The impact of ocean sound speed variability on the uncertainty of geoacoustic parameter estimates N. Ross Chapman and Yong-Min Jiang University of Victoria Victoria BC Canada University of Victoria, Victoria, BC, Canada Work supported by ONR Chapman and Jiang ASA157, Portland, Oregon, 18 - 22 May 2009
Objective: • Describe a simple technique to account for unknown ocean sound speed profile in matched field inversion sound speed profile in matched-field inversion � Invert for an effective SSP that creates a range independent propagation environment p p g • Method: � Use EOFs to parameterize the SSP � Use EOFs to parameterize the SSP � What information is necessary? � Large data set of SSP over extended space and time � Limited data set in vicinity of experiment in space and time � Li it d d t t i i i it f i t i d ti • Hypothesis is that limited set will be adequate if changes in SSP are not large Chapman and Jiang ASA157, Portland, Oregon, 18 - 22 May 2009
Experimental site: Acoustic array (MPL): Acoustic array (MPL): • VLA1 • 16 sensors, 3.75 m separation • the bottom one is 8.2 m from the bottom one is 8.2 m from the sea floor Source ship stations, distance to VLA1: VLA1: • WP21, 1 km • WP22, 3 km • Wp23 5 km • Wp23, 5 km Water depth: • ~79.0m Signal frequencies (CW tonals): • LF: 53, 103, 203, and 253 Hz • MF: 303, 403 503, 703 and 303, 03 503, 03 953 Hz Chapman and Jiang ASA157, Portland, Oregon, 18 - 22 May 2009
Geoacoustic model • Invert for: • geometric parameters of the experiment and the experiment, and • geoacoustic model parameters • Approach: Bayesian Matched field inversion Geoacoustic model for the SW06 site Chapman and Jiang ASA157, Portland, Oregon, 18 - 22 May 2009
Recap: Is SSP at the source all we need? p SSPs measured at source and VLA1 Ambiguity surface of MFP (source localization) Chapman and Jiang ASA157, Portland, Oregon, 18 - 22 May 2009
SSP data: SSPs measured at source and VLA1 (derived from CTDs) (derived from CTDs) SSPs measured at SHARK, SW31 and SW32 Chapman and Jiang ASA157, Portland, Oregon, 18 - 22 May 2009
Approaches: pp • Limited set from CTDs measured at source ship stations p • change only in the thermocline • requires fewer EOFs (only 4 EOFs) • Full set from oceanographic moorings and CTDs from source ship stations source ship stations • cover whole water column • need more EOFs need more EOFs • how to decide how many EOFs? Chapman and Jiang ASA157, Portland, Oregon, 18 - 22 May 2009
Comparison of the energy fit versus the number p gy of EOFs used: Chapman and Jiang ASA157, Portland, Oregon, 18 - 22 May 2009
Example: 8EOFs for large SSP sample set: Base line profile 1 EOFs 2 EOFs 3 EOFs 0 0 0 0 20 20 20 20 40 40 40 40 60 60 60 60 1480 1500 1520 1540 1480 1500 1520 1540 1480 1500 1520 1540 1480 1500 1520 1540 4 EOFs 5 EOFs 6 EOFs 7 EOFs 0 0 0 0 20 20 20 20 40 40 40 40 60 60 60 60 1480 1500 1520 1540 1480 1500 1520 1540 1480 1500 1520 1540 1480 1500 1520 1540 8 EOFs 9 EOFs 10 EOFs 11 EOFs 0 0 0 0 0 0 0 0 20 20 20 20 40 40 40 40 60 60 60 60 60 60 60 60 1480 1500 1520 1540 1480 1500 1520 1540 1480 1500 1520 1540 1480 1500 1520 1540 Chapman and Jiang ASA157, Portland, Oregon, 18 - 22 May 2009
Comparison of effective SSP for 1 km data Comparison of effective SSP for 1 km data Small SSP data set Large SSP data set Marginal distributions of SSPs Chapman and Jiang ASA157, Portland, Oregon, 18 - 22 May 2009
Inter-parameter correlations for 1 km data – EOF EOFs vs. geometric and geoacoustic parameters t i d ti t Small SSP data set Large SSP data set 2D Marginal distributions of EOFs Chapman and Jiang ASA157, Portland, Oregon, 18 - 22 May 2009
Results – comparison of Bayesian geoacoustic Inversion by using different SSP data set at 1 km Inversion by using different SSP data set at 1 km Small SSP set Large SSP set Large SSP set 1D Marginal distributions of geometric parameters Chapman and Jiang ASA157, Portland, Oregon, 18 - 22 May 2009
Results – comparison of Bayesian geoacoustic Inversion by using different SSP data set at 1 km Inversion by using different SSP data set at 1 km Small SSP set Large SSP set 1D Marginal distributions of geoacoustic parameters Chapman and Jiang ASA157, Portland, Oregon, 18 - 22 May 2009
Results – comparison of Bayesian geoacoustic p y g Inversion using small SSP data set at 1, 3 and 5 km Chapman and Jiang ASA157, Portland, Oregon, 18 - 22 May 2009
Breakdown – 3 km site, small SSP data set 2D Marginal distributions of EOFs with geometric parameters Chapman and Jiang ASA157, Portland, Oregon, 18 - 22 May 2009
Breakdown – 3 km site, large SSP data set 2D Marginal distributions of EOFs with geometric parameters Chapman and Jiang ASA157, Portland, Oregon, 18 - 22 May 2009
Conclusions: • Water column sound speed profile has significant effect on geometric parameters and therefore • affects geoacoustic parameter estimates • has great impact on matched field processing based has great impact on matched field processing based source localization • Geoacoustic parameter estimates using different SSP observations • Geoacoustic parameter estimates using different SSP observations are consistent with each other • for small SSP variations over the propagation path, the most relevant SSPs are more effective • for large SSP variations, single effective SSP may not be adequate q Chapman and Jiang ASA157, Portland, Oregon, 18 - 22 May 2009
Acknowledgments: • Office of Naval Research: for sponsoring the research • William Hodgkiss and Peter Gerstoft from MPL for providing the acoustic data • David Knobles from ARL for providing navigation and source depth data • John Goff from Institute of Geophysics, University of Texas at Austin for providing geophysical chirp seismic reflection data f idi h i l hi i i fl i d • Arthur Newhall from WHOI f for providing oceanographic observation data idi hi b ti d t Chapman and Jiang ASA157, Portland, Oregon, 18 - 22 May 2009
Recommend
More recommend
