TOMOGRAPHIC INVERSION FOR GEOACOUSTIC PARAMETERS IN SHALLOW WATER
This paper examines the linearized tomographic inversion of simulated data for a shallow water, multi-array, multi-source scenario. The environments represented include simulations of (1) highly idealized constant regions as well as (2) the Haro Strait Test of June 1996 which displays range, depth, and azimuthal variability, i.e., 3-D dependence on environmental parameters where these parameters can include water depths and multiple sediment sound-speed profiles, densities, depths, and attenuations. This tomographic inversion method is independent of the number of parameters to be determined. However, the method does assume that some inversion method (such as RIGS, simulated annealing, genetic algorithms, etc.) has already estimated range-independent average source-to-receiver environmental parameters. These average parameters are then input into the tomographic inversion which relies on a matrix of path-cell distances. The matrix condition number, Λ, is a determining feature for the inversion accuracy where Λ is a function of source and receiver distributions and their subsequent path distances through the region cells. Additionally, the accuracy of the input estimates for the average geoacoustic properties is also an important factor in the final 3-D tomographic inversion accuracy. Results using this (linearized) tomography inversion method show a potential for excellent error estimates (much less than 1%) for the environmental parameters assuming exact, idealized input values. Errors are still quite reasonable (well under 10%) if more realistic, i.e., erroneous, input values are assumed. This paper will conclude with a discussion of upcoming future directions.