Bayesian tracking of time or space varying environment from ship noise recorded on a drifting vector sensor

In this paper, a Bayes filter is adapted to determine environmental variation using ship noise data measured on a vector sensor. As compared to batch processing approaches, the Bayesian framework can monitor ocean processes or environments that substantially vary in time or space. The use of vertical impedance (a ratio of pressure and vertical particle velocity) is emphasized here, which is shown to be source spectral independent but highly sensitive to environmental properties. The scenario tested model is inspired from environmental and acoustic data collected at the Amazon River mouth in June 2012. Results show that Bayesian approach can effectively resolve environmental properties variations along range, suggesting the feasibility of using this approach for complex range-dependent environmental characterization.