Abstract:

When modeling ocean-acoustic systems, the environmental information necessary to compute the acoustic field properties is assumed to be accurately specified by various parameters, fields, and boundary conditions. However, in real world applications, these quantities are subject to uncertainties due to our incomplete knowledge of the waveguide environment. This form of uncertainty involves errors that are quite distinct from the numerical errors that can arise when a mathematical model is discretized, implemented on a computer, and solved with finite precision arithmetic. In effect, the environmental uncertainty introduces spurious degrees of freedom into the system. In order to make reliable simulation-based predictions, this uncertainty needs to be quantified and incorporated in the simulation process itself. The idea of embedding uncertainty into the simulation framework and elevating its status to a subject worth studying on its own merits represents a paradigm shift that has stimulated research in several disciplines. This talk will overview recent approaches to this problem in an ocean-acoustic context and give examples of computations that have incorporated environmental uncertainty in the numerical computation of acoustic field properties. [Work supported by the Office of Naval Research.]