Nicholas C. Makris
Naval Res. Lab., Washington D.C. 20375
A technique is under development to optimize experimental design for estimation of 3-D sound-speed structure by inversion of acoustic data. The motivation is to take advantage of a priori knowledge of invariant environmental parameters to estimate the minimum number of sensors necessary and their optimal deployment geometry for a well constrained inversion. First, static environmental information, such as bathymetry, geoacoustic parameters of the sediment, and mean sound-speed structure of the water column, is input to an appropriate range-dependent acoustic model for a given sensor deployment geometry. Next, a theoretical lower bound on estimation error for the water-column sound-speed structure is obtained via the Cramer--Rao bound. The deployment geometry is then perturbed until the estimation error is within acceptable bounds for oceanographic and acoustic modeling. However, the choice of sound-speed parametrization can also severely affect the accuracy of an inversion. For example, an empirical orthogonal function (EOF) representation typically has higher resolution for fewer parameters than a discrete cell representation. But this is at the cost of more limiting assumptions. These issues are addressed by computing the theoretical lower bound on estimation error for discrete cell, EOF, and Fourier internal wave representations of 3-D sound-speed structure.