Geoacoustic model for the New Jersey Shelf by inverting airgun data

Abstract

This paper describes geoacoustic inversion of airgun data acquired during the SWARM95 experiment. Hybrid optimization and Bayesian inversion techniques were applied to three airgun data sets recorded by a vertical line array. Optimization results are used to show the consistency of the estimates from all of the shots in terms of histograms and standard deviations of the inverted geoacoustic model parameters. The inversion results from the Bayesian approach are used to show the uncertainties of the estimates in terms of marginal distributions, MAP estimates, and credibility intervals. In the Bayesian inversion, full data error covariance matrices were estimated by ensemble averaging the covariance of the residuals of the measured and modeled data of inversions from many shots. The numbers of shots in the ensemble averages were determined by checking the temporal coherence of the signal. Statistical tests were used to test the validity of the assumptions in the Bayesian approach after incorporating full data error covariance matrices. With these inversion techniques, equivalent geoacoustic models with/without shear wave estimates are extracted for this experimental site. The frequency dependence of the p-wave attenuation, and the correlation between the geoacoustic parameters are obtained from the inversion results. [Work supported by ONR.]