Inaccurate results from benchmark accurate ocean acoustic propagation models: It’s the implementation, not the model

Abstract

Despite numerous publications that document the benchmark accuracy of several ocean acoustic propagation models, some users of these models continue to get inaccurate results. The problem is not with the model, but with the implementation of the model. The most common mistake in implementing range-dependent acoustic models (e.g., parabolic equation models) is the selection of model parameters (e.g., step size, grid spacing, attenuating bottom depth). Another common mistake is lack of sufficient environmental data to the model. The number of implementation errors could be significantly reduced, if the user would abandon the erroneous belief that the model is a ‘‘black box’’ that will always produce the correct answer in a single run. Knowledgeable users methodically vary the input parameters and make repeated runs until model convergence is evident. It is possible to preselect model parameters, and couple the acoustic model with its environmental databases, so that a general user can obtain an accurate result. The U.S. Navy has done this with its Navy Standard Propagation models for nearly a decade. The models and their revisions are documented, configuration managed, and verified by applications to over 500 test cases. Selected examples will be presented. [Work supported by ONR/NRL.]