Matched-field source localization methods attempt to estimate the rang
e and depth of a source in an acoustic waveguide. These methods give g
ood results when the waveguide parameters are known precisely; however
, matched-field methods have been shown to be very sensitive to model
mismatch resulting from errors in the assumed environmental parameters
. Described in this paper is an approach which minimizes model mismatc
h, caused by uncertainty in the sound velocity profile, by jointly est
imating the environmental parameters and source location. An efficient
way to initialize the maximum-likelihood search by projecting the rec
eived data onto subspaces corresponding to regions in parameter space
is proposed. Monte Carlo simulation results in a shallow-water environ
ment are reported. (C) 1996 Acoustical Society of America.