Dynamically Orthogonal Reduced-Order Models for Stochastic Underwater Sound Propagation

Accurate and efficient acoustic fields' predictions are ubiquitous for numerous applications of underwater communication and sensing. Yet, the task of modeling underwater sound propagation remains challenging due to the physical complexities arising from the multi-scale ocean-acoustic effects and typically large domains of interest. In addition, the incomplete knowledge of the ocean environment, and specifically the uncertainties associated with the ocean state as well as the topography and seabed properties, lead to significant uncertainties in the acoustic predictions. In this work, we present a rigorous, accurate, and efficient technique to quantify the effect of such environmental uncertainties, and predict the propagating stochastic acoustic waves and their underlying probability distributions. This technique uses a dynamic model-order reduction and uncertainty quantification methodology, the Dynamically Orthogonal (DO) equations, to efficiently model the environmental uncertainties and predict the stochastic acoustic waves. We apply this technique to model sound propagation in an uncertain ocean-seabed environment using wide-angle parabolic equation and ray-based methods. We then showcase this technique in idealized and realistic ocean cases.