3aUW8. Source localization bias due to forward model uncertainties in matched-field processing.

Session: Wednesday Morning, December 3


Author: N. Ross Chapman
Location: School of Earth and Ocean Sci., Univ. of Victoria, P.O. Box 3055, Victoria, BC V8W 3P6, Canada
Author: Ronald T. Kessel
Location: School of Earth and Ocean Sci., Univ. of Victoria, P.O. Box 3055, Victoria, BC V8W 3P6, Canada

Abstract:

Matched-field source localization (MFSL) in range-dependent media requires fast realistic forward models of sound propagation. No model replicates the real environment in all its detail, however, which effectively means that the source is being localized in the wrong environment, at a position that may not agree with the actual. As a rule, the agreement improves as the model is made more realistic, but this usually increases the computation time and complexity of the model, especially in range-dependent environments. MFSL must, therefore, balance accuracy and speed against model realism. The objective here is to demonstrate and account for consistent errors in the estimated range and depth of a source when the forward model differs from the actual environment. Range bias is demonstrated for data collected in the Pacific Shelf sea trials, for example, when bathymetric variation was at first ignored for speed and simplicity in the MFSL analysis, and then taken into account using a simple bias rule. Range and depth biases due to incorrectly modeled sediment properties and bathymetry in range-dependent media are explored more generally using the theory of adiabatic modes and numerical simulations. [Work supported by ONR and Department of National Defence, Canada.]


ASA 134th Meeting - San Diego CA, December 1997