Abstract:
Nonuniqueness is an insidious problem for all inversion methods, and it appears to be particularly virulent for the shallow water geoacoustic applications. For such scenarios the combination of many unknowns (from a few to many dozens) plus the nature of the solution space (which can be highly nonconvex and rippled) results in potentially untenable problems. This work will discuss MFP efforts via simulated data to address (1) the hunt for the ``true'' solution, (2) the suppression of false solutions via broadband processing (incoherent as well as coherent) and via multiple, distributed sources, and (3) the evolution of problem difficulty with increases in the number of problem parameters and with errors in ``known'' parameters. [Work supported by ONR.]