Bayesian geoacoustic inversion of ship noise on a horizontal array

Abstract

This paper applies geoacoustic inversion to low-frequency narrow-band acoustic data from a quiet surface ship recorded on a bottom-moored horizontal line array in shallow water. A Bayesian matched-field inversion method is employed which quantifies geoacoustic uncertainties and allows for meaningful comparison of inversion results from different data sets. Geoacoustic inversion results for ship-noise data are compared with inversion results for multitone data from a towed controlled source collected in the same experiment, and with independent geophysical measurements. To increase the information content of low-level ship-noise data, the effect of including multiple, independent data segments in the inversion is investigated and shown to significantly reduce geoacoustic parameter uncertainties. Geoacoustic uncertainties are also shown to depend on ship range and orientation, with increased uncertainties for long ranges and for the ship stern oriented away from the array.

Three-dimensional source tracking in an uncertain environment via Bayesian marginalization

“…8 Acoustic data were collected with an 18-element HLA of length 900 m deployed on the relatively flat seabed at a depth of approximately 282 m. The acoustic data considered here were due to a continuous-wave tone emitted by a towed acoustic source (depth 54 m) at a frequency of 80 Hz for source-to-array (closest end) ranges of 4.0-4.7 km, and ship noise due to the R/V H U SVERDRUP II (speed 5.2 kn) at ranges of 5.1-5.8 km at three frequencies (40 Hz, 50 Hz, and 144 Hz), both data sets collected along a radial track oriented at 30°angle with respect to the array endfire. Each data set is comprised of nine data segments extending over a time span of 4 min 42 s over which the source/ship moved approximately 755 m in range.…”

“…The lower layer is described by constant sound-speed and attenuation values that are identical to those at the base of the upper layer, and by an independent density 2 . Table 1 lists the geoacoustic model parameters, their estimated values from Bayesian matched-field inversion, 8 and the uniform prior bounds employed for tracking. Numerical grids from 1 to 9 km in range (50 m spacing) and from 6 to 270 m in depth (2 m spacing) were used for source coordinates.…”

“…8 The parameters that describe the upper seabed layer are the layer thickness ͑h͒, sound speed at the top and bottom of the layer (c 1 and c 2 ), attenuation coefficient at the top and bottom of the layer (␣ 1 and ␣ 2 ), and a depth-independent density 1 . The lower layer is described by constant sound-speed and attenuation values that are identical to those at the base of the upper layer, and by an independent density 2 .…”

“…Under the assumption of uncorrelated, complex Gaussian-distributed errors with unknown source amplitude and phase and unknown error variances the appropriate data misfit function, derived in Ref. 8…”

Maximum entropy inference of seabed attenuation parameters using ship radiated broadband noise

“…(2) shows a clear ambiguity between source strength and loss parameters in G(f i ), such as the attenuation (and indirectly the sediment sound speed), it is clear that v i must be part of the statistical inference problem; namely, both source levels and waveguide parameters need to be treated as random variables. Tollefson and Dosso were the first to apply a Bayesian inference approach, rather than a geoacoustic inversion method, to obtain information from ship radiated noise about both the a) Electronic mail: dpknobles@yahoo.com geoacoustic properties of the seabed and v i for selected frequencies, 8 but, like Hamilton, they assumed that c was equal to unity.…”