Bayesian hindcast of acoustic transmission loss in the western Pacific Ocean

Abstract

AbstractA Bayesian network is developed to demonstrate the feasibility of using environmental acoustic feature vectors (EAFVs) to predict underwater acoustic transmission loss (TL) versus range at two locations for a single acoustic source depth and frequency. Features for the networks are chosen based on a sensitivity analysis. The final network design resulted in a well‐trained network, with high skill, little gain error, and low bias. The capability presented here shows promise for expansion to a more generalized approach, which could be applied at varying locations, depths and frequencies to estimate acoustic performance over a highly variable oceanographic area in real‐time or near‐real‐time.