2025_programme: Bayesian Inversion for Geoacoustic Parameters in the Baltic Sea

• Day: June 16, Monday
    Location / Time: B. ERATO at 15:10-15:30
• Last minutes changes: -
• Session: 07. Inverse Problems in Acoustical Oceanography
  Organiser(s): Julien Bonnel, Stan Dosso
  Chairperson(s): Julien Bonnel, Stan Dosso
• Lecture: Bayesian Inversion for Geoacoustic Parameters in the Baltic Sea
  Paper ID: 2299
  Author(s): Siobhán Correnty, Ylva Ljungberg Rydin, Magnus Lundberg Nordenvad, Isaac Skog, Martin Östberg
  Presenter: Siobhán Correnty
  Abstract: Environmental parameters, such as acoustic seabed parameters, significantly influence the propagation of acoustic waves in shallow water. Collecting in-situ measurements of these parameters is time-consuming and costly, making it challenging to survey large geographic areas. An alternative approach to obtaining information about environmental parameters is through acoustic inversion methods. While optimization-based inversion methods have long been used to estimate environmental parameters from acoustic data, they typically provide only point estimates of the parameters and, at best, second-order statistics. Consequently, the uncertainty in the estimated parameters cannot be fully quantified. Additionally, these methods make it difficult to incorporate prior information about the model parameters in a structured way.\n \nRecent advancements in Monte Carlo methods have enabled the application of Bayesian inference techniques to complex models, allowing for the calculation of full posterior probability densities of the model parameters given observed data [1]. In this work, we investigate how Monte Carlo methods can perform Bayesian estimation of environmental parameters in acoustic propagation models. More specifically, an acoustic propagation model based on wave number integration and parabolic equations is considered, and a Metropolis-Hastings algorithm is employed to calculate the posterior probability densities of the seabed parameters given transmission-loss data; a likelihood function like the one presented in [2] is utilized to weigh the measurements. Using the implemented Bayesian inversion method, we conduct simulation experiments that emulate the acoustic environment of the Baltic Sea to examine how uncertainties in the measured data affect the estimation of seabed parameters. Understanding this is essential to adopting an information theoretical experiment design approach in the data collection process and maximizing the information gathered about the acoustic seabed parameters. \n \n[1] A. Wigren, J. Wågberg, F. Lindsten, A. G. Wills, and T. B. Schön, Nonlinear system identification: Learning while\nrespecting physical models using a sequential Monte Carlo method, IEEE Control Systems Magazine, 2022.\n \n[2] G. Zheng, H. Zhu, X. Wang, S. Khan, N. Li, and Y. Xue. Bayesian inversion for geoacoustic param-\neters in shallow sea. Sensors (Basel), 2020.\n
    Download the full paper
  This paper is a candidate for the "Prof. John Papadakis award for the best paper presented by a young acoustician(under 40)"
• Corresponding author: Dr Siobhán Correnty
  Affiliation: Swedish Defence Research Agency (FOI)
  Country: Sweden