Geoacoustic inversion with strongly correlated data errors

Jan Dettmer, Stan E. Dosso, Charles W. Holland


Pre-processing of synthetic single bounce reflection-loss data with strongly correlated data errors to improve application of a nonlinear Bayesian inversion to recover geoacoustic parameters from a viscoelastic model was considered. Correlated Gaussian errors were generated using a realistic synthetic covariance matrix derived from experimental measurements. It was found that the data used contained eight frequencies in a band from 300 to 1600 Hz with different numbers of angles at each frequency. The correct treatment of error correlations for uncertainties estimation from the data was found to be essential.


Acoustic noise; Correlation methods; Data acquisition; Error analysis; Iterative methods; Matrix algebra; Natural frequencies; Optimization; Parameter estimation; Viscoelasticity; Cholesky decomposition; Covariance matrix; Data residuals; Geoacoustic inversions

Full Text:



  • There are currently no refbacks.