scale below the pulse length of the acoustic source. A practical approach to model selection is used, employing the Bayesian information criterion to decide on the number of sediment layers needed to sufficiently fit the data while satisfying parsimony to avoid overparametrization. Posterior parameter inference is carried out using an efficient Metropolis-Hastings algorithm for high-dimensional models, and results are presented as marginal-probability depth distributions for sound velocity, density, and attenuation. The approach is applied to plane-wave reflection-coefficient inversion of single-bounce data collected on the Malta Plateau, Mediterranean Sea, which indicate complex fine structure close to the water-sediment interface. This fine structure is resolved in the geoacoustic inversion results in terms of four layers within the upper meter of sediments. The inversion results are in good agreement with parameter estimates from a gravity core taken at the experiment site.

Trans-dimensional matched-field geoacoustic inversion with hierarchical error models and interacting Markov chains

This paper develops a trans-dimensional approach to matched-field geoacoustic inversion, including interacting Markov chains to improve efficiency and an autoregressive model to account for correlated errors. The trans-dimensional approach and hierarchical seabed model allows inversion without assuming any particular parametrization by relaxing model specification to a range of plausible seabed models (e.g., in this case, the number of sediment layers is an unknown parameter). Data errors are addressed by sampling statistical error-distribution parameters, including correlated errors (covariance), by applying a hierarchical autoregressive error model. The well-known difficulty of low acceptance rates for trans-dimensional jumps is addressed with interacting Markov chains, resulting in a substantial increase in efficiency. The trans-dimensional seabed model and the hierarchical error model relax the degree of prior assumptions required in the inversion, resulting in substantially improved (more realistic) uncertainty estimates and a more automated algorithm. In particular, the approach gives seabed parameter uncertainty estimates that account for uncertainty due to prior model choice (layering and data error statistics). The approach is applied to data measured on a vertical array in the Mediterranean Sea.

Remote sensing of sediment density and velocity gradients in the transition layer

The geoacoustic properties of marine sediments, e.g., bulk density and compressional velocity, commonly exhibit large variations in depth near the water-sediment interface. This layer, termed the transition layer, is typically of 0(10(-1)-10(0)) m in thickness. Depth variations within the transition layer may have important implications for understanding and modeling acoustic interaction with the seabed, including propagation and reverberation. In addition, the variations may contain significant clues about the underlying depositional or erosional processes. Characteristics of the transition layer can be measured directly (e.g., coring) or remotely. Remote measurements have the advantage of sampling without disturbing the sediment properties; they also have the potential to be orders of magnitude faster and less expensive than direct methods. It is shown that broadband seabed reflection data can be exploited to remotely obtain the depth dependent density and velocity profiles in the transition layer to high accuracy. A Bayesian inversion approach, which accounts for correlated data errors, provides estimates and uncertainties for the geoacoustic properties. These properties agree with direct (i.e., core) measurements within the uncertainty estimates.

Uncertainty estimation in seismo-acoustic reflection travel time inversion.

This paper develops a nonlinear Bayesian inversion for high-resolution seabed reflection travel time data including rigorous uncertainty estimation and examination of statistical assumptions. Travel time data are picked on seismo-acoustic traces and inverted for a layered sediment sound-velocity model. Particular attention is paid to picking errors which are often biased, correlated, and nonstationary. Non-Toeplitz data covariance matrices are estimated and included in the inversion along with unknown travel time offset (bias) parameters to account for these errors. Simulated experiments show that neglecting error covariances and biases can cause misleading inversion results with unrealistically high confidence. The inversion samples the posterior probability density and provides a solution in terms of one- and two-dimensional marginal probability densities, correlations, and credibility intervals. Statistical assumptions are examined through the data residuals with rigorous statistical tests. The method is applied to shallow-water data collected on the Malta Plateau during the SCARAB98 experiment.

Bayesian geoacoustic inversion using wind-driven ambient noise.

This paper applies Bayesian inversion to bottom-loss data derived from wind-driven ambient noise measurements from a vertical line array to quantify the information content constraining seabed geoacoustic parameters. The inversion utilizes a previously proposed ray-based representation of the ambient noise field as a forward model for fast computations of bottom loss data for a layered seabed. This model considers the effect of the arrays finite aperture in the estimation of bottom loss and is extended to include the wind speed as the driving mechanism for the ambient noise field. The strength of this field relative to other unwanted noise mechanisms defines a signal-to-noise ratio, which is included in the inversion as a frequency-dependent parameter. The wind speed is found to have a strong impact on the resolution of seabed geoacoustic parameters as quantified by marginal probability distributions from Bayesian inversion of simulated data. The inversion method is also applied to experimental data collected at a moored vertical array during the MAPEX 2000 experiment, and the results are compared to those from previous active-source inversions and to core measurements at a nearby site.

