Linearized Bayesian Inversion for Experiment Geometry at the New England Mud Patch

Abstract

This paper presents a linearized Bayesian approach to invert acoustic arrival-time data for high-precision estimation of experiment geometry and uncertainties for geoacoustic inversion applications. The data considered here were collected as part of the 2017 Seabed Characterization Experiment at the New England Mud Patch for the purpose of carrying out broadband reflection-coefficient inversion. The calculation of reflection coefficients requires accurate knowledge of the survey geometry. To provide this, a Bayesian ray-based inversion is developed here that estimates source–receiver ranges, source depths, receiver depths, and water depths at reflection points along the track to much higher precision than prior information based on GPS and bathymetry measurements. Near the closest point of approach, where rays are near vertical, data information is low and inaccurate range estimates are improved using priors from analytic predictions based on nearby sections of the track. Uncertainties are obtained using analytic linearized estimates, and verified with nonlinear analysis. The high-precision experiment geometry is subsequently used to calculate grazing angles, with angle uncertainties computed using Monte Carlo methods.