Bayesian inversion of seabed reflection data

Stan E. Dosso, Charles W. Holland

Research output: Chapter in Book/Report/Conference proceedingChapter

Abstract

This paper applies nonlinear Bayesian inversion to seabed reflection data from two sites in the Strait of Sicily to estimate visco-elastic parameters of the upper-most sediments. At one site the seabed consists of fine-grained, low-velocity sediments, resulting in reflecti v-ity data (bottom loss versus grazing angle) with a well-defined angle of intromission. At the second site the seabed consists of high-velocity sediments, resulting in a critical angle. The Bayesian inversion provides maximum a posteriori parameter estimates with uncertainties quantified in terms of standard deviations and marginal probability distributions. Data uncertainties are quantified using several approaches, including analysis of exper-mental errors, maximum-likelihood estimation, and treating uncertainties as nuisance parameters in the Bayesian inversion. Statistical tests are applied to the data residuals to validate the assumed uncertainty distributions. Excellent results (i.e., small uncertainties) are obtained for sediment compressional velocity, compressional attenuation, and density; shear parameters are poorly determined although low shear-wave velocities are indicated. The Bayesian analysis provides a quantitative comparison of inversion results for the two sites, and indicates that the geoacoustic information content is significantly higher for angle-of- intromission data.

Original languageEnglish (US)
Title of host publicationAcoustic Sensing Techniques for the Shallow Water Environment
Subtitle of host publicationInversion Methods and Experiments
PublisherSpringer Netherlands
Pages17-27
Number of pages11
ISBN (Print)1402043724, 9781402043727
DOIs
StatePublished - 2006

All Science Journal Classification (ASJC) codes

  • Earth and Planetary Sciences(all)
  • Environmental Science(all)

Fingerprint Dive into the research topics of 'Bayesian inversion of seabed reflection data'. Together they form a unique fingerprint.

Cite this