Bayesian matched-field geoacoustic inversion

This paper describes a Bayesian approach to matched-field inversion (MFI) of ocean acoustic data for seabed geoacoustic properties. In a Bayesian formulation, the unknown environmental and experimental parameters are considered random variables constrained by noisy data and prior information, and the goal is to interpret the multi-dimensional posterior probability density (PPD). The PPD is typically characterized in terms of point estimates, marginal distributions, and posterior correlations (or joint statistics). Computing these requires numerical optimization and integration of the PPD, which are carried out efficiently here using adaptive hybrid optimization and Metropolis‐ Hastings sampling in principal-component space, respectively. Likelihood and misfit functions for multi-frequency MFI with incomplete source spectral information are derived based on the assumption of complex Gaussiandistributed data errors with covariance matrices estimated from residual analysis; posterior statistical tests are applied to validate these estimates and assumptions. Model selection is carried out by applying the Bayesian information criterion to determine the simplest seabed parameterization consistent with the resolving power of the data. Bayesian MFI is illustrated for shallow-water acoustic data measured in the Mediterranean Sea. (Some figures in this article are in colour only in the electronic version)

Joint time/frequency-domain inversion of reflection data for seabed geoacoustic profiles and uncertainties

This paper develops a joint time/frequency-domain inversion for high-resolution single-bounce reflection data, with the potential to resolve fine-scale profiles of sediment velocity, density, and attenuation over small seafloor footprints (approximately 100 m). The approach utilizes sequential Bayesian inversion of time- and frequency-domain reflection data, employing ray-tracing inversion for reflection travel times and a layer-packet stripping method for spherical-wave reflection-coefficient inversion. Posterior credibility intervals from the travel-time inversion are passed on as prior information to the reflection-coefficient inversion. Within the reflection-coefficient inversion, parameter information is passed from one layer packet inversion to the next in terms of marginal probability distributions rotated into principal components, providing an efficient approach to (partially) account for multi-dimensional parameter correlations with one-dimensional, numerical distributions. Quantitative geoacoustic parameter uncertainties are provided by a nonlinear Gibbs sampling approach employing full data error covariance estimation (including nonstationary effects) and accounting for possible biases in travel-time picks. Posterior examination of data residuals shows the importance of including data covariance estimates in the inversion. The joint inversion is applied to data collected on the Malta Plateau during the SCARAB98 experiment.

Tsunami waveform inversion for sea surface displacement following the 2011 Tohoku earthquake: Importance of dispersion and source kinematics

This paper considers the importance of model parameterization, including dispersion, source kinematics, and source discretization, in tsunami source inversion. We implement single and multiple time window methods for dispersive and nondispersive wave propagation to estimate source models for the tsunami generated by the 2011 Tohoku-Oki earthquake. Our source model is described by sea surface displacement instead of fault slip, since sea surface displacement accounts for various tsunami generation mechanisms in addition to fault slip. The results show that tsunami source models can strongly depend on such model choices, particularly when high-quality, open-ocean tsunami waveform data are available. We carry out several synthetic inversion tests to validate the method and assess the impact of parameterization including dispersion and variable rupture velocity in data predictions on the inversion results. Although each of these effects has been considered separately in previous studies, we show that it is important to consider them together in order to obtain more meaningful inversion results. Our results suggest that the discretization of the source, the use of dispersive waves, and accounting for source kinematics are all important factors in tsunami source inversion of large events such as the Tohoku-Oki earthquake, particularly when an extensive set of high-quality tsunami waveform recordings are available. For the Tohoku event, a dispersive model with variable rupture velocity results in a profound improvement in waveform fits that justify the higher source complexity and provide a more realistic source model.

Trans-dimensional joint inversion of seabed scattering and reflection data.

This paper examines joint inversion of acoustic scattering and reflection data to resolve seabed interface roughness parameters (spectral strength, exponent, and cutoff) and geoacoustic profiles. Trans-dimensional (trans-D) Bayesian sampling is applied with both the number of sediment layers and the order (zeroth or first) of auto-regressive parameters in the error model treated as unknowns. A prior distribution that allows fluid sediment layers over an elastic basement in a trans-D inversion is derived and implemented. Three cases are considered: Scattering-only inversion, joint scattering and reflection inversion, and joint inversion with the trans-D auto-regressive error model. Including reflection data improves the resolution of scattering and geoacoustic parameters. The trans-D auto-regressive model further improves scattering resolution and correctly differentiates between strongly and weakly correlated residual errors.