Study Of Multi-mode Interface-wave Dispersion Curves Inversion Based On Nonlinear Bayesian Theory

The detection and recognition study of seafloor is one of the most active front fields of the marine sciences research, which has an important academic significance and application value in the exploration and exploitation of seabed recourses, ocean environment monitoring, marine engineering construction and military operations. The inversion of interface wave dispersion curves is one of the most effective techniques to study the seabed structures and detect the seafloor resources. Recently, interface-wave inversion work has been getting more and more attention to estimate the shear-wave velocity profile structures. Shear-wave velocity provides a good indicator of sediment rigidity for seafloor geotechnical applications and shear-wave transition in shallow seabed can represent an important ocean acoustic loss mechanism which must be considered in propagation modeling and sonar performance predictions.The extraction resolution of interface-wave dispersion curves is a key of success or failure in inversion. The dispersion curves were extracted normally from the seismic interface-wave propagating in seabed sediments by a mean of a natural or artificial earthquake focus. The resolution of dispersion curves is affected by the distributions of source and stations. It is very important to develop a high accuracy and resolution method independent of the seismic focus to obtain the interface-wave dispersion curves and the sediment parameters are estimated by inverting the dispersion curves. In additional, inversion of interface-wave dispersion curves is typically a highly nonlinear, multi-parameter, and multi-optima inversion problem, as well as most other marine geophysical optimization problems. Usually there is not an only optimal solution due to the measurement and calculation errors. The inversion results of traditional linear search methods depend on the choice of a starting model and the accuracy of the partial derivatives, which are prone to being trapped by local minima. The available nonlinear inversion techniques such as genetic algorithm (GA) and simulated annealing (SA), have proven to be quite useful for global optimization problems, but these techniques can only determine the best-fit model and slower convergence and not provide quantitatively nonlinear uncertainty estimation of model parameters. It is necessary to develop a quantitative nonlinear uncertainty estimation approach combined global and local optimization to seabed interface-wave dispersion inversion in a rigorous manner.This paper applies a dataset of ocean seismic ambient noise data recorded by ocean bottom seismic cable to extract multi-mode interface-wave dispersion curves. The nonlinear Bayesian inversion is applied to estimate seabed sediment parameters and their uncertainties from interface-wave dispersion curves. The main contents of this paper include analysis of tranditional inversion theory and methods, interface-wave dispersion curves extracting from the seismic ambient noise data, the expressions and computional method of dispersion equations, nonlinear Bayesian inversion formulation and the study of multi-mode interface-wave dispersion curves inversion.The main contributions are as follows:1.A high accuracy and resolution approach to extract the interface-wave dispersion curves are applied from the ocean seismic ambient noise data. The fundamental principles of the cross-correlation, Green's function and time-frequency analysis are illustrated and the general process to extract the interface-wave dispersion curves is summarized. The ambient noise has the dispersion characteristics because of its random sources and the effects of multiple scattering, so the resolution of dispersion is independent of the seismic sources. The multi-mode interface-wave dispersion curves are obtained by processing the noise signals recorded by an ocean bottom seismic.2.An appropriate forward model is built and resolved to obtain the phase velocity dispersion of interface-wave on multilayered media. The forward model is a prerequisite for inversion and determines the precision and reliability of inversion. The Thompson-Haskill matrix formalism are used to compute the phase velocity dispersion equations for elastic plane waves propagated in a semi-infinite media with n parallel, homogeneous, istropic layers. The numerical case is applied to certify the effective of this method and illustrate this method has relative simple forms, obvious physical meanings of parameters, and the advantages of easy computing and programming. 3.A nonlinear Bayesian inversion approach is applied in this paper. In a Bayesian formulation, the multi-dimensional PPD represents the general solution to an inversion problem. The maximum a posterior (MAP) estimates are determined by minimizing misfit function numerically using adaptive simplex simulated annealing (ASSA), an effective hybrid optimization algorithm that combines the local downhill-simplex method with a very fast simulated annealing global search. The Bayesian information criterion (BIC) is applied to determine the optimal model that fully explains the observed data by the different parameterizations. An efficient and complete Metropolis Hastings sampling (MHS) algorithm is applied for computing 1-D and 2-D parameter marginal probability distributions. Bayesian inversion technique not only provides the MAP estimation, but also estimates the uncertainties of unkown parameters.4.The mechanism of multi-mode interface-wave dispersion curves is studied and analyzed. It is usually considered that the fundamental mode of interface-wave dominates the recorded wavefield and higher modes can be ignored. However, the higher modes contribute significant amounts of energy at higher frequencies and the higher modes of interface wave can be applied to improve the accuracy of the inverted shear-wave velocity profiles. In this paper, inversion studies are carried out for both the fundamental mode alone and for the first three modes to determine the best parameterization for the shear-wave velocity profiles considering the different parameterizations. The results show that the multi-mode of interface-wave is more sensitive to the seabed sediment structures than the fundamental mode. The multi-mode interface-wave dispersion curves inversion can obtain more shear-wave velocity profile structures, especially the near-interface structures with low-velocities layers; improve the resolution and reliability of shear-wave velocities.The innovation of the dissertation is appling nonlinear Bayesion inversion to estimate seabed shear-wave velocity profiles and their uncertainties using multi-mode interface wave dispersion curves extracted from ocean ambient noise data.