Variational Bayesian geoacoustic inversion using modal dispersion

Abstract Bayesian matched-field inversion based on ocean acoustic data has been extensively studied. However, the high computing cost limit the application of conventional techniques under the Bayesian framework. This study explores the application of Variational Bayesian Monte Carlo (VBMC) optimization to geoacoustic inversion using modal dispersion data. The VBMC method offers a more efficient alternative to traditional Monte Carlo Methods by approximating the posterior distribution with a simpler form. Our approach is validated through simulations that mimic the Shallow Water Acoustic Experiment 2006. We demonstrate the effectiveness of VBMC in estimating geoacoustic parameters, leveraging its superior computational efficiency and accuracy compared to traditional Gibbs sampling methods.