Posterior-covariance analysis under varying survey geometry and navigation accuracy in moving-platform ocean acoustic tomography

Abstract

Ocean acoustic tomography (OAT) is an acoustic remote-sensing technique that maps the ocean’s interior sound-speed structure from travel-time measurements, complementing satellite and surface remote sensing that observe mainly the sea surface. On moving ship platforms it becomes a marine acoustic survey whose accuracy hinges on platform positioning and navigation (e.g., GNSS). When the source and receiver move simultaneously (dual-ship), position and velocity uncertainties become entangled with the environmental sound-speed perturbation, degrading the inversion. Within a linear- Gaussian Bayesian framework, this work formulates a unified observation model that jointly contains the environmental sound-speed parameters and the geometric and clock parameters of two moving platforms. The geometric sensitivities are assembled by central finite differencing of the forward travel-time model, independent of the forward solver. Because the posterior covariance depends only on the observation operator, the data-error covariance and the prior, and not on the data realization, the environmental posterior uncertainty is evaluated for four geometries (fixed, single-ship, dual-ship parallel and dual-ship converging) across navigation accuracies and propagation distances. Simulations indicate that coverage gained from platform motion lowers the environmental posterior uncertainty, but the gain depends on the track pattern: parallel tracks that translate the aperture outperform converging tracks; relaxing the navigation prior from centimeter to meter level raises the uncertainty, more strongly at short range. For a target whole-section uncertainty of about 1.4 m/s, the parallel geometry tolerates a roughly three-fold looser positioning prior than the fixed array, providing a quantitative posterior basis for choosing the survey geometry and its navigation-accuracy requirement.