Quantifying the uncertainty of geoacoustic parameter estimates for the New Jersey shelf by inverting air gun data.

ABSTRACT

This paper describes geoacoustic inversion of low frequency air gun data acquired during an experiment on the New Jersey shelf. Hybrid optimization and Bayesian inversion techniques based on matched field processing were applied to multiple shots from three air gun data sets recorded by a vertical line array in a long-range shallow water geometry. For the Bayesian inversions, full data error covariance matrix was estimated from a set of consecutive shots that had high temporal coherence and small spatial variation in source position. The effect of different data error information on the geoacoustic parameter uncertainty estimates was investigated by using the full data error covariance matrix, a diagonalized version of the full error covariance, and a diagonal matrix with identical variances. The comparison demonstrated that inversion using the full data error information provided the most reliable parameter uncertainty estimates. The inversions were highly sensitive to the near sea floor geoacoustic parameters, including sediment attenuation, of a simple single-layer geoacoustic model. The estimated parameter values of the model were consistent with depth averaged values (over wavelength scales) of a high resolution geoacoustic model developed from extensive ground truth information. The interpretation of the frequency dependence of the estimated attenuation is also discussed.