Inversion of bottom/subbottom statistical parameters from acoustic backscatter data

Inversion of statistical parameters of a bottom/subbottom scattering model is investigated by using genetic algorithms for both synthetic and real data. The bottom/subbottom scattering model used in the calculations is a modified version of Lyons, Anderson, and Dwan [J. Acoust. Soc. Am. 79, 1410–1422 (1986)] in which correlation between subbottom density, compressibility, and sound-speed fluctuations is established through Wood’s equation [A Textbook of Sound (Macmillan, New York, 1941)], and volume inhomogeneities are described by von Karman type autocorrelation functions [T. von Karman, J. Mar. Res. 7, 252–264 (1948)]. The inversion is posed as an optimization problem which is solved (by a controlled Monte Carlo search using genetic algorithms) to find the optimum set of statistical parameters that minimizes the quadratic deviation between measured and calculated backscatter data. A posteriori probabilities calculated at the end of the search are used for error estimation and indication of relative importance of model parameters. This information helps to further assess the relative importance of two major scattering mechanisms due to bottom roughness and subbottom inhomogeneities. Such assessment is successfully demonstrated by using synthetic and real backscatter data for sandy, silty, and muddy sediments. Finally, inverted statistical parameters for a sandy site at Biscayne Bay, Florida confirm the results of simultaneous tomographic measurements indicating that scattering by subbottom inhomogeneities plays a minor role for this particular site.