Application of a stochastic inversion method for estimating the sound speed in the ocean bottom

Abstract

A stochastic inversion method is applied to the problem of determining sound-speed profile in bottom sediment from modal eigenvalues, which can be observed from the FFT beamformed output of a horizontal array in a range-independent shallow-water waveguide. Under a noisy field, each observed eigenvalue has different variance depending on the signal-to-noise ratio (SNR) of modal peaks in the output wave-number spectrum; this difference in SNR originates from the relation between mode shape and source/receiver depths. Thus by using those variances in a covariance matrix involved in stochastic inversion method, it is numerically demonstrated that we can improve the estimation for the solution as compared to that obtained by using a uniform covariance matrix. It is also shown that combination of the same modal eigenvalues observed at the different source or receiver depth can reduce the variance of estimated solution. Application of the stochastic inversion method to real data is also discussed together with comparison with a regularization method. [Work support by JDA.]