Geoacoustic parameter estimation through iteratively particle filtering of solitary data set

In Bayesian inversion schemes, geoacoustic parameter values and associated uncertainties normally are obtained through sampling approaches, e.g., Metropolis-Hastings (M-H) sampling. This paper presents an attempt to solve this problem through applying a particle filter (PF) to solitary data input as collected in the Yellow Sea, China, 2002. In the experiment, a 38-g explosive charge was used as a broadband source that detonated at an approximate depth of 7 m. The recording system is a moored vertical line array composed of 30 pressure sensors. Comparison between the performances of PF and M-H sampling is first made with numerical tests based on the Yellow Sea environmental model, and then are applied to the real at-sea data. Results from both synthetic and real data processing demonstrate the PF is an efficient approach in geoacoustic parameters estimation.