Environmental inversion using high-resolution matched-field processing

Abstract

This paper considers the inversion of experimental field data collected with light receiving systems designed to meet operational requirements. Such operational requirements include system deployment in free drifting configurations and a limited number of acoustic receivers. A well-known consequence of a reduced spatial coverage is a poor sampling of the vertical structure of the acoustic field, leading to a severe ill-conditioning of the inverse problem and data to model cost function with a massive sidelobe structure having many local extrema. This causes difficulties to meta-heuristic global search methods, such as genetic algorithms, to converge to the true model parameters. In order to cope with this difficulty, broadband high-resolution processors are proposed for their ability to significantly attenuate sidelobes, as a contribution for improving convergence. A comparative study on simulated data shows that high-resolution methods did not outperform the conventional Bartlett processor for pinpointing the true environmental parameter when using exhaustive search. However, when a meta-heuristic technique is applied for exploring a large multidimensional search space, high-resolution methods clearly improved convergence, therefore reducing the inherent uncertainty on the final estimate. These findings are supported by the results obtained on experimental field data obtained during the Maritime Rapid Environmental Assessment 2003 sea trial.

Broadband synthetic aperture geoacoustic inversion

“…Comparing Eq. (16) with other models [25][26][27] and examining Eq. (15), the source spectrum is coupled within, and noncommutable with, the integral of the Green's function due to source/receiver motion.…”

“…[25][26][27] Here, this model is improved by including source/receiver motion and realistic noise assumptions,…”

“…6,26 Under the uniform Doppler assumption, the source spectrum is frequency shifted as if in free space [Eq. (27)]. The resulting source spectrum S½x ðv s Þ s then is propagated through the normal mode model, where all the modes are given the same Doppler shift for a given frequency,…”

Acoustic observatories for ocean tomography: Multi-array matched-field tomography

“…The data model considered herein is an extension of the broadband data model discussed in Ref. [10], which is a concatenation of single frequency components of the signal observed across a vertical array. In the present case the idea is to include frequency components of a signal observed across two or more vertical arrays.…”

“…The parameter estimators described in section II-B use the broadband Bartlett processor [10] as the functional for comparison of the replica fields with the observed field based on the multi-array data model:…